From: KOTH Subject: KOTH.ORG: Status - 94 No Pspace 07/03/06 Date: Mon, 3 Jul 2006 00:25:57 -0400 (EDT) Message-ID: <200607030409.k634906e022740@asgard.t-b-o-h.net> Weekly Status on 07/03/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG 94 No Pspace CoreWar Hill: Last battle concluded at : Sun Jul 2 07:12:49 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 43/ 50/ 7 s-test Sascha Zapf 137 3 2 40/ 45/ 16 Pro Radii inversed 135 10 3 30/ 25/ 46 Monster_Human_Grunt inversed 134 265 4 25/ 18/ 57 8kSquare Roy van Rijn 132 1 5 40/ 48/ 12 Hallucination Scanner inversed 131 138 6 29/ 27/ 44 Hullab3loo Roy van Rijn 131 180 7 35/ 39/ 26 Twilight S.Fernandes 131 150 8 28/ 24/ 48 Monster_Alien_Grunt inversed 131 266 9 27/ 23/ 51 Bluebell Christian Schmidt 130 153 10 26/ 22/ 52 Song of the blue sea Miz 130 137 11 35/ 41/ 24 Gabble Christian Schmidt 130 31 12 25/ 21/ 55 Last Judgement Christian Schmidt 129 450 13 23/ 18/ 59 Burning Metal inversed 127 5 14 28/ 29/ 44 N e i t h inversed 127 26 15 22/ 17/ 62 dd-test Sascha Zapf 127 14 16 22/ 17/ 61 D3vilstick Roy van Rijn 126 179 17 36/ 45/ 19 ChimeraQueen Nenad Tomasev 126 145 18 24/ 23/ 53 MoonOfChaos Nenad Tomasev 126 228 19 25/ 24/ 51 Amber inversed 126 4 20 24/ 22/ 55 Raging Gale inversed 125 6 21 12/ 17/ 70 f-test Fluffy 108 0 From: KOTH Subject: KOTH.ORG: Status - MultiWarrior 94 07/03/06 Date: Mon, 3 Jul 2006 00:26:09 -0400 (EDT) Message-ID: <200607030403.k63430NS022281@asgard.t-b-o-h.net> Weekly Status on 07/03/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Multiwarrior 94 CoreWar Hill: Last battle concluded at : Thu Jun 8 05:01:40 EDT 2006 # Name Author Score Age 1 Urgle Daniel Rivas 43 7 2 CLP-shot again G.Labarga 38 1 3 kingdom of the grasshoppe simon wainwright 34 122 4 simply believe John Metcalf 29 4 5 JustADirtyClearTest Nenad Tomasev 28 56 6 nameless fragment S.Fernandes 27 26 7 Fluffy Paper VI Jens Gutzeit 25 27 8 Diptera Nenad Tomasev 24 41 9 rooftop pursuit John Metcalf 23 3 10 the price of hostility John Metcalf 21 14 11 the watching eye John Metcalf 20 0 From: KOTH Subject: KOTH.ORG: Status - Standard 07/03/06 Date: Mon, 3 Jul 2006 00:26:10 -0400 (EDT) Message-ID: <200607030400.k63400Nq022058@asgard.t-b-o-h.net> Weekly Status on 07/03/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Standard KotH CoreWar Hill : Last battle concluded at : Sun Jul 2 08:28:00 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 39/ 24/ 37 Test Alexander (Sasha) Wa 154 190 2 40/ 29/ 30 The Next Step '88 David Houston 151 66 3 42/ 35/ 23 My 1st try Christian Schmidt 148 82 4 35/ 23/ 41 test G.Labarga 147 33 5 35/ 24/ 40 Guardian Ian Oversby 147 250 6 36/ 25/ 39 The Hurricaner G.Labarga 146 37 7 43/ 40/ 18 Scan the Can Christian Schmidt 145 31 8 37/ 29/ 35 SoundOfDarkness Nenad Tomasev 145 14 9 44/ 45/ 11 Blacksoil inversed 144 1 10 44/ 44/ 11 Speeed 88mph Christian Schmidt 144 50 11 40/ 38/ 22 Moonwipe Christian Schmidt 142 47 12 42/ 42/ 16 July Nenad Tomasev 141 23 13 37/ 33/ 31 The Seed Roy van Rijn 140 68 14 28/ 17/ 55 Utterer '88 Christian Schmidt 140 25 15 39/ 39/ 21 Hexamorph inversed 139 4 16 31/ 25/ 44 Scopulos pluviae G.Labarga 137 17 17 42/ 47/ 11 Scan Test C 6 Steve Gunnell 137 89 18 40/ 43/ 16 war in the name of music John Metcalf 137 9 19 34/ 32/ 34 A.I.P. Christian Schmidt 137 58 20 43/ 49/ 8 Replihater Some Redcoder 136 21 21 2/ 82/ 16 kensi K.V. Andronov 21 0 From: KOTH Subject: KOTH.ORG: Status - ICWS Experimental 94 07/03/06 Date: Mon, 3 Jul 2006 00:26:08 -0400 (EDT) Message-ID: <200607030406.k63460ir022453@asgard.t-b-o-h.net> Weekly Status on 07/03/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG ICWS Experimental 94 CoreWar Hill: Last battle concluded at : Sat May 27 16:29:52 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 46/ 35/ 19 Fatamorgana X Zul Nadzri 157 10 2 44/ 36/ 20 The X Machine Zul Nadzri 152 34 3 44/ 36/ 19 Ogre Christian Schmidt 152 171 4 37/ 22/ 42 xd100 test David Houston 152 20 5 28/ 11/ 61 Evol Cap 4 X John Wilkinson 146 292 6 42/ 40/ 18 Eliminator X Zul Nadzri 145 35 7 41/ 39/ 20 Bewitching S.Fernandes 144 1 8 34/ 25/ 41 KAT v5 Dave Hillis 142 155 9 39/ 37/ 24 Trefoil F 13 Steve Gunnell 140 106 10 35/ 29/ 36 Olivia X Ben Ford 140 104 11 41/ 43/ 16 Giant Hazy Test 13 Steve Gunnell 140 50 12 41/ 43/ 15 O_Fortuna3X Nenad Tomasev 139 6 13 40/ 42/ 18 Black Moods Ian Oversby 139 219 14 38/ 38/ 23 Trefoil Test F 14 Steve Gunnell 139 7 15 37/ 37/ 26 test Some Redcoder 138 4 16 21/ 6/ 73 Evolve X v4.0 John Wilkinson 137 240 17 38/ 41/ 21 Controlled Aggression Ian Oversby 135 223 18 31/ 28/ 41 Glenstorm John Metcalf 135 85 19 38/ 42/ 20 Simply Intelligent Zul Nadzri 134 16 20 39/ 46/ 14 Fatal Choice Some Redcoder 133 5 21 8/ 79/ 12 Reaper II Slaytanist 38 0 From: "Roy" Subject: Multi process qscan Date: 7 Jul 2006 11:41:32 -0700 Message-ID: <1152297692.719272.280090@h48g2000cwc.googlegroups.com> Hi everybody, This evening Fluffy was talking to me about his qscan article and he asked if I had any odd ideas he could explore. Thats when I suggested a multi process qscanner. The big advantage is that is can withstand simple dat attacks in the early phase. After a lot of tinkering with the processes I came up with the following design I'm quite proud of. It splits the qscan in half and scans the upper and lower half seperatly.... OOPS! just realized something, the following warrior's qscan is probably broken, the qz mutation isn't done in the right way... I hate it I don't fully understand qscans :( But well, someone can probably fix that for me...! The point is my following qscan is just simple cut in two and has a smart booting mechanism for multi-process warriors. If the upper of lower half of the qscan sees something or gets bombed the other part still starts the warrior normally! Without further bad english and blabla, the warrior: just look at how it works :-) Its just Maelstrom (one of my favourites) with the SplitQScan. ;redcode-94nop ;name ;author Roy van Rijn ;assert 1 zero equ qbomb qtab3 equ qbomb qbomb dat >qoff , >qc2 hDist equ 788 iDist equ hDist-1520 cDist equ hDist-3634 mov.i 2 , 3 ;might move 2,3 or 3,3 depends on qEnd1 and qEnd2 sBoot spl 1 , {qb1 qtab2 spl 1 , -1 pStone spl #0 mov bomb , >ptr add.x imp , ptr ptr jmp imp-iStep*8 , >sStep-6 bomb dat >1 , }1 imp mov.i #sStep-1 , iStep for 3 dat 0 , 0 rof nstep1 equ 2413 cstep1 equ 4704 tstep1 equ 3278 pap spl @8 , }tstep1 mov.i }-1 , >-1 nothA spl cstep1 , 0 mov.i >-1 , }-1 nothB spl @0 , }nstep1 mov.i }-1 , >-1 mov.i #1138 , <1 djn.b -2 , #1618 for 25 dat 0 , 0 rof qc2 equ ((1+(qtab3-qptr)*qy)%CORESIZE) qb1 equ ((1+(qtab2-1-qptr)*qy)%CORESIZE) qb2 equ ((1+(qtab2-qptr)*qy)%CORESIZE) qb3 equ ((1+(qtab2+1-qptr)*qy)%CORESIZE) qa1 equ ((1+(qtab1-1-qptr)*qy)%CORESIZE) qa2 equ ((1+(qtab1-qptr)*qy)%CORESIZE) qz equ 2108 qy equ 243 ;qy*(qz-1)=1 ;q0 mutation qgo spl qGo2 sne qptr+qz*qa1 , qptr+qz*qa1+qb2 seq <(qtab1-1) , qptr+qz*(qa1-1)+qb2 djn.a q0 , {q0 sne qptr+qz*qc2 , qptr+qz*qc2+qb2 seq qEnd2 qGo2 sne qptr+qz*qb1 , qptr+qz*qb1+qb1 seq <(qtab2-1) , qptr+qz*(qb1-1)+(qb1-1) jmp q0 , {q1 sne qptr+qz*qb2 , qptr+qz*qb2+qb2 seq qptr , qptr+qz+(qb2-1) jmp q2 , Subject: Re: Multi process qscan Date: 7 Jul 2006 12:45:35 -0700 Message-ID: <1152301534.957802.233330@s13g2000cwa.googlegroups.com> I've switched it around a little bit and added the final jmz line too (I deleted it the first time). Now the qscan should work a little more like it it supposed to do. But still not good enough... I'll just call it a PoC (Proof of Concept) for now ;-) The problem is this: sne/seq/jmp 2x becomes: sne sne seq seq jmp jmp But in the second 'seq' lines there are mutations :( So the qscanner isn't working properly. What we really need (Fluffy pointed that out) are two seperate scanlines with one common bomb-engine. Instead of the q4^5 then we can have all sne/seq/jmp instead of seq/jmp's at the end. But I'll just leave that as a challenge because I'm way to lazy for that right now :P The improved PoC: ;redcode-94nop ;name PoC q4.5/2 ;author Roy van Rijn ;assert 1 zero equ qbomb qtab3 equ qbomb qbomb dat >qoff , >qc2 hDist equ 788 iDist equ hDist-1520 cDist equ hDist-3634 mov.i 2 , 3 ;might move 2,3 or 3,3 depends on qEnd1 and qEnd2 sBoot spl 1 , {qb1 qtab2 spl 1 , -1 pStone spl #0 mov bomb , >ptr add.x imp , ptr ptr jmp imp-iStep*8 , >sStep-6 bomb dat >1 , }1 imp mov.i #sStep-1 , iStep for 3 dat 0 , 0 rof nstep1 equ 2413 cstep1 equ 4704 tstep1 equ 3278 pap spl @8 , }tstep1 mov.i }-1 , >-1 nothA spl cstep1 , 0 mov.i >-1 , }-1 nothB spl @0 , }nstep1 mov.i }-1 , >-1 mov.i #1138 , <1 djn.b -2 , #1618 for 24 dat 0 , 0 rof qc2 equ ((1+(qtab3-qptr)*qy)%CORESIZE) qb1 equ ((1+(qtab2-1-qptr)*qy)%CORESIZE) qb2 equ ((1+(qtab2-qptr)*qy)%CORESIZE) qb3 equ ((1+(qtab2+1-qptr)*qy)%CORESIZE) qa1 equ ((1+(qtab1-1-qptr)*qy)%CORESIZE) qa2 equ ((1+(qtab1-qptr)*qy)%CORESIZE) qz equ 2108 qy equ 243 ;qy*(qz-1)=1 ;q0 mutation qgo spl qGo2 sne qptr+qz*qa1 , qptr+qz*qa1+qb2 seq <(qtab1-1) , qptr+qz*(qa1-1)+qb2 djn.a q0 , {q0 sne qptr+qz*qa2 , qptr+qz*qa2+qb2 seq qptr , qptr+qz+(qb2-1) jmp q2 , qEnd1 qoff equ -87 qstep equ -7 qtime equ 19 q0 mul.b *2 , qptr q2 sne {qtab1 , @qptr q1 add.b qtab2 , qptr mov qtab3 , @qptr qptr mov qbomb , }qz sub #qstep , qptr djn -3 , #qtime jmn.a sBoot-1 , pap2 end qgo From: =?iso-8859-2?Q?=A3ukasz_Grabu=F1?= Subject: [OT] Ant wars Date: Sat, 8 Jul 2006 07:53:33 +0200 Message-ID: <3c7mzyfbpwqo$.dlg@grabun.com> A bit OT, I guess. Have a look at: http://ant-wars.net/ Quote from the site: "Ant Wars is a competition which pits clever programs against each other to do battle and compete for food in virtual worlds. Each contestant is a species of ant, which can visualize only the world immediately around him and pheromones left by fellow and enemy ants. Using this information, the ant brain (a simple state machine) must guide the ant towards collecting food at his home ant hill, while fending off or attacking enemies." -- �ukasz Grabu�, http://www.grabun.com/ From: KOTH Subject: KOTH.ORG: Status - Standard 07/10/06 Date: Mon, 10 Jul 2006 15:54:22 -0400 (EDT) Message-ID: <200607100400.k6A400SL002789@asgard.t-b-o-h.net> Weekly Status on 07/10/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Standard KotH CoreWar Hill : Last battle concluded at : Tue Jul 4 16:52:51 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 36/ 25/ 39 Test Alexander (Sasha) Wa 148 193 2 38/ 30/ 32 The Next Step '88 David Houston 146 69 3 40/ 36/ 25 My 1st try Christian Schmidt 144 85 4 33/ 26/ 40 The Hurricaner G.Labarga 141 40 5 33/ 26/ 41 test G.Labarga 140 36 6 33/ 26/ 41 Guardian Ian Oversby 140 253 7 34/ 29/ 37 SoundOfDarkness Nenad Tomasev 139 17 8 40/ 41/ 20 Scan the Can Christian Schmidt 139 34 9 31/ 28/ 46 Scopulos pluviae G.Labarga 138 20 10 41/ 46/ 12 Speeed 88mph Christian Schmidt 136 53 11 42/ 48/ 10 Monolith inversed 135 2 12 26/ 18/ 56 Utterer '88 Christian Schmidt 135 28 13 37/ 39/ 24 Moonwipe Christian Schmidt 135 50 14 38/ 47/ 20 July Nenad Tomasev 133 26 15 39/ 44/ 17 Hexamorph inversed 133 1 16 34/ 34/ 32 The Seed Roy van Rijn 133 71 17 24/ 16/ 60 Raging Gale '88 inversed 133 3 18 41/ 55/ 9 Replihater Some Redcoder 132 24 19 39/ 48/ 12 Scan Test C 6 Steve Gunnell 130 92 20 35/ 45/ 20 war in the name of music John Metcalf 126 12 21 2/ 1/ 2 Hexamorph inversed 6 7 From: KOTH Subject: KOTH.ORG: Status - MultiWarrior 94 07/10/06 Date: Mon, 10 Jul 2006 15:54:21 -0400 (EDT) Message-ID: <200607100403.k6A430BZ002963@asgard.t-b-o-h.net> Weekly Status on 07/10/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Multiwarrior 94 CoreWar Hill: Last battle concluded at : Thu Jun 8 05:01:40 EDT 2006 # Name Author Score Age 1 Urgle Daniel Rivas 43 7 2 CLP-shot again G.Labarga 38 1 3 kingdom of the grasshoppe simon wainwright 34 122 4 simply believe John Metcalf 29 4 5 JustADirtyClearTest Nenad Tomasev 28 56 6 nameless fragment S.Fernandes 27 26 7 Fluffy Paper VI Jens Gutzeit 25 27 8 Diptera Nenad Tomasev 24 41 9 rooftop pursuit John Metcalf 23 3 10 the price of hostility John Metcalf 21 14 11 the watching eye John Metcalf 20 0 From: KOTH Subject: KOTH.ORG: Status - ICWS Experimental 94 07/10/06 Date: Mon, 10 Jul 2006 15:54:20 -0400 (EDT) Message-ID: <200607100406.k6A460nL003242@asgard.t-b-o-h.net> Weekly Status on 07/10/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG ICWS Experimental 94 CoreWar Hill: Last battle concluded at : Sat May 27 16:29:52 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 46/ 35/ 19 Fatamorgana X Zul Nadzri 157 10 2 44/ 36/ 20 The X Machine Zul Nadzri 152 34 3 44/ 36/ 19 Ogre Christian Schmidt 152 171 4 37/ 22/ 42 xd100 test David Houston 152 20 5 28/ 11/ 61 Evol Cap 4 X John Wilkinson 146 292 6 42/ 40/ 18 Eliminator X Zul Nadzri 145 35 7 41/ 39/ 20 Bewitching S.Fernandes 144 1 8 34/ 25/ 41 KAT v5 Dave Hillis 142 155 9 39/ 37/ 24 Trefoil F 13 Steve Gunnell 140 106 10 35/ 29/ 36 Olivia X Ben Ford 140 104 11 41/ 43/ 16 Giant Hazy Test 13 Steve Gunnell 140 50 12 41/ 43/ 15 O_Fortuna3X Nenad Tomasev 139 6 13 40/ 42/ 18 Black Moods Ian Oversby 139 219 14 38/ 38/ 23 Trefoil Test F 14 Steve Gunnell 139 7 15 37/ 37/ 26 test Some Redcoder 138 4 16 21/ 6/ 73 Evolve X v4.0 John Wilkinson 137 240 17 38/ 41/ 21 Controlled Aggression Ian Oversby 135 223 18 31/ 28/ 41 Glenstorm John Metcalf 135 85 19 38/ 42/ 20 Simply Intelligent Zul Nadzri 134 16 20 39/ 46/ 14 Fatal Choice Some Redcoder 133 5 21 8/ 79/ 12 Reaper II Slaytanist 38 0 From: KOTH Subject: KOTH.ORG: Status - 94 No Pspace 07/10/06 Date: Mon, 10 Jul 2006 15:54:19 -0400 (EDT) Message-ID: <200607100409.k6A490ur003393@asgard.t-b-o-h.net> Weekly Status on 07/10/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG 94 No Pspace CoreWar Hill: Last battle concluded at : Sun Jul 9 10:23:33 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 29/ 19/ 52 8kSquare Roy van Rijn 138 2 2 44/ 50/ 6 s-test Sascha Zapf 137 4 3 31/ 26/ 43 Monster_Human_Grunt inversed 136 266 4 32/ 28/ 40 Amicable Antagonist Roy van Rijn 135 1 5 32/ 29/ 40 Hullab3loo Roy van Rijn 134 181 6 30/ 27/ 43 Monster_Alien_Grunt inversed 134 267 7 40/ 46/ 15 Pro Radii inversed 133 11 8 28/ 24/ 48 Song of the blue sea Miz 132 138 9 27/ 22/ 52 Last Judgement Christian Schmidt 132 451 10 26/ 20/ 55 Burning Metal inversed 132 6 11 36/ 41/ 22 Gabble Christian Schmidt 131 32 12 27/ 24/ 49 MoonOfChaos Nenad Tomasev 131 229 13 28/ 25/ 48 Bluebell Christian Schmidt 131 154 14 36/ 40/ 24 Twilight S.Fernandes 131 151 15 30/ 31/ 39 N e i t h inversed 130 27 16 24/ 19/ 57 D3vilstick Roy van Rijn 130 180 17 38/ 46/ 17 ChimeraQueen Nenad Tomasev 130 146 18 24/ 18/ 59 dd-test Sascha Zapf 129 15 19 28/ 26/ 46 Amber inversed 129 5 20 39/ 50/ 11 Hallucination Scanner inversed 128 139 21 35/ 50/ 15 pursuit of the ducklings John Metcalf 119 0 From: Tuc at Beach House Subject: 88 Hill issues Date: Mon, 10 Jul 2006 20:51:29 -0400 (EDT) Message-ID: <200607110042.k6B0gQ1k002072@himinbjorg.tucs-beachin-obx-house.com> Hi all, It seems there have been some issues on the 88 Hill. I recreated the stats of warriors 17 and 18, and I hope it will begin to work again. Please let me know. Thanks, Tuc From: "iapdk@admin.drake.edu" Subject: Re: Multi process qscan Date: 11 Jul 2006 15:32:35 -0700 Message-ID: <1152657155.426116.133710@m79g2000cwm.googlegroups.com> Roy wrote: > Hi everybody, > > This evening Fluffy was talking to me about his qscan article and he > asked if I had any odd ideas he could explore. Thats when I suggested a > multi process qscanner. The big advantage is that is can withstand > simple dat attacks in the early phase. Now that would be cheating on all the other qscans that depend on those early shots for their high win scores. When QuickFreeze was written it was obvious that the scan component was giving up unnecessary losses because it was so big, so it got broken up into two separate scans. Today's qscans are very compact, but still give up a few particularly to the other qscans. Paul K From: "Roy" Subject: Re: Multi process qscan Date: 11 Jul 2006 23:52:46 -0700 Message-ID: <1152687166.095865.180290@m73g2000cwd.googlegroups.com> Hehe thats not cheating, thats trying to get some more wins! But there are a couple of problems with it: 1.It doesn't beat normal qscans (yet?) 2. When everybody uses it is becomes more useless tieing more instead of win/loss (win/loss at 50/50 is much better for the scores) So if it would work better it only works against normal qscans. iapdk@admin.drake.edu wrote: > Roy wrote: > > Hi everybody, > > > > This evening Fluffy was talking to me about his qscan article and he > > asked if I had any odd ideas he could explore. Thats when I suggested a > > multi process qscanner. The big advantage is that is can withstand > > simple dat attacks in the early phase. > > Now that would be cheating on all the other qscans that depend on those > early shots for their high win scores. > > When QuickFreeze was written it was obvious that the scan component was > giving up unnecessary losses because it was so big, so it got broken up > into two separate scans. Today's qscans are very compact, but still > give up a few particularly to the other qscans. > > Paul K From: Tuc at Beach House Subject: IRC.KOTH.ORG down Date: Thu, 13 Jul 2006 23:35:22 -0400 (EDT) Message-ID: <200607140322.k6E3MBqg007239@himinbjorg.tucs-beachin-obx-house.com> Hi, The webserver running irc.koth.org went down, I'm looking to try to get it back up remotely. Unfortunately I am not in a position to go on site so it might be down until I can get someone to. Tuc From: Tuc at Beach House Subject: IRC.KOTH.ORG up Date: Fri, 14 Jul 2006 00:31:22 -0400 (EDT) Message-ID: <200607140403.k6E43AfO009904@himinbjorg.tucs-beachin-obx-house.com> Hi, The system fortunately rebooted, but with an issue. I've fixed it and it should be fine again. Thanks, Tuc From: Tuc at Beach House Subject: IRC down again Date: Fri, 14 Jul 2006 01:40:09 -0400 (EDT) Message-ID: <200607140532.k6E5WfgA001095@himinbjorg.tucs-beachin-obx-house.com> Hi, Looks like I spoke too soon previously. The system is down yet again, and appears hard down. Tuc From: =?ISO-8859-1?Q?Tuc=20at=20Beach=20House?= Subject: IRC status Date: Fri, 14 Jul 2006 10:25:26 -0400 (EDT) Message-ID: <200607141411.k6EEBmxE047889@vjofn.tucs-beachin-obx-house.com> The server came up again but went back down. I can not diagnose remotely, and the local staff doesn't have knowledge to do it. It will need to wait for me to be on site. I will do my best to get there soon. Tuc -- Sent from my Treo From: Tuc at Beach House Subject: Temporary IRC server Date: Sun, 16 Jul 2006 11:52:29 -0400 (EDT) Message-ID: <200607161536.k6GFainQ012902@himinbjorg.tucs-beachin-obx-house.com> Hi, The parts to fix the down'd server didn't arrive, so I've set up a temporary IRC server. There is no logging, sorry. irc2.koth.org If your DNS servers haven't refreshed, you can use : 192.136.64.204 As always... Please contact me if necessary. Tuc From: Tuc at Beach House Subject: Just wanna say Date: Sun, 16 Jul 2006 11:56:31 -0400 (EDT) Message-ID: <200607161546.k6GFkEVW013091@himinbjorg.tucs-beachin-obx-house.com> Hi, Just wanted to say I'm sorry about all the recent KOTH.ORG issues. 88 hill - It appears that somehow 2 stats files were corrupt, which caused WEB and potentially hill issues 94NOP - It seemed to be COMPLETELY crashed, I had to copy a backup file in. IRC - The server that runs IRC on rebooted, then crashed/died completely. I believe the 88/94NOP were a result of the maintenance window to fix the file server. I thought it would have been able to handle it, but I was wrong. In the future I will make sure that the hills are shut down while I do this. I haven't completely decided what to do about the IRC server. I was only running it on one of the servers because when I decided what server would do what, Frigga got the IRC server and Odin got the hills. As for the future plans, I don't know that I really have any. Had Odin gone down, I would have brought the hills up on Frigga immediately. I realize that the IRC is important to everyone, but not "mission critical". As an aside, I don't feel like I'm getting much feedback on things. Maybe you do it all through JKW, but I'd like to know whats going on and what your looking for, etc. I can't promise anything, especially since now I'm busier than ever, but I'd like to know what people are thinking and wanting. Tuc From: Philip Kendall Subject: Re: Just wanna say Date: 16 Jul 2006 17:16:22 GMT Message-ID: On 2006-07-16, Tuc at Beach House wrote: > > As an aside, I don't feel like I'm getting much feedback on things. The first thing I learned as a (l)user: don't hassle admins when things are going wrong :-) Cheers, Phil -- Philip Kendall http://www.srcf.ucam.org/~pak21/ From: "Roy" Subject: Re: Just wanna say Date: 16 Jul 2006 23:44:26 -0700 Message-ID: <1153118666.126167.309490@m79g2000cwm.googlegroups.com> Yeah, we just give you the time/space to fix things. Its a hobby to all of us, so we don't want to rush you nor anybody else. And don't forget there are a lot of scriptkiddies here who might be able to help you with stuff, so if you are having a specific problem, you can always post it here..! Roy Philip Kendall wrote: > On 2006-07-16, Tuc at Beach House wrote: > > > > As an aside, I don't feel like I'm getting much feedback on things. > > The first thing I learned as a (l)user: don't hassle admins when things > are going wrong :-) > > Cheers, > > Phil > > -- > Philip Kendall > http://www.srcf.ucam.org/~pak21/ From: KOTH Subject: KOTH.ORG: Status - Standard 07/17/06 Date: Mon, 17 Jul 2006 11:42:17 -0400 (EDT) Message-ID: <200607170400.k6H400dH041704@asgard.t-b-o-h.net> Weekly Status on 07/17/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Standard KotH CoreWar Hill : Last battle concluded at : Sun Jul 16 17:12:18 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 38/ 24/ 38 Test Alexander (Sasha) Wa 152 201 2 35/ 19/ 46 Raging Gale '88 inversed 152 6 3 39/ 29/ 31 The Next Step '88 David Houston 150 77 4 40/ 31/ 29 White Fire inversed 149 3 5 36/ 26/ 37 Triangular Sun inversed 147 2 6 43/ 40/ 17 Hexamorph inversed 145 5 7 41/ 36/ 23 My 1st try Christian Schmidt 145 93 8 44/ 44/ 12 Monolith inversed 145 10 9 35/ 25/ 40 The Hurricaner G.Labarga 144 48 10 41/ 40/ 19 Scan the Can Christian Schmidt 143 42 11 33/ 25/ 42 test G.Labarga 141 44 12 28/ 16/ 55 Utterer '88 Christian Schmidt 140 36 13 33/ 26/ 41 Guardian Ian Oversby 140 261 14 34/ 29/ 37 SoundOfDarkness Nenad Tomasev 139 25 15 36/ 34/ 30 Vampire Knight inversed 138 1 16 35/ 33/ 32 The Seed Roy van Rijn 138 79 17 38/ 39/ 23 Moonwipe Christian Schmidt 137 58 18 41/ 46/ 13 Scan Test C 6 Steve Gunnell 137 100 19 42/ 47/ 12 Speeed 88mph Christian Schmidt 136 61 20 38/ 41/ 20 July Nenad Tomasev 135 34 21 2/ 98/ 0 0 0 7 0 From: KOTH Subject: KOTH.ORG: Status - MultiWarrior 94 07/17/06 Date: Mon, 17 Jul 2006 11:42:16 -0400 (EDT) Message-ID: <200607170403.k6H430eW042196@asgard.t-b-o-h.net> Weekly Status on 07/17/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Multiwarrior 94 CoreWar Hill: Last battle concluded at : Thu Jun 8 05:01:40 EDT 2006 # Name Author Score Age 1 Urgle Daniel Rivas 43 7 2 CLP-shot again G.Labarga 38 1 3 kingdom of the grasshoppe simon wainwright 34 122 4 simply believe John Metcalf 29 4 5 JustADirtyClearTest Nenad Tomasev 28 56 6 nameless fragment S.Fernandes 27 26 7 Fluffy Paper VI Jens Gutzeit 25 27 8 Diptera Nenad Tomasev 24 41 9 rooftop pursuit John Metcalf 23 3 10 the price of hostility John Metcalf 21 14 11 the watching eye John Metcalf 20 0 From: KOTH Subject: KOTH.ORG: Status - ICWS Experimental 94 07/17/06 Date: Mon, 17 Jul 2006 11:42:15 -0400 (EDT) Message-ID: <200607170406.k6H460lu042381@asgard.t-b-o-h.net> Weekly Status on 07/17/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG ICWS Experimental 94 CoreWar Hill: Last battle concluded at : Sat May 27 16:29:52 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 46/ 35/ 19 Fatamorgana X Zul Nadzri 157 10 2 44/ 36/ 20 The X Machine Zul Nadzri 152 34 3 44/ 36/ 19 Ogre Christian Schmidt 152 171 4 37/ 22/ 42 xd100 test David Houston 152 20 5 28/ 11/ 61 Evol Cap 4 X John Wilkinson 146 292 6 42/ 40/ 18 Eliminator X Zul Nadzri 145 35 7 41/ 39/ 20 Bewitching S.Fernandes 144 1 8 34/ 25/ 41 KAT v5 Dave Hillis 142 155 9 39/ 37/ 24 Trefoil F 13 Steve Gunnell 140 106 10 35/ 29/ 36 Olivia X Ben Ford 140 104 11 41/ 43/ 16 Giant Hazy Test 13 Steve Gunnell 140 50 12 41/ 43/ 15 O_Fortuna3X Nenad Tomasev 139 6 13 40/ 42/ 18 Black Moods Ian Oversby 139 219 14 38/ 38/ 23 Trefoil Test F 14 Steve Gunnell 139 7 15 37/ 37/ 26 test Some Redcoder 138 4 16 21/ 6/ 73 Evolve X v4.0 John Wilkinson 137 240 17 38/ 41/ 21 Controlled Aggression Ian Oversby 135 223 18 31/ 28/ 41 Glenstorm John Metcalf 135 85 19 38/ 42/ 20 Simply Intelligent Zul Nadzri 134 16 20 39/ 46/ 14 Fatal Choice Some Redcoder 133 5 21 8/ 79/ 12 Reaper II Slaytanist 38 0 From: KOTH Subject: KOTH.ORG: Status - 94 No Pspace 07/17/06 Date: Mon, 17 Jul 2006 11:42:14 -0400 (EDT) Message-ID: <200607170409.k6H490c0043231@asgard.t-b-o-h.net> Weekly Status on 07/17/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG 94 No Pspace CoreWar Hill: Last battle concluded at : Fri Jul 14 20:01:40 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 44/ 50/ 6 s-test Sascha Zapf 138 4 2 28/ 19/ 53 8kSquare Roy van Rijn 138 2 3 32/ 26/ 43 Monster_Human_Grunt inversed 138 266 4 40/ 45/ 15 Pro Radii inversed 136 11 5 32/ 28/ 40 Amicable Antagonist Roy van Rijn 136 1 6 30/ 26/ 44 Monster_Alien_Grunt inversed 135 267 7 31/ 28/ 40 Hullab3loo Roy van Rijn 134 181 8 27/ 21/ 52 Last Judgement Christian Schmidt 133 451 9 37/ 41/ 22 Gabble Christian Schmidt 132 32 10 28/ 24/ 48 Song of the blue sea Miz 131 138 11 36/ 40/ 24 Twilight S.Fernandes 131 151 12 30/ 30/ 39 N e i t h inversed 130 27 13 27/ 24/ 49 MoonOfChaos Nenad Tomasev 130 229 14 27/ 24/ 49 Bluebell Christian Schmidt 130 154 15 28/ 26/ 46 Amber inversed 130 5 16 25/ 20/ 55 Burning Metal inversed 130 6 17 39/ 49/ 11 Hallucination Scanner inversed 130 139 18 37/ 45/ 17 ChimeraQueen Nenad Tomasev 130 146 19 24/ 18/ 58 D3vilstick Roy van Rijn 129 180 20 23/ 18/ 59 dd-test Sascha Zapf 129 15 21 30/ 51/ 19 SBtst_5466h6 inversed 108 0 From: Tuc at Beach House Subject: KOTH.ORG renewed for another year Date: Mon, 17 Jul 2006 23:55:42 -0400 (EDT) Message-ID: <200607180343.k6I3hlxu019986@himinbjorg.tucs-beachin-obx-house.com> Hi, FYI If anyone cared.........I just renewed KOTH.ORG (A bit late) for another year so it'll now come due again 7/13/2007. The server is still not up with the original KOTH server, they are having problems finding a motherboard. Tuc/TBOH From: sayembara@gmail.com Subject: Re: KOTH.ORG renewed for another year Date: 18 Jul 2006 02:56:34 -0700 Message-ID: <1153216594.396032.184540@p79g2000cwp.googlegroups.com> Excellent, I really appreciate your effort :) Tuc at Beach House wrote: > Hi, > > FYI If anyone cared.........I just renewed KOTH.ORG (A bit late) > for another year so it'll now come due again 7/13/2007. > > The server is still not up with the original KOTH server, they > are having problems finding a motherboard. > > Tuc/TBOH From: SimFlyer Subject: Re: Just wanna say Message-ID: Date: Tue, 18 Jul 2006 10:09:41 -0700 On Sun, 16 Jul 2006 11:56:31 -0400 (EDT), wrote: > As an aside, I don't feel like I'm getting much feedback on things. > Maybe you do it all through JKW, but I'd like to know whats going on and > what your looking for, etc. I can't promise anything, especially since now > I'm busier than ever, but I'd like to know what people are thinking and > wanting. IRC is handy (sometimes) to get clarification about some technical point about Corewars. I can leave a session logged and usually find and answer later. The 'hills' are what keep Corewars alive. I haven't been active lately, but when I get back to GA programming, I will be noisy. Without the hills, my interest in Corewars would have never revived. I am aware of the difficulties in maintaining servers, remote or otherwise, and can have plenty of patience awaiting the outcome. -- --- an evolved lifeform -- Posted via a free Usenet account from http://www.teranews.com From: Sascha Zapf Subject: 476 - Making of ... Date: Tue, 18 Jul 2006 18:50:29 +0200 Message-ID: 476 - Making of.... 3p(c) which is the small brother of paper(paper(paper(clear))) was a anti-imp coreclear paper with a big 3c-decoymaker in front. 26 27.4/ 28.8/ 43.7 3p(c) Sascha Zapf 126.0 116 Nice age for tiny but i expected more. So i began to search the weaknesses of 3p(c). 3p(c) was optimized against the first version of Christian Schmidts tiny-fsh, but now there is a second version, much more up to date. So lets have a look to the result against both fsh's. Scoring of the whole warrior Fsh Win Loss Tie Score SD ===================================================== fshtiny0.1 36.48% 24.75% 38.78% 148.21 45.15 fshtiny0.2 33.49% 26.70% 39.82% 140.28 43.93 Scoring of the ripped basewarrior ( Boot and Silk ) Fsh Win Loss Tie Score SD ===================================================== fshtiny0.1 32.68% 25.24% 42.08% 140.11 47.20 fshtiny0.2 29.62% 27.43% 42.96% 131.80 46.97 3p(c) looses many wins and earn only some ties more. Fizmo tries to hold the composition of the different strategies as on the actual hill's. So if your warrior scores more worse against a newer fsh, he probably will be pushed off the hill in the next time. So i take the scores from fshtiny0.2 for the further development of 476. The second table are the results from the ripped mainwarrior. Here are the score of the fsh classes. Class WLT% Score ======================================================================= Coreclears, Oneshot's and Twoshot 38.57/ 32.20/ 29.23 144.94 Clear-directing Scanner 40.80/ 38.90/ 20.30 142.70 Paper 26.51/ 12.35/ 61.15 140.67 Stones with a-imps 20.30/ 11.00/ 68.70 129.60 Stones with b-imps 20.40/ 17.40/ 62.20 123.40 Paper with 3/7-point imps 11.53/ 1.73/ 86.73 121.33 Paper with seperate stone 13.70/ 11.50/ 74.80 115.90 Scanner 27.33/ 58.30/ 14.37 96.37 Bomber and Vampires without imps 21.87/ 69.67/ 8.47 74.07 Only weaknesses against Clear-directing Scanner like tiny zooom and hired sword and against the oneshots. Ties, ties and ties against all kind of Papers and the Stone/Imps. Powerful against the scanners like Origin of Storm, solo or clockwork. That's what i started with. My maintarget was to replace the 3c decoymaker by a 2c qbomber which scores much better against tiny-hills then on the 8000er or 55440er hills. The decoymaker only seems to work against one class of scanner. I started the development from this basewarrior code. pgo spl 2 spl 1 spl 1 pAw1 equ 100 pEnd equ cc+1 mov {cp , {pBo1 pBo1 jmp pEnd+pAw1 ;constants for the coreclear paper tstep1 equ 38 nstep1 equ 125 cstep2 equ 355 cp spl @pEnd , cp nothB spl @nothB , nothB mov.i #1 , <1 cc djn.b -2 , #cstep2 end pgo Some words about Score and SD. Most important criterion while optimizing is the Score. But in the earlier days of optimax, while testing the dynamic targets, Fizmo and me decided that we need a way to avoid 'magic numbers'. If you get a magic number against one warrior or a whole class the scores grows up unnatural. Thats why i have implemented the column SD, which is simple the standard deviation all over the scores of the randry. That works fine. If you see that the SD of one of the randry's is more then 15% above the average SD can see that there is a magic number and you shouldn't use him for further developments. In the last i give the SD more attention. After a run in a multi step optimizing like describe in this making of i always choose the randry with the lowest SD from the top five for the next step. Sometimes there are only 0.1 points but sometimes there is a difference about 0.5 points. I future i will do some experiments with class related SD, which should tell us something about how sensitive a warrior is against changings in the class balance on his hill. Step 1: New Silkstep's Nearly 50% of the fsh's warriors have changed. Some very powerful paper's like 3p(c) or some evolved paper are new. So new silksteps are undispensable. I had optimized the steps in three batches. Bootdistance was fixed to 100. 1) Main silkstep ( tstep1 ) 2) Center silkstep ( nstep1 ) 3) Anti-imp bombing ( cstep2 ) For each batch optimax runs for approximate 1 hour on 600Mhz. After that i choose the randry with the lowest SD out of the Phase4 Topten. Optimax settings ================ Phase 2 Opponent Digital Swarm Phase 2 Target 90 Static Phase 2 Rounds 200 Phase 3 Opponents pap:pws:pwi Phase 3 Target 100 Static Phase 3 Rounds 400 Phase 4 Opponents 100% of fshtiny0.2 Phase 4 Rounds 400 Results ======= Fsh Win Loss Tie Score SD ===================================================== fshtiny0.2 30.23% 26.04% 43.73% 134.41 42.94 O.K, only three points more at the score, but, and thats more important, the SD, the standart deviation allover the score decreased by three points too. You can see, most classes lose some points. Last two columns are showing the score compared to before this run and from the beginning of the development Class WLT% Score run all =============================================================================== Clear-directing Scanner 40.80/ 42.40/ 16.80 139.20 -2.45% -2.45% Coreclears, Oneshot's and Twoshot 34.85/ 33.48/ 31.68 136.22 -6.02% -6.02% Paper 24.44/ 13.64/ 61.92 135.24 -1.19% -1.18% Stones with a-imps 16.80/ 11.40/ 71.80 122.20 -5.71% -5.71% Paper with 3/7-point imps 10.47/ 1.73/ 87.80 119.20 -1.76% -1.76% Stones with b-imps 19.20/ 19.60/ 61.20 118.80 -3.73% -3.73% Paper with seperate stone 14.05/ 13.75/ 72.20 114.35 -1.34% -1.34% Scanner 32.83/ 52.93/ 14.23 112.73 16.77% 16.77% Bomber and Vampires without imps 19.27/ 71.17/ 9.57 67.37 -9.05% -9.05% Every class but the scanners loose some power. Stones and the Oneshots the most. Better spreading makes it more difficult to hit and to find the right one when the papers are coloring the core fast enough. Scanners like Clockwork and Origin of Storm are scoring better because of the smaller targets which let them not so long in their bombing runs. Step 2: Bootdistances and Decoy's In most cases i optimize bootdistance and all kind of decoys at once. Because of the big qbomber i want to use, i decide to insert a colored decoy in front of the bootcode. The jmp for start the booted silk was changed to djn, with a little dekrement train which should be a cheap additional target for scissors. n for 9 dat n,-n rof pgo spl 2 spl 1 spl 1 pAw1 equ !(20-780) pAw2 equ !(20-780) pEnd equ cc+1 mov {cp , {pBo1 pBo1 djn.f pEnd+pAw1 , cp nothB spl @nothB, nothB mov.i #1, <1 cc djn.b -2, #cstep2 end pgo Optimax settings ================ Phase 2 Opponent T766 Phase 2 Target 120 Static Phase 2 Rounds 200 Phase 3 Opponents clr:scn Phase 3 Target 120 Static Phase 3 Rounds 400 Phase 4 Opponents 100% of fshtiny0.2 Phase 4 Rounds 400 After a first run of 120 mins and over 1500 Randrys the genoms of the topten warriors looks like that. pAw1 pAw2 !(20-780) !(20-780) 91 370 90 480 96 492 89 384 105 323 90 265 96 526 44 333 98 87 94 339 As you can see that 90% of the pAw1 genes are in the range 85-105 and 90% of the pAw2 genes are in the range 250-500. So i had decided to reduce the range of the operators. pAw1 from !(20-780) to !(85-100) pAw2 from !(20-780) to !(250-500) Results ======= I gave them another hour to run. After that, surprise surprise Fsh Win Loss Tie Score SD ===================================================== fshtiny0.2 32.67% 24.12% 43.21% 141.23 45.01 Class WLT% Score run all =============================================================================== Paper 24.67/ 15.71/ 59.63 133.63 -1.19% -5.00% Clear-directing Scanner 35.80/ 44.70/ 19.50 126.90 -8.84% -11.07% Stones with a-imps 17.70/ 10.90/ 71.40 124.50 -1.88% -3.94% Stones with b-imps 21.40/ 19.60/ 59.00 123.20 3.70% -0.16% Coreclears, Oneshot's and Twoshot 30.09/ 37.63/ 32.28 122.55 -10.04%-15.45% Paper with 3/7-point imps 10.87/ 3.13/ 86.00 118.60 -0.50% -2.25% Paper with seperate stone 12.35/ 13.90/ 73.75 110.80 -3.10% -4.40% Scanner 26.40/ 58.97/ 14.63 93.83 -16.77% -2.64% Bomber and Vampires without imps 20.67/ 70.77/ 8.57 70.57 4.75% -4.73 All kind of Scanner are the big looser of this run because creating some decoy or leave behind 20 very colorful cell without active processes will fool the scissors most of the corewar world. Step 3: 2c qbomber Now i change the 9 instruction colored decoy in front of the bootstrap to a 2c decoymaker in the simple from - mov.i coreclear paper ;strategy 3p(c) reworked ;assert CORESIZE==800 ; --- 2c qbomber --- start mov.i <538, 706 mov.i <125, 773 mov.i <183, 372 mov.i <48 , 188 mov.i <443, 398 mov.i <569, 220 mov.i <534, 329 mov.i <336, 523 mov.i <97 , 249 ; --- create 6 parallel processes pgo spl 2 , <318 spl 1 , <513 spl 1 , <466 ; --- boot and decoy constants pAw1 equ 97 pAw2 equ 490 pEnd equ cc+1 ; --- boot away and start --- mov {cp , {pBo1 pBo1 djn.f pEnd+pAw1 , cp nothB spl @nothB , nothB mov.i #1 , <1 cc djn.b -2 , #cstep2 end start ========8<========8<========8<========8<========8<========8<========8<======== SAL's loveletter to me... ;-) "476" by "Sascha Zapf" has challenged the Tiny hill ;strategy qbomber -> coreclear paper ;strategy 3p(c) reworked Current hill status: # %W/ %L/ %T Name Author Score Age 1 32.8/ 24.5/ 42.6 476 Sascha Zapf 141.2 1 2 34.1/ 27.9/ 38.0 desTiny Sascha Zapf 140.3 26 3 29.8/ 21.4/ 48.7 Digital Swarm John Metcalf 138.2 207 4 32.7/ 27.2/ 40.1 Endless pain G.Labarga 138.1 93 5 39.6/ 41.3/ 19.1 Sneaky Spike 2 Roy van Rijn 137.9 37 6 38.2/ 39.5/ 22.2 White Noise (RBv1.5r10) The MicroGP Corewa 136.9 92 7 39.4/ 42.4/ 18.1 Four Winds John Metcalf 136.4 120 8 37.3/ 39.2/ 23.5 Blue bubble Dr.Gapp 135.4 17 9 28.5/ 22.4/ 49.1 Yet Another Tiny Paper Jens Gutzeit 134.7 3 10 28.6/ 22.9/ 48.5 Backdraft Sascha Zapf 134.2 16 11 37.5/ 41.6/ 20.8 Diptera Nenad Tomasev 133.4 2 12 23.6/ 14.4/ 62.0 TL5 G.Labarga 132.8 5 13 41.3/ 49.9/ 8.8 Muskrat John Metcalf 132.8 125 14 26.6/ 20.6/ 52.8 Moomintroll John Metcalf 132.6 56 15 38.3/ 44.1/ 17.7 against all odds John Metcalf 132.4 6 16 22.4/ 12.4/ 65.2 tiny Blowrag Metcalf/Schmidt 132.4 221 17 37.9/ 43.8/ 18.3 interlocking John Metcalf 131.9 4 18 32.6/ 33.5/ 33.9 All over the core G.Labarga 131.8 14 19 30.2/ 29.1/ 40.7 Hades nebula Sascha Zapf 131.4 19 20 37.1/ 42.9/ 20.0 Hired Sword John Metcalf 131.2 58 21 28.0/ 25.4/ 46.6 MTP Nenad Tomasev 130.6 25 22 37.0/ 43.5/ 19.6 holograph Simon Wainwright 130.4 36 23 36.9/ 44.0/ 19.0 sixteen and counting down Simon Wainwright 129.8 48 24 37.3/ 44.9/ 17.8 WriteLn('Victory'); Nenad Tomasev 129.8 30 25 37.0/ 44.2/ 18.8 Milkshake V LAchi 129.7 15 26 37.0/ 44.3/ 18.8 Tiny BiShot 2.0 Christian Schmidt 129.7 221 Challenge results: Score # %W/ %L/ %T Name Author Taken Given 3 33.9/ 24.8/ 41.3 desTiny Sascha Zapf 143.1 115.7 6 32.7/ 26.5/ 40.9 White Noise (RBv1.5r10) The MicroGP Corewa 138.9 120.3 7 36.5/ 35.6/ 27.9 Four Winds John Metcalf 137.5 134.7 5 24.0/ 14.9/ 61.1 Digital Swarm John Metcalf 133.1 105.9 24 24.0/ 15.6/ 60.4 MTP Nenad Tomasev 132.4 107.2 9 22.9/ 13.9/ 63.2 Yet Another Tiny Paper Jens Gutzeit 131.8 105.0 17 23.9/ 21.0/ 55.1 All over the core G.Labarga 126.9 118.1 4 22.9/ 22.1/ 55.0 Endless pain G.Labarga 123.8 121.2 11 23.7/ 24.0/ 52.3 Backdraft Sascha Zapf 123.5 124.3 14 16.6/ 10.4/ 73.0 TL5 G.Labarga 122.8 104.2 16 17.3/ 12.2/ 70.5 Moomintroll John Metcalf 122.3 107.1 13 13.5/ 8.7/ 77.8 tiny Blowrag Metcalf/Schmidt 118.2 104.0 8 28.3/ 39.3/ 32.4 Blue bubble Dr.Gapp 117.2 150.4 18 30.0/ 45.1/ 24.9 against all odds John Metcalf 114.9 160.1 2 30.1/ 45.5/ 24.4 Sneaky Spike 2 Roy van Rijn 114.8 160.8 1 15.7/ 17.9/ 66.4 476 Sascha Zapf 113.4 120.2 22 25.9/ 43.8/ 30.3 sixteen and counting down Simon Wainwright 108.1 161.7 21 17.7/ 29.5/ 52.8 Hades nebula Sascha Zapf 105.8 141.4 23 28.1/ 50.5/ 21.5 WriteLn('Victory'); Nenad Tomasev 105.7 172.9 15 25.4/ 46.3/ 28.3 Diptera Nenad Tomasev 104.5 167.1 25 25.9/ 49.0/ 25.1 Milkshake V LAchi 102.9 172.1 20 25.7/ 52.1/ 22.3 Hired Sword John Metcalf 99.3 178.5 19 20.9/ 44.4/ 34.7 holograph Simon Wainwright 97.5 167.9 26 24.6/ 51.9/ 23.5 Tiny BiShot 2.0 Christian Schmidt 97.3 179.3 12 25.4/ 55.4/ 19.2 interlocking John Metcalf 95.4 185.4 10 22.5/ 72.7/ 4.9 Muskrat John Metcalf 72.3 222.9 There a many scanner on the tiny-hill today, so i had expected a high position. I hope i have shown how to do developing and optimizing a new warrior together. After each step in development you get best results so far, and so you'll know how to go on. You see new weaknesses or if something not in order with your warrior. O.K, where's my crown...Need it fast because i don't think that i'm koth for a long time. White Noise or Digital Swarm are hungry ;-) Hope you enjoy this article Sascha Zapf May be the core with you! P.S: 476 was a battle droid in the game 'Paradroid' from Andrew Braybrook my favorite computer game on the C64. From: Sascha Zapf Subject: Re: Just wanna say Date: Thu, 20 Jul 2006 13:07:43 +0200 Message-ID: > As an aside, I don't feel like I'm getting much feedback on things. > Maybe you do it all through JKW, but I'd like to know whats going on and > what your looking for, etc. I can't promise anything, especially since now > I'm busier than ever, but I'd like to know what people are thinking and > wanting. > > Tuc IRC is very important for all off us. eMails are nice but really conversation and Brainstorming is only possible with IRC. And the Hill's are the Spirit of Corewar..we need them. Keep them alive thanx Sascha -- News From: KOTH Subject: KOTH.ORG: Status - Standard 07/24/06 Date: Mon, 24 Jul 2006 00:22:19 -0400 (EDT) Message-ID: <200607240400.k6O400Sl087723@asgard.t-b-o-h.net> Weekly Status on 07/24/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Standard KotH CoreWar Hill : Last battle concluded at : Mon Jul 17 12:23:00 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 34/ 25/ 41 Test Alexander (Sasha) Wa 143 202 2 31/ 19/ 49 Raging Gale '88 inversed 143 7 3 37/ 32/ 30 White Fire inversed 142 4 4 36/ 30/ 34 The Next Step '88 David Houston 141 78 5 33/ 27/ 40 Triangular Sun inversed 138 3 6 42/ 46/ 12 Monolith inversed 138 11 7 40/ 43/ 18 Hexamorph inversed 137 6 8 38/ 39/ 24 My 1st try Christian Schmidt 137 94 9 39/ 42/ 20 Scan the Can Christian Schmidt 136 43 10 31/ 27/ 43 The Hurricaner G.Labarga 135 49 11 29/ 26/ 45 test G.Labarga 133 45 12 30/ 27/ 43 Guardian Ian Oversby 132 262 13 24/ 17/ 59 Utterer '88 Christian Schmidt 131 37 14 30/ 30/ 39 SoundOfDarkness Nenad Tomasev 131 26 15 31/ 33/ 36 usual fee, plus expenses John Metcalf 130 1 16 33/ 36/ 31 Vampire Knight inversed 130 2 17 39/ 49/ 12 Speeed 88mph Christian Schmidt 130 62 18 35/ 41/ 24 Moonwipe Christian Schmidt 130 59 19 39/ 48/ 13 Scan Test C 6 Steve Gunnell 130 101 20 32/ 34/ 34 The Seed Roy van Rijn 129 80 21 36/ 43/ 21 July Nenad Tomasev 128 35 From: KOTH Subject: KOTH.ORG: Status - MultiWarrior 94 07/24/06 Date: Mon, 24 Jul 2006 00:22:18 -0400 (EDT) Message-ID: <200607240403.k6O430wD087931@asgard.t-b-o-h.net> Weekly Status on 07/24/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Multiwarrior 94 CoreWar Hill: Last battle concluded at : Thu Jun 8 05:01:40 EDT 2006 # Name Author Score Age 1 Urgle Daniel Rivas 43 7 2 CLP-shot again G.Labarga 38 1 3 kingdom of the grasshoppe simon wainwright 34 122 4 simply believe John Metcalf 29 4 5 JustADirtyClearTest Nenad Tomasev 28 56 6 nameless fragment S.Fernandes 27 26 7 Fluffy Paper VI Jens Gutzeit 25 27 8 Diptera Nenad Tomasev 24 41 9 rooftop pursuit John Metcalf 23 3 10 the price of hostility John Metcalf 21 14 11 the watching eye John Metcalf 20 0 From: KOTH Subject: KOTH.ORG: Status - ICWS Experimental 94 07/24/06 Date: Mon, 24 Jul 2006 00:22:17 -0400 (EDT) Message-ID: <200607240406.k6O460q2088116@asgard.t-b-o-h.net> Weekly Status on 07/24/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG ICWS Experimental 94 CoreWar Hill: Last battle concluded at : Sat May 27 16:29:52 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 46/ 35/ 19 Fatamorgana X Zul Nadzri 157 10 2 44/ 36/ 20 The X Machine Zul Nadzri 152 34 3 44/ 36/ 19 Ogre Christian Schmidt 152 171 4 37/ 22/ 42 xd100 test David Houston 152 20 5 28/ 11/ 61 Evol Cap 4 X John Wilkinson 146 292 6 42/ 40/ 18 Eliminator X Zul Nadzri 145 35 7 41/ 39/ 20 Bewitching S.Fernandes 144 1 8 34/ 25/ 41 KAT v5 Dave Hillis 142 155 9 39/ 37/ 24 Trefoil F 13 Steve Gunnell 140 106 10 35/ 29/ 36 Olivia X Ben Ford 140 104 11 41/ 43/ 16 Giant Hazy Test 13 Steve Gunnell 140 50 12 41/ 43/ 15 O_Fortuna3X Nenad Tomasev 139 6 13 40/ 42/ 18 Black Moods Ian Oversby 139 219 14 38/ 38/ 23 Trefoil Test F 14 Steve Gunnell 139 7 15 37/ 37/ 26 test Some Redcoder 138 4 16 21/ 6/ 73 Evolve X v4.0 John Wilkinson 137 240 17 38/ 41/ 21 Controlled Aggression Ian Oversby 135 223 18 31/ 28/ 41 Glenstorm John Metcalf 135 85 19 38/ 42/ 20 Simply Intelligent Zul Nadzri 134 16 20 39/ 46/ 14 Fatal Choice Some Redcoder 133 5 21 8/ 79/ 12 Reaper II Slaytanist 38 0 From: KOTH Subject: KOTH.ORG: Status - 94 No Pspace 07/24/06 Date: Mon, 24 Jul 2006 00:22:16 -0400 (EDT) Message-ID: <200607240409.k6O4904p088256@asgard.t-b-o-h.net> Weekly Status on 07/24/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG 94 No Pspace CoreWar Hill: Last battle concluded at : Sun Jul 23 14:21:16 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 45/ 49/ 6 s-test Sascha Zapf 142 4 2 33/ 24/ 43 Monster_Human_Grunt inversed 141 266 3 33/ 27/ 40 Amicable Antagonist Roy van Rijn 140 1 4 33/ 27/ 40 Hullab3loo Roy van Rijn 140 181 5 32/ 25/ 44 Monster_Alien_Grunt inversed 139 267 6 28/ 18/ 53 8kSquare Roy van Rijn 138 2 7 41/ 44/ 15 Pro Radii inversed 137 11 8 29/ 22/ 49 Song of the blue sea Miz 136 138 9 28/ 21/ 51 Last Judgement Christian Schmidt 136 451 10 27/ 19/ 54 Burning Metal inversed 136 6 11 29/ 23/ 48 Bluebell Christian Schmidt 135 154 12 32/ 29/ 40 N e i t h inversed 134 27 13 28/ 23/ 49 MoonOfChaos Nenad Tomasev 134 229 14 26/ 18/ 56 D3vilstick Roy van Rijn 134 180 15 41/ 48/ 12 Hallucination Scanner inversed 133 139 16 25/ 17/ 58 dd-test Sascha Zapf 133 15 17 36/ 39/ 25 Twilight S.Fernandes 132 151 18 28/ 24/ 47 Amber inversed 132 5 19 36/ 41/ 23 Gabble Christian Schmidt 132 32 20 38/ 45/ 17 ChimeraQueen Nenad Tomasev 130 146 21 9/ 73/ 19 Second_try Spiritofsparta 45 0 From: "Neogryzor" Subject: Re: KOTH.ORG renewed for another year Date: 24 Jul 2006 05:27:13 -0700 Message-ID: <1153744033.287768.293690@h48g2000cwc.googlegroups.com> Tuc at Beach House ha escrito: > Hi, > > FYI If anyone cared.........I just renewed KOTH.ORG (A bit late) > for another year so it'll now come due again 7/13/2007. > > The server is still not up with the original KOTH server, they > are having problems finding a motherboard. > > Tuc/TBOH Great, thanks a lot! From: Tuc at Beach House Subject: IRC.KOTH.ORG back up Date: Mon, 24 Jul 2006 18:12:56 -0400 (EDT) Message-ID: <200607242143.k6OLh6EH001948@himinbjorg.tucs-beachin-obx-house.com> Hi, The IRC.KOTH.ORG is back up. The logger is running too.. I'll be taking irc2 down in the next 24 hours or so. Tuc From: "=?iso-8859-1?q?M._Weing=E4rtner?=" Subject: Re: KOTH Hill discussion Date: 26 Jul 2006 12:42:54 -0700 Message-ID: <1153942974.265711.247330@i3g2000cwc.googlegroups.com> Hi. Although I'm new to corewar (active, reading on websites and irc for a long time) and without having written much yet I'd like to write my opinion on this. I don't think it's "unfair" having handwritten and evolved warriors fighting against each other. They are both warriors, using their own strategy. On the other hand I'd personally prefer a Noo/e hill populated with handwritten warriors. Old school warriors have an author behind them, the author thinks a lot (maybe?) about strategies, tries to learn, tries to optimize his code, tries to find optimal constants, tries to learn about the opponents strategies, ...they (warriors) have kind of "character" which gives a fight between old school warriors something "special". Please don't get this wrong - optimized/evolved warriors do have many of those things, too. Evolving warriors is a very, very interesting part of corewar. But for me they are missing the character component. For me having a old school hill, a evolved warriors hill and a mixed hill would be a great addition, but maybe leads to a problem having way too much hills - and not enough warriors to populate them ? From: Tuc at Beach House Subject: KOTH Hill discussion Date: Wed, 26 Jul 2006 14:12:18 -0400 (EDT) Message-ID: <200607261759.k6QHx7aY028836@himinbjorg.tucs-beachin-obx-house.com> Hi, I was talking to someone on IRC, and they mentioned something I want to bring up for discussion. Being my usual self, on top of everything (Said with my tongue shoved WAY UP my cheek) I've come to learn there are redcode optimizers available to the general public, most notably "Optimax". What seems to be happening is that warriors written by hand are losing out more and more to the ones that were written and programatically optimized. My *PERSONAL* feeling on this is that it is unfair to those that still do it the old fashioned way, and unfair to force them to use it to keep up with the rest. What are peoples feelings about this? Would you want a ban on its usage? Would you want "no-optimize" hill(s)? Am I just being stupid and should go back to making sure the servers stay up? Tuc From: sayembara@gmail.com Subject: Re: KOTH Hill discussion Date: 26 Jul 2006 21:28:43 -0700 Message-ID: <1153974523.499831.126250@m73g2000cwd.googlegroups.com> May be those 'old' warriors did being optimized in some way...may be spreadsheet way? :) BTW, since it is not that clear how to identified optimized warriors, I say we keep existing format...let it open for hand coded or program optimized warriors. BTW, I have not use Optimax for any submission yet...probably not using it in the future. The change of warriors in the hill will show increasing 'strength' or changing strategies. /Zul Nadzri From: dhillismail@netscape.net Subject: Re: KOTH Hill discussion Date: 26 Jul 2006 22:04:35 -0700 Message-ID: <1153976675.066554.227680@i3g2000cwc.googlegroups.com> Attempts to make evolver only contests have always been hampered by the lack of any clear line demarcating purely-evolved from mostly-evolved-but-partly-designed warriors. At the other end of the spectrum, exactly when does testing become tainted by optimization? (Time for a fuzzy logic hill.) There's pretty much the same problem with "original" warriors. Dave Hillis From: "Cristian210273" Subject: Re: KOTH Hill discussion Date: 26 Jul 2006 23:11:42 -0700 Message-ID: <1153980701.971852.222700@i3g2000cwc.googlegroups.com> > What are peoples feelings about this? Would you want a ban on > its usage? Would you want "no-optimize" hill(s)? Am I just being stupid > and should go back to making sure the servers stay up? > > Tuc Tuc, I'm totally new to this field, but I think that you have to think of the chess game: while more and more sophisticated and pretty strong computer chess game are designed, human players still beat the machine. I think that in the KOTH competition we are in a similar situation...... Cristian From: "Cristian210273" Subject: AI warrior. Date: 26 Jul 2006 23:14:41 -0700 Message-ID: <1153980881.242196.313970@s13g2000cwa.googlegroups.com> Hi everybody. I'm totally new to this field, so I'm sorry if the question sounds silly...... Does anybody have designed a warrior using AI techniques (like fuzzy logic or Neural Network)? Do you think that such a technology could be too much "heavy" in terms of number of instructions, memory usage,...., to be used to design a warrior? Any help or hint would be appreciated. Thank you! From: Sascha Zapf Subject: Re: KOTH Hill discussion Date: Wed, 26 Jul 2006 23:45:42 +0200 Message-ID: Tuc at Beach House wrote: > Hi, > > I was talking to someone on IRC, and they mentioned something I > want to bring up for discussion. > > Being my usual self, on top of everything (Said with my tongue > shoved WAY UP my cheek) I've come to learn there are redcode optimizers > available to the general public, most notably "Optimax". > > > What seems to be happening is that warriors written by hand > are losing out more and more to the ones that were written and > programatically optimized. > > > My *PERSONAL* feeling on this is that it is unfair to those > that still do it the old fashioned way, and unfair to force them to > use it to keep up with the rest. > No one is forced to use the optimizer or evolver software that is arround! Highend optimized Warriors are as weak as each other. One Example is the return of the dodgers like "Bottomless Pit" from Fizmo and Roy or "Interlocking" from John Metcalf. The success of this strategy is the consequence of more and more high optimized paper. These paper's are overwriting own instances occasionally, so it's the best for dodger. I think Corewars is changing like the whole IT. At the beginning, where a small benchmark against 12 Warrior needs more then 30 minutes the Player had to choose their constants by hand. But now ever > What are peoples feelings about this? Would you want a ban on > its usage? Would you want "no-optimize" hill(s)? Am I just being stupid > and should go back to making sure the servers stay up? Sometimes while creating a new warrior or change an older one i insert a random value for bootdistance or for a decoymaker. When the warrioir is tested with cdb i start the optimizer and wait for the best constants. And sometimes after 10 hour of optimizing you will see that the "random" choosed constant is the best so far. So how would you decide if a warrior is optimized or not. On the other hand, i'm as a hardcore optimizer, didn't have any chance at the actual tournament's as csec 2005 or the running one. Because i didn't have any fsh for the gives hill-specifications to test my warrior and i don't able to choose good constans by hand. > > Tuc I think that there is another competion grown the last years. Not only paper - scissor - stone but also evolver - optimizer - handmade Sascha -- News From: achillu@tin.it Subject: Re: KOTH Hill discussion Date: 27 Jul 2006 00:21:08 -0700 Message-ID: <1153984868.224412.275520@m79g2000cwm.googlegroups.com> Tuc at Beach House wrote: > Hi, There is a huge difference between optimized and evolved warriors, and Optimax is not an evolving tool at all! IMO optimaxed warriors are just like any other handwritten warriors and Optimax is just a tool you can use to adjust some numbers in your handwritten warrior, but sure you will not change the "spirit" or the "character" of your warriors by simply changing those numbers. Using Optimax you do not create "new" warriors, but you simply try to get the best (say) "My Cool Paper" out of the bunch of all possible variants of "My Cool Paper", and you will never get an "Evolved Cool Paper". From: SimFlyer Subject: Re: KOTH Hill discussion Message-ID: Date: Thu, 27 Jul 2006 05:23:42 -0700 On Wed, 26 Jul 2006 14:12:18 -0400 (EDT), wrote: > > What seems to be happening is that warriors written by hand > are losing out more and more to the ones that were written and > programatically optimized. > > My *PERSONAL* feeling on this is that it is unfair to those > that still do it the old fashioned way, and unfair to force them to > use it to keep up with the rest. GIGO still rules here. Meaning that unless it's a semi-decent warrior in the first place, optimization will not help. I can run hand-optimization (and have done so) simply by making a large number of hill submissions with a different 'constant' value each time. Program tools just save time and load on the hills for that same process. "keeping up with the rest" has never been on my agenda. -- --- an evolved lifeform -- Posted via a free Usenet account from http://www.teranews.com From: SimFlyer Subject: Re: AI warrior. Message-ID: Date: Thu, 27 Jul 2006 05:35:24 -0700 On 26 Jul 2006 23:14:41 -0700, wrote: > > Does anybody have designed a warrior using > AI techniques (like fuzzy logic or Neural Network)? > Do you think that such a technology could be > too much "heavy" in terms of number of instructions, > memory usage,...., to be used to design a warrior? > > Any help or hint would be appreciated. Go hang out on the nano hill. I was fooling around with some numbers the other day and it seems that you could test every possible warrior for that hill in about 18 months (for my machine). The 5 instruction limit make that hill makes it a good place for evolvers, including the human based variants. 'Neural Networks' - wrong hammer for this nail. You should start by studying the redcode instruction set. Sifting thru the archives will be helpful for you, it's an old FAQ. Personally, I'd recomment the Koza approach. -- --- an evolved lifeform -- Posted via a free Usenet account from http://www.teranews.com From: "iapdk@admin.drake.edu" Subject: Re: KOTH Hill discussion Date: 27 Jul 2006 06:17:04 -0700 Message-ID: <1154006224.149117.40110@h48g2000cwc.googlegroups.com> I used Optimax for 2by4k (paper) and Codgeriffic (dodger). The constants it chose did poorly on the hill. Often it would narrow in on a set of numbers that completely whipped one or two opponents and did marginally well against the rest. Realistically, given that the programs you optimize against are not the ones on the hill, the best use of Optimax for me was to rule out bad numbers and give the program a fair chance to succeed on the merits of its strategy. Paul K From: "Cristian210273" Subject: Re: AI warrior. Date: 27 Jul 2006 07:08:54 -0700 Message-ID: <1154009333.832125.119790@p79g2000cwp.googlegroups.com> > > The 5 instruction limit make that hill makes it a good place for > evolvers, including the human based variants. > > 'Neural Networks' - wrong hammer for this nail. > Thank you! I will investigate more deeply..... However there was a misunderstanding: with "warrior using AI techniques" I meant "warrior whose code implements some AI inside, like fuzzy logic". From: "inversed" Subject: optimization equ evil Date: 27 Jul 2006 15:13:45 -0700 Message-ID: <1154038425.636171.120880@i42g2000cwa.googlegroups.com> Tuc started that theme, so here goes ... IMO optimax is a big problem nowdays, having more negative than positive effects. Negative effects came from the fact that people don't think how to make good constants (and therefore don't think of their warrior's specifics and structure), they just bruteforce them, and that some people just recycle strategies. I classify authors in this manner: - Handcoders (who may use optimization to improve their warrior's performance) - Evolvers, who _usually_ have bad redcode knowledge and do not write a single line by themselves - Optimizers, who optimize-to-death already existing structures instead of writing new warriors, maybe multiple times! (I'll not point my finger ...) Tuc suggested using special hill, but my opinion is "no": given a load file, you can't objectively distinguish between optimized and non-optimized warrior, so there is only one choice: don't use optimax, but write a good warrior. Impossible, you say? Look at the hills then. I don't use optimax, I don't optimize for weeks, but I dominate 94, 94nop, 88 and LP. I use simple program to replace !a b! string with random number in [a..b] range + my own optimization techniques + corestep (or my own similair program) to test some selected constants, and CoreWin to test randries (usually no more than 1000). So my warriors are live prove that there is no need for NoOpt hill. The problem is that spirit of corewars changed :( Optimize-to-death is not the thing that was envisioned by Dewney. You don't need to run 10^F#$%INGHUGEPOWER instructions to make a good warrior. That very popular "clone_other_warrior -> add_modern_quickscan -> reoptimize_ten_times -> submit_to_the_hill" thingy is JUST F*&^ING AWFUL!!! So, I suggest for all optimizers to repent and become handcoders (or quit). Simple example of how abovementioned approach harmed corewars: people started to optimize dat-throwing papers against imps, but no one tryed to understand how these heavy-optimized papers work, what are they, why they loose to Moore-style papers, and why they beat imps. So huge subclass of silkish papers was just left unnoticed. I want to see corewars as something more than tweaking quckscans (wrote by others) and seeking good constants. There are still a lot of new strategies to explore even at 94nop, believe me! "The net is vast and infinite" From: "Peter Duniho" Subject: Re: optimization equ evil Date: Thu, 27 Jul 2006 18:56:12 -0700 Message-ID: <12cirm9fbliu53b@corp.supernews.com> Okay, a brief respite from my usual lurking... I don't have a strong opinion about the optimization/evolving issue. The purity of hand-coded-only appeals to me, but I also agree that a) once you see the outcome of an optimizer and/or evolver, there should be no reason one cannot hand-code an answer to it, and b) the outcome of the optimizer and/or evolver is only as good as the tools themselves and the person using them (so there is still some sense of skill and competition there). But, one issues that I haven't seen anyone mention and as an idle observer would love to see more conversation about is whether anyone has actually analyzed the output of an optimizer (especially...I'm less interested in the evolver idea, but I suppose that could be relevant too), to try to understand and describe why one constant works better than another. Are the constants only "better" with respect to the existing warriors being fought? Or are some constants inherently better than others, and if so, why? I realize that in broad strokes, there are some constants that would be clearly bad and not useful (for example, constants that result in only spotty coverage of the core). But it seems as though among the set of constants that are not obviously bad, there are still some that do better than others. Especially as a beginner, I would be very interested to know why, and it seems to me that looking at the results of an optimizer would help that, especially if it is the author of the warrior who is doing that analysis and explaining it to the community generally. So how about it? Anyone interested in shifting the direction of the discussion from "should these warriors exist and be permitted to fight hand-coded ones" to the more interesting one of "how is it that one random number is superior to a different but similar random number?" Oh, and I have to say, I don't see how anyone can claim to know what Mr. Dewdney would say about optimizers or evolvers. Short of asking him, it seems presumptuous to invoke his supposed views on the matter. Pete From: bvowk@math.ualberta.ca Subject: Re: optimization equ evil Date: 27 Jul 2006 20:26:11 -0700 Message-ID: <1154057171.360448.259320@b28g2000cwb.googlegroups.com> inversed wrote: > - Evolvers, who _usually_ have bad redcode knowledge and do not write > a single line by themselves Ouch. You could just whisper it behind my back! We evolvers have feelings too, we're not just redcode manipulating automatons you know! From: SimFlyer Subject: Re: AI warrior. Message-ID: Date: Thu, 27 Jul 2006 23:26:34 -0700 On 27 Jul 2006 07:08:54 -0700, wrote: > > Thank you! I will investigate more deeply..... > However there was a misunderstanding: > with "warrior using AI techniques" I meant > "warrior whose code implements some AI > inside, like fuzzy logic". No. I suspected that, which is why I recommended studying Redcode first. -- --- an evolved lifeform -- Posted via a free Usenet account from http://www.teranews.com From: SimFlyer Subject: Re: optimization equ evil Message-ID: Date: Thu, 27 Jul 2006 23:56:35 -0700 On 27 Jul 2006 15:13:45 -0700, wrote: > > - Evolvers, who _usually_ have bad redcode knowledge and do not write > a single line by themselves Writing a line isn't the point afterall. But 'bad redcode' knowledge exists because of some confusion in the specs. I remember looking into the several MARS and noting that they had comments about code interpretation differences. Since I was on linux, and was planning on using Common Lisp, I realized that I needed a verification suite to know that my new MARS would give identical results to the hills (whatever they might run). It finally boiled down to needing to use pMARS somehow, there isn't a verification suite for Redcode, and I recall reading 'the code does what pMARS does, the spec seems to require something different' comments. A large set of sample redcode and its correct assembler would clear up some of the questions. I ended up single-stepping test code in Corewin to see how it worked. I've read http://vyznev.net/corewar/guide.html enough to see how to efficiently organize an evolver. I understood it well enough to see how several simple warriors functionally block out. I'm still not interested in writing a single line. As for 'getting rid of' some of the corewar users for 'being bad', I think the community is too small now.... -- --- an evolved lifeform -- Posted via a free Usenet account from http://www.teranews.com From: SimFlyer Subject: Re: optimization equ evil Message-ID: Date: Thu, 27 Jul 2006 23:58:13 -0700 On Fri, 28 Jul 2006 01:07:27 +0200, wrote: > > Discover new strategys is a hard job. Most of us are don't have the time for > do that. That's what evolvers are after...or at least one of the goals. -- --- an evolved lifeform -- Posted via a free Usenet account from http://www.teranews.com From: SimFlyer Subject: Re: optimization equ evil Message-ID: Date: Fri, 28 Jul 2006 00:00:48 -0700 On Thu, 27 Jul 2006 18:56:12 -0700, wrote: > > I realize that in broad strokes, there are some constants that would be > clearly bad and not useful (for example, constants that result in only > spotty coverage of the core). But it seems as though among the set of > constants that are not obviously bad, there are still some that do better > than others. Especially as a beginner, I would be very interested to know > why, and it seems to me that looking at the results of an optimizer would > help that, especially if it is the author of the warrior who is doing that > analysis and explaining it to the community generally. Having gone thru this myself, my best advice is simply pick a simple warrior and do this yourself. You will gain far more than reading someone elses analysis. -- --- an evolved lifeform -- Posted via a free Usenet account from http://www.teranews.com From: "Fluffy" Subject: Re: KOTH Hill discussion Date: 28 Jul 2006 00:40:02 -0700 Message-ID: <1154072401.933362.83630@h48g2000cwc.googlegroups.com> Hi, it is an interesting discussion, but I think, that it cannot lead anywhere. There is no way to decide whether a warrior is "optimized" or not. There isn't even a way to decide what "optimized" means. Is it "optimized" when you have used OptiMax or any other program, that randomly checks values? Is it "optimized" when you send a warrior several times to the hill and use different constants every time? > What seems to be happening is that warriors written by hand > are losing out more and more to the ones that were written and > programatically optimized. > > My *PERSONAL* feeling on this is that it is unfair to those > that still do it the old fashioned way, and unfair to force them to > use it to keep up with the rest. I might be wrong, but isn't it the core idea to look for the "best" warrior? If I can beat handwritten warriors with my "optimized" ones, they are obviously better :-) There is nothing unfair about it. Regards, Fluffy From: "Peter Duniho" Subject: Re: optimization equ evil Date: Fri, 28 Jul 2006 01:04:30 -0700 Message-ID: <12cjh8pbs0u1jf2@corp.supernews.com> "SimFlyer" wrote in message news:slrnecjdh0.8lv.spambait@phoenix.clouddancer.com... > Having gone thru this myself, my best advice is simply pick a simple > warrior and do this yourself. You will gain far more than reading > someone elses analysis. Actually, I do not believe your assertion to be true. It may be true for someone who is already experienced in the ways of RedCode. But for a beginner to successfully decipher why one constant is better than another, without having the necessary background in why the warrior works at all, is highly unlikely. In fact, one of the best ways for a beginner to learn, other than through the usual hard work and perseverance (a tried and true method, to be sure) is to take advantage of expert advice and insight. Granted, not all experts are competent to convey their expertise to others, but many are and it's worthwhile for them to try in any case. As an added benefit, it is a much more efficient way for a beginner to learn (trial and error works, but it's very slow), and it would help make the community seem less exclusive (assuming that what you as a community actually want are more participants...I don't know, maybe you actually like that hardly anyone knows about what you're doing). Just my two cents. Pete From: Sascha Zapf Subject: Re: optimization equ evil Date: Fri, 28 Jul 2006 01:07:27 +0200 Message-ID: inversed wrote: > Tuc started that theme, so here goes ... > > IMO optimax is a big problem nowdays, having more negative than > positive effects. Negative effects came from the fact that people don't > think how to make good constants (and therefore don't think of their > warrior's specifics and structure), they just bruteforce them, and that > some people just recycle strategies. I'm very unhappy about your jugdement on optimax. If someone is new to corewars he think about new ideas and strategies. Maybe he test a very good idea but had no idea how to choose good constants. In this case optimax can help to find good constants to see if that new idea can be successfull. Another example is my article about 476. In fact warriors are composed of more than one part. If you took the silk from 3(p)c and i change the decoymaker to an quickbomber it is a new warrior. So what is wrong ? > > I classify authors in this manner: > - Handcoders (who may use optimization to improve their warrior's > performance) > - Evolvers, who _usually_ have bad redcode knowledge and do not write > a single line by themselves > - Optimizers, who optimize-to-death already existing structures > instead of writing new warriors, maybe multiple times! (I'll not point > my finger ...) > > Tuc suggested using special hill, but my opinion is "no": given a load > file, you can't objectively distinguish between optimized and > non-optimized warrior, so there is only one choice: don't use optimax, > but write a good warrior. Impossible, you say? Look at the hills then. > I don't use optimax, I don't optimize for weeks, but I dominate 94, > 94nop, 88 and LP. I use simple program to replace !a b! string with > random number in [a..b] range + my own optimization techniques + > corestep (or my own similair program) to test some selected constants, > and CoreWin to test randries (usually no more than 1000). So my > warriors are live prove that there is no need for NoOpt hill. > > The problem is that spirit of corewars changed :( Optimize-to-death is > not the thing that was envisioned by Dewney. You don't need to run > 10^F#$%INGHUGEPOWER instructions to make a good warrior. That very > popular "clone_other_warrior -> add_modern_quickscan -> > reoptimize_ten_times -> submit_to_the_hill" thingy is JUST F*&^ING > AWFUL!!! So, I suggest for all optimizers to repent and become > handcoders (or quit). I'm sad about you wan't to boot out members of our small community only if they don't play the game how you understand it. But think, the competition is not only paper - scissor - stone. If you are good enough to make good constants for your warrior you are good enough to make good constants against "high-optimized" warriors, too. Iirc corewars has its roots in the time where computerworms are new on the spot. Artificial life seems one of the influences too. I don't thing that A.K. Dewney has thought what cw will be in 20 years. So it is uselees to think about his conceivability. O.K, clone a warrior and only reoptimize is awful. But when some changes are made, sometimes one instruction shorter or longer, than it is a new warrior. I think it's o.k if such a warrior will be optimized and send to the hill. > > Simple example of how abovementioned approach harmed corewars: people > started to optimize dat-throwing papers against imps, but no one tryed > to understand how these heavy-optimized papers work, what are they, why > they loose to Moore-style papers, and why they beat imps. So huge > subclass of silkish papers was just left unnoticed. > > I want to see corewars as something more than tweaking quckscans (wrote > by others) and seeking good constants. There are still a lot of new > strategies to explore even at 94nop, believe me! > > "The net is vast and infinite" Discover new strategys is a hard job. Most of us are don't have the time for do that. Sascha -- News From: "Fluffy" Subject: Re: optimization equ evil Date: 28 Jul 2006 01:13:19 -0700 Message-ID: <1154074399.400328.131530@m79g2000cwm.googlegroups.com> Hi, > IMO optimax is a big problem nowdays, having more negative than > positive effects. Negative effects came from the fact that people don't > think how to make good constants (and therefore don't think of their > warrior's specifics and structure), they just bruteforce them, and that > some people just recycle strategies. I have no problem with recycling strategies. It increases the odds, that someone finds a warrior which exploits that. Do you remember Type-1 on nano? :-) > - Evolvers, who _usually_ have bad redcode knowledge and do not write > a single line by themselves What's wrong with having no knowledge about Redcode? Writing a good evolver is really hard. (I've tried and so far failed!) > The problem is that spirit of corewars changed :( Optimize-to-death is > not the thing that was envisioned by Dewney. You don't need to run > 10^F#$%INGHUGEPOWER instructions to make a good warrior. That very > popular "clone_other_warrior -> add_modern_quickscan -> > reoptimize_ten_times -> submit_to_the_hill" thingy is JUST F*&^ING > AWFUL!!! So, I suggest for all optimizers to repent and become > handcoders (or quit). ??? It is just a game. There are many ways to have fun with it. > Simple example of how abovementioned approach harmed corewars: people > started to optimize dat-throwing papers against imps, but no one tryed > to understand how these heavy-optimized papers work, what are they, why > they loose to Moore-style papers, and why they beat imps. So huge > subclass of silkish papers was just left unnoticed. Some other example ... it seems, that we now know how to theoretically choose good step constants for a paper. But there is still no practical (!) way to find them apart from randomly testing values. That is why I use my own optimizer :-) > I want to see corewars as something more than tweaking quckscans (wrote > by others) and seeking good constants. There are still a lot of new > strategies to explore even at 94nop, believe me! I like tweaking (other people's) quickscanners and optimizing until I've found better constants. That is fun for me. I really like the idea of OptiMax. It is a nice tool. And if you don't like other people to use optimizers then there is always one way to prevent them from using optimizers further. Push their warrior off the hill! (I really should get onto 94nop again ;-) > "The net is vast and infinite" Nope. Regards, Fluffy From: "inversed" Subject: Re: optimization equ evil Date: 28 Jul 2006 09:19:14 -0700 Message-ID: <1154103554.073939.294900@i42g2000cwa.googlegroups.com> The point is not to don't use optimax at all, but to use it wisely: do not *misuse* or *overuse*. Optimization is the last thing you should think about, warrior's structure is much more important. Misuse - Using optimax when it's not needed. Finding scanner's boot distance with optimax is surely a misuse. It only matters against oneshots (maybe except scanners with zooom-trick or transparent decoy) and can be done by bisection with ~15 tests. For many scanners and stones, amount of possible constants is quite small, so you can do testing by hand, using Optima or Corestep. In fact, Optimax is most usefull for paperish warriors. Overuse - Spending much more time for optimization than needed. For a typical paper, after some optimization there are some top-ranking randries that have similair scores (and SD) against different strategies (with differences near 1..2 points). When that happens, there is no need to optimize further. Sascha: > If someone is new to > corewars he think about new ideas and strategies. Maybe he test a very good > idea but had no idea how to choose good constants. > In this case optimax can help to find good constants to see if that new idea > can be successfull. Handpicked constants are okay for that purpose: make some constants, test with graphical display to be sure that everything is fine (nice scan pattern, stone in pws works long enough, ...), change constants if it's not, then test it if it can be successful (against some testset or hill), and THEN finally optimize it if it looks promising. > Another example is my article about 476. In fact warriors are composed of > more than one part. If you took the silk from 3(p)c and i change the > decoymaker to an quickbomber it is a new warrior. So what is wrong ? With only 3 lines dealing with actual warrior's code, it's not a Making Of, but an Optimax Tutorial. Plus it overuses Optimax. Finding 3 decrement positions for tiny using 4.2 GHz for an hour is definately an overuse - I'm sure that same results could've been obtained much faster, especially with such powrfull PC - 5..10 mins. > I'm sad about you wan't to boot out members of our small community only if > they don't play the game how you understand it. They should stay if they're thinking they're acting right. > O.K, clone a warrior and only reoptimize is awful. But when some changes are > made, sometimes one instruction shorter or longer, than it is a new > warrior. I think it's o.k if such a warrior will be optimized and send to > the hill. If this one instruction affects warrior's structure and actual performance after booting, I count it as a new warrior. If it only puts some data in free fields or does any other insignificant change, I count it as clone. > Discover new strategys is a hard job. Most of us are don't have the time for > do that. "I have seven children, I have three jobs, I have no time to think, but I want to have something on the hill, so I'll just reuse some code". This has no point at all. Thinking and breif testing doesn't requires much time. And it's better to have 1 original warrior than 10 clones. And if you don't optimize that much, you'll have time to discover new strategies :) Pete: > But, one issues that I haven't seen anyone mention and as an idle observer > would love to see more conversation about is whether anyone has actually > analyzed the output of an optimizer (especially...I'm less interested in the > evolver idea, but I suppose that could be relevant too), to try to > understand and describe why one constant works better than another. Are the > constants only "better" with respect to the existing warriors being fought? > Or are some constants inherently better than others, and if so, why? You're wrong, there is a Score Surfaces project, along with some explanations on RGC: http://corewars.jgutzeit.de/score_surfaces/index.en.html Bvowk: > Ouch. You could just whisper it behind my back! We evolvers have > feelings too, we're not just redcode manipulating automatons you know! That's why I underlined word "usually". Not always, but more often that not. "Bad redcode knowledge" may be exaggerated, but handcoding for hills and evolving _usually_ don't mix. SimFlyer: > As for 'getting rid of' some of the corewar users for 'being bad', I > think the community is too small now.... I don't propose to forcedly get rid of 'baddies', I just openly ask everyone to think if they should do CW. For a community it's better to be small than to be impure. > there isn't a verification suite for Redcode, There is validate.red for '88. Also, http://vyznev.net/corewar/guide.html explains everything pretty well, and most MARSes behave exactly as described there (with exceptions of very old or those who use modifyed redcode). > That's what evolvers are after...or at least one of the goals. Heh, than try discovering implauncher with an evolver. For "real" coresizes, its mostly work for humans. Fluffy: > I have no problem with recycling strategies. It increases the > odds, that someone finds a warrior which exploits that. Do > you remember Type-1 on nano? :-) Yes, but that just doesn't feels right ... > It is just a game. There are many ways to have fun with it. Doing it that way is no fun at all. > And if you don't like other people to use optimizers then > there is always one way to prevent them from using optimizers > further. Push their warrior off the hill! Ok, you've asked for it >:) >> "The net is vast and infinite" > > Nope. ? From: achillu@tin.it Subject: Re: optimization equ evil Date: 28 Jul 2006 15:50:00 -0700 Message-ID: <1154127000.380206.8390@h48g2000cwc.googlegroups.com> Inversed, I don't think your points are wrong, but I think that you are pointing those out in the wrong way: > For a community it's better to > be small than to be impure. Well... some guy with funny mustaches said something similar last century in the 30s... but that was absolutely not a good start! I agree that Optimax should not be misused or overused, anyway it is up to people how to use or misuse or overuse one's own computing power. Actually, I often use my Athlon XP Mobile 803MhZ to play Railroad Tycoon 3; this is sure a misuse, anyway am I wrong or bad? > Doing it that way is no fun at all. Yes, I agree, I have no fun in using Optimax that way, too; anyway I know people that do have fun using Optimax as well as I know people that do not have fun playing Railroad Tycoon 3. Peace. I personally tried to use Optimax several times but I never got the feeling for it, maybe because I am a formal mathematician and I see nothing stochastic (measurable) in the randries. Nevertheless in Italy we say that "a donkey hitting random on a keyboard may eventually happen to write the Divine Comedy" so I do not think that Optimax is bad. Optimax is just an unaware tool, it may happen to write the "Divine Comedy" for corewars but Optimax alone will never have the feeling to understand that; the feeling is in the people reading the results. I agree that we risk that most people just does not want to read the results. I agree with most points you pointed out, but I am not as absolute as you are. People may happen to use the tool in the wrong way, and this is what usually happens when people is facing a new tool. This does not mean that the tool is stupid, this simply means that people should learn how to use this tool. Most of your points suggest the correct direction anyway and I think that most redcoders will follow your hints when using Optimax or any other optimization tool. I just completely disagree with your point of view on evolvers; they are some kind of meta-programmers. I think that you cannot meta-do anything if you cannot do; and meta-doing is always harder than doing, both formal and procedural. And an evolver program is some kind of aware of what is the result, and this is the opposite of the unaware Optimax. So I welcome any new tool for redcoders; if I happen to see people using the tools in the wrong way, well the answer is simple: I will use the tools in the correct way and I will have better results. Or I will use better tools. Fortunately, those are just tools and they do not harm the corewar foundations or the corewar community: they may just harm the warriors they produce; sure no one using them will never harm the warriors that I will produce in any way. So I do not complain if anyone is having fun in using tools to produce one's own crippled warriors; I will have my own fun in using the tools that I prefer (possibly starting with pencil and paper) to produce my own warriors. Possibly crippled... but that's the challenge! From: "martinus" Subject: corewars simulator overview Date: 29 Jul 2006 06:13:30 -0700 Message-ID: <1154178810.597660.91350@s13g2000cwa.googlegroups.com> Hello everybody, I have finally taken the time to write a short overview about the different mars simulators I have written/modified some years ago.You can now get exhaust-ma, qmars, and exMars from here: http://martin.ankerl.org/2006/07/29/corwars-simulators-overview/ Thanks to Pavel =8Aavara qMars should now compile in visual studio 2005 and comes with a project file. Martin From: Sascha Zapf Subject: Re: optimization equ evil Date: Sat, 29 Jul 2006 11:08:54 +0200 Message-ID: inversed wrote: > The point is not to don't use optimax at all, but to use it wisely: do > not *misuse* or *overuse*. Optimization is the last thing you should > think about, warrior's structure is much more important. I think that is a very restricted view of the usage of optimax. Be sure that i'm not only take all constants from a already written warrior and exchange it with the optimax random operators. That will be more evolving than optimizing. Papersteps are the only kind of constants where this "bruteforcing" is o.k. In all other cases i try to understand what the current constant does. In a stone for example. What does this step, when did he selfmutate the stone body, oder how long is the bombing run before start the clear. After i have understood that i write the formula down and insert the right operators. I don't understand your problem. Optimax only helps you to process benchmarks after a given pattern. It is kind of shell-script for renaming 5000 files taken from your digicam with the ugly dp00002703 etc - names. Not more. It only rates the randry against a given benchmark. Which benchmark and what to do is always done by me. If i see a strong weakness against pws and try to fix that und will have a nice html where i can see the consequence of my try to increase warrior strength against pws. There is nothing magic in it. It is just some kind of luxury for me. > > Misuse - Using optimax when it's not needed. Finding scanner's boot > distance with optimax is surely a misuse. It only matters against > oneshots (maybe except scanners with zooom-trick or transparent decoy) > and can be done by bisection with ~15 tests. For many scanners and > stones, amount of possible constants is quite small, so you can do > testing by hand, using Optima or Corestep. In fact, Optimax is most > usefull for paperish warriors. > > Overuse - Spending much more time for optimization than needed. For a > typical paper, after some optimization there are some top-ranking > randries that have similair scores (and SD) against different > strategies (with differences near 1..2 points). When that happens, > there is no need to optimize further. I can remenber runs which are running for hours with nice scores, and suddenly a warrior appears which scoren over 3% better then the rest. Find out the reason and use it for the further development is something that only happen with optimax ( for me ) So when i start optimax in the evening, why don't let it run until the next day when i came back from work? > > Sascha: > >> If someone is new to >> corewars he think about new ideas and strategies. Maybe he test a very >> good idea but had no idea how to choose good constants. >> In this case optimax can help to find good constants to see if that new >> idea can be successfull. > > Handpicked constants are okay for that purpose: make some constants, > test with graphical display to be sure that everything is fine (nice > scan pattern, stone in pws works long enough, ...), change constants if > it's not, then test it if it can be successful (against some testset or > hill), and THEN finally optimize it if it looks promising. > >> Another example is my article about 476. In fact warriors are composed of >> more than one part. If you took the silk from 3(p)c and i change the >> decoymaker to an quickbomber it is a new warrior. So what is wrong ? > > With only 3 lines dealing with actual warrior's code, it's not a Making > Of, but an Optimax Tutorial. Plus it overuses Optimax. Finding 3 > decrement positions for tiny using 4.2 GHz for an hour is definately an > overuse - I'm sure that same results could've been obtained much > faster, especially with such powrfull PC - 5..10 mins. No one can rate a score of 145 after 10 minutes and 8000*8000 possible kombiantions. After 2.5 Years with optimax i know that on 94nop after 3 hour and two optimized constants nothing epoch-making will happen. On tiny after 30 minutes. But i don't wan't to sit beside my tower to stop after 15 minutes only because of self-discipline. > >> I'm sad about you wan't to boot out members of our small community only >> if they don't play the game how you understand it. > > They should stay if they're thinking they're acting right. > >> O.K, clone a warrior and only reoptimize is awful. But when some changes >> are made, sometimes one instruction shorter or longer, than it is a new >> warrior. I think it's o.k if such a warrior will be optimized and send to >> the hill. > > If this one instruction affects warrior's structure and actual > performance after booting, I count it as a new warrior. If it only puts > some data in free fields or does any other insignificant change, I > count it as clone. It is very hard to jugde about significant's while chaning something. If i put a constant to a free field to reduce a loop it is a great change. If i free a b-field that was carrying a bombing step and change it to an imp-gate it will have a big effekt on enemys-imp. So even it is only one changed adressing modes it can be a big and very significant change. > >> Discover new strategys is a hard job. Most of us are don't have the time >> for do that. > > "I have seven children, I have three jobs, I have no time to think, but > I want to have something on the hill, so I'll just reuse some code". > This has no point at all. Thinking and breif testing doesn't requires > much time. And it's better to have 1 original warrior than 10 clones. > And if you don't optimize that much, you'll have time to discover new > strategies :) As written above: There is no need to sit beside the PC when he is optimizing. So i don't have more time when not optimizing that much. Discovering new strategies is the kombination of deep thoughts and long cdb-session which can't be done by the way while working or care for three children ( I have three children between 3 and 11 ) As one consequence of this discussion i have removed my two warriors on the 94nop, because that where only tests for new snippets and the original warrior was changed less then 20%. You will noticed that your Warrior are koth and vice-koth again. Congrats.. I will have a deep look onto my 94draft warriors too if i will find the sources. For tiny only "Endless Power" is kind of that clones + misusing optimax, so i will remove it too. Next point is - if i really write optimax2 i don't know if i should publish it. The consequence of that would be that a big part of newbies to that game won't have success on of the hill beside sal's beginner and will withdraw disappointent before they understand the game enough writing good warrior without tools like optimax or mopt etc. My personally way to play cw is that when i see a warrior or a snippet which can make a kind of warrior better, or decrease weakness of strategie a against b i looks for good warriors as a base for implentation. I insert and change them and then i will give that in optimax' hand's. But if i must think about a complete new warrior i will never find the time to do that. There is alive beside cw, you know? -- News From: "fizmo" Subject: Re: optimization equ evil Date: 29 Jul 2006 14:44:52 -0700 Message-ID: <1154209492.349426.284710@m73g2000cwd.googlegroups.com> Hi folks, back from vacation now and looking now through rgc. Just my two cents in this discussion: Before the release of optimax people wrote their very own, _secret_ optimizers to dominate the hills. Sascha and I just changed this fact to make the very first cw optimizer public available. So one can say we care about equality of opportunity in corewar ;-) How the people use optimax is there very own risk. So please don't outlaw this helpful tool. inversed: I find your opinion somehow a bit narrow-minded, because a good idea not necessarily make a good warrior and a good warrior isn't necessarily based on a good, unique idea. Usually the success lies also in the numerous minutiae inside a warrior, which are usually as hard to find as a new ideas. Optimax usually helps saving time to figure out how well the idea really is, and if the idea looks promising to save time to make it competitive for the hill. To be honest, my most creative time in corewar came after I started using optimax and I really extensively use it until that time. So, it seems that I am in inversed's classification just a dumb optimizer ;-) Anyway, as Jens already stated, corewar is just a game and everybody understand the game different. I think nobody should judge about other players understanding of this game...... Fizmo From: dhillismail@netscape.net Subject: Re: AI warrior. Date: 29 Jul 2006 19:01:22 -0700 Message-ID: <1154224882.680221.96730@m73g2000cwd.googlegroups.com> SimFlyer wrote: > On 26 Jul 2006 23:14:41 -0700, wrote: ... > Go hang out on the nano hill. I was fooling around with some numbers > the other day and it seems that you could test every possible warrior > for that hill in about 18 months (for my machine). Hmm... I did some back of the envelope calculations double D = 16*80*80*7*7*8; Console.WriteLine(D); // 40 million possible lines of redcode on Nano Dn = D*D*D*D*D * 5; // require 5 lines of code, 5 possible starting lines Console.WriteLine(Dn); // 5.2E+38 possible warriors on Nano double Ds = 60*60*24*30*18; Console.WriteLine(Ds + " secs per month"); // 46 million seconds in 18 months D = Dn/Ds; Console.WriteLine(D + " warriors per sec"); // 1.1E+31 warriors evaluated per second and it looks to me like you'd need to test 10 raised to the power of 31 warriors every second. What kind of computer do you have? ;-) Dave Hillis From: dhillismail@netscape.net Subject: Re: AI warrior. Date: 29 Jul 2006 19:14:50 -0700 Message-ID: <1154225690.585911.177300@m73g2000cwd.googlegroups.com> Cristian210273 wrote: > > > > The 5 instruction limit make that hill makes it a good place for > > evolvers, including the human based variants. > > > > 'Neural Networks' - wrong hammer for this nail. > > > Thank you! I will investigate more deeply..... > However there was a misunderstanding: > with "warrior using AI techniques" I meant > "warrior whose code implements some AI > inside, like fuzzy logic". You probably already know that Redcode is Turing complete, and also that while it allows multiple processes, they aren't simultaneous-there is no parallel architecture speedup. Whether one of these techniques can be fast and compact enough for a warrior is the answer to a more detailed question: what kind of neural net? on-line training?, etc. My suggestion is to look at P-space warriors and see if you find something that interests you. Dave Hillis From: "inversed" Subject: Re: optimization equ evil Date: 29 Jul 2006 19:18:55 -0700 Message-ID: <1154225935.116117.202340@m73g2000cwd.googlegroups.com> > As one consequence of this discussion i have removed my two warriors on the > 94nop, because that where only tests for new snippets and the original > warrior was changed less then 20%. You will noticed that your Warrior are > koth and vice-koth again. Congrats.. I really didn't wanted to offend someone personally, neither to start ;killing epidemy. I think you should send those warriors back (20% can make a huge difference + I don't enjoy being KOTH in that way + people will blame me for killing them). > Anyway, as Jens already stated, corewar is just a game and everybody > understand the game different. I think nobody should judge about other > players understanding of this game...... Oh okay, at least you got my PoV. I just want the community to be creative. From: sayembara@gmail.com Subject: Re: optimization equ evil Date: 29 Jul 2006 20:33:51 -0700 Message-ID: <1154230431.490755.167940@b28g2000cwb.googlegroups.com> I don't use Optimax, but I welcome everyone who wants to use it. In fact, not only Optimax, but other kinds of optimizations. I prefer to see dynamic hill than static hill. I look Optimax as an "automatic constant changer" where traditionally (as what I am doing right now) players change the constant manually, until the constant is best enough. Of course there are some that use some sort of programming. After that warrior is kicked off the hill, the constants are worked on again and resubmit for second chance of getting on the hill; or the structure is modified and the same cycle starts again. From: Sascha Zapf Subject: Re: optimization equ evil Date: Sun, 30 Jul 2006 11:08:58 +0200 Message-ID: inversed wrote: >> As one consequence of this discussion i have removed my two warriors on >> the 94nop, because that where only tests for new snippets and the >> original warrior was changed less then 20%. You will noticed that your >> Warrior are koth and vice-koth again. Congrats.. > > I really didn't wanted to offend someone personally, neither to start > ;killing epidemy. I think you should send those warriors back (20% can > make a huge difference + I don't enjoy being KOTH in that way + people > will blame me for killing them). I've killed the warriors because in some points you are right. Both were only tests - successful warriors just with an added stone and a new decoymaker. I want to see how do they score against this Hill. At home on my fsh they score poor ~135. I was very surprised about being koth with s-test. But it's not o.k for me i a have felt bad when i think to the day when i decide to publish the source. I don't want to blame you - i only try to not blame me. > >> Anyway, as Jens already stated, corewar is just a game and everybody >> understand the game different. I think nobody should judge about other >> players understanding of this game...... > > Oh okay, at least you got my PoV. I just want the community to be > creative. Creativity in a 20 year old game can be a free b-field in the 5liner-stone class for an imp-gate, or a better way to design a centersilk. I'm sury there are some major-strategies left to discover but i rate simple improvements as innovation too. I personally prize your influence very high. I was very impressed after you have publish your thoughts about the score-surfaces. Your Warriors are unique and very special. I'm happy that someone like you is part of our small community. I do not force the player to use only the random operator in optimax. Optimax is only a worker. Sitting beside the editor and calculate good steps for a scanner is hard enough - why not use !{241|1023|2027|3049} for testing them instead of open scanner.red with editor change constant save mts Subject: Re: AI warrior. Date: 30 Jul 2006 23:15:18 -0700 Message-ID: <1154326518.195548.167200@s13g2000cwa.googlegroups.com> > > You probably already know that Redcode is Turing complete, > Dave, please help me! What does this mean? (Sorry if this question sounds silly...) > Whether one of these techniques > can be fast and compact enough for a warrior is the answer to a more > detailed question: what kind of neural net? on-line training?, etc. > My suggestion is to look at P-space warriors and see if you find > something that interests you. > Yes, my idea was to find an AI tecnique compact enough to be coded in a warrior (I was thinking at fuzzy logic: a membership function should not be so heavy in terms of number of lines of code....). I will take a lokk at P-space: can you point me to a reference? Thanks! From: KOTH Subject: KOTH.ORG: Status - MultiWarrior 94 07/31/06 Date: Mon, 31 Jul 2006 00:19:13 -0400 (EDT) Message-ID: <200607310403.k6V430GV039254@asgard.t-b-o-h.net> Weekly Status on 07/31/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Multiwarrior 94 CoreWar Hill: Last battle concluded at : Mon Jul 24 12:25:15 EDT 2006 # Name Author Score Age 1 Urgle Daniel Rivas 38 7 2 Fluffy Paper VI Jens Gutzeit 38 27 3 kingdom of the grasshoppe simon wainwright 32 122 4 JustADirtyClearTest Nenad Tomasev 27 56 5 the price of hostility John Metcalf 24 14 6 CLP-shot again G.Labarga 23 1 7 Diptera Nenad Tomasev 22 41 8 simply believe John Metcalf 22 4 9 rooftop pursuit John Metcalf 22 3 10 nameless fragment S.Fernandes 15 26 11 shorty krister 0 0 From: KOTH Subject: KOTH.ORG: Status - ICWS Experimental 94 07/31/06 Date: Mon, 31 Jul 2006 00:19:12 -0400 (EDT) Message-ID: <200607310406.k6V460pn039464@asgard.t-b-o-h.net> Weekly Status on 07/31/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG ICWS Experimental 94 CoreWar Hill: Last battle concluded at : Sat May 27 16:29:52 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 46/ 35/ 19 Fatamorgana X Zul Nadzri 157 10 2 44/ 36/ 20 The X Machine Zul Nadzri 152 34 3 44/ 36/ 19 Ogre Christian Schmidt 152 171 4 37/ 22/ 42 xd100 test David Houston 152 20 5 28/ 11/ 61 Evol Cap 4 X John Wilkinson 146 292 6 42/ 40/ 18 Eliminator X Zul Nadzri 145 35 7 41/ 39/ 20 Bewitching S.Fernandes 144 1 8 34/ 25/ 41 KAT v5 Dave Hillis 142 155 9 39/ 37/ 24 Trefoil F 13 Steve Gunnell 140 106 10 35/ 29/ 36 Olivia X Ben Ford 140 104 11 41/ 43/ 16 Giant Hazy Test 13 Steve Gunnell 140 50 12 41/ 43/ 15 O_Fortuna3X Nenad Tomasev 139 6 13 40/ 42/ 18 Black Moods Ian Oversby 139 219 14 38/ 38/ 23 Trefoil Test F 14 Steve Gunnell 139 7 15 37/ 37/ 26 test Some Redcoder 138 4 16 21/ 6/ 73 Evolve X v4.0 John Wilkinson 137 240 17 38/ 41/ 21 Controlled Aggression Ian Oversby 135 223 18 31/ 28/ 41 Glenstorm John Metcalf 135 85 19 38/ 42/ 20 Simply Intelligent Zul Nadzri 134 16 20 39/ 46/ 14 Fatal Choice Some Redcoder 133 5 21 8/ 79/ 12 Reaper II Slaytanist 38 0 From: KOTH Subject: KOTH.ORG: Status - 94 No Pspace 07/31/06 Date: Mon, 31 Jul 2006 00:19:11 -0400 (EDT) Message-ID: <200607310409.k6V491xA039789@asgard.t-b-o-h.net> Weekly Status on 07/31/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG 94 No Pspace CoreWar Hill: Last battle concluded at : Sun Jul 30 13:41:22 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 34/ 23/ 42 Monster_Human_Grunt inversed 146 270 2 30/ 16/ 54 8kSquare Roy van Rijn 143 6 3 33/ 24/ 44 Monster_Alien_Grunt inversed 142 271 4 33/ 26/ 40 Amicable Antagonist Roy van Rijn 141 5 5 34/ 27/ 40 Hullab3loo Roy van Rijn 140 185 6 28/ 16/ 56 D3vilstick Roy van Rijn 140 184 7 42/ 44/ 14 Pro Radii inversed 139 15 8 29/ 19/ 52 Last Judgement Christian Schmidt 139 455 9 30/ 21/ 48 Bluebell Christian Schmidt 139 158 10 30/ 22/ 48 Song of the blue sea Miz 139 142 11 28/ 17/ 55 Burning Metal inversed 138 10 12 30/ 21/ 49 MoonOfChaos Nenad Tomasev 138 233 13 39/ 40/ 21 Gabble Christian Schmidt 138 36 14 33/ 28/ 40 N e i t h inversed 137 31 15 38/ 40/ 22 Twilight S.Fernandes 136 155 16 30/ 24/ 46 Amber inversed 135 9 17 40/ 48/ 12 Hallucination Scanner inversed 132 143 18 38/ 45/ 16 ChimeraQueen Nenad Tomasev 131 150 19 26/ 22/ 52 P Fluffy 130 2 20 27/ 53/ 20 spaghetti coding John Metcalf 101 1 21 12/ 84/ 4 tester Spiritofsparta 39 3 From: KOTH Subject: KOTH.ORG: Status - Standard 07/31/06 Date: Mon, 31 Jul 2006 00:23:15 -0400 (EDT) Message-ID: <200607310400.k6V400C3039077@asgard.t-b-o-h.net> Weekly Status on 07/31/06 -=- irc.KOTH.org is up! Meetings held in #corewars -=- Tons of new features on www.KOTH.org/koth.html pages -=- *FAQ* page located at: www.KOTH.org/corewar-faq.html Current Status of the KOTH.ORG Standard KotH CoreWar Hill : Last battle concluded at : Sun Jul 30 19:26:54 EDT 2006 # %W/ %L/ %T Name Author Score Age 1 34/ 18/ 49 Raging Gale '88 inversed 149 12 2 38/ 27/ 35 The Next Step '88 David Houston 148 83 3 39/ 31/ 30 White Fire inversed 148 9 4 35/ 23/ 42 Test Alexander (Sasha) Wa 147 207 5 44/ 44/ 12 Monolith inversed 145 16 6 33/ 23/ 45 The Hurricaner G.Labarga 143 54 7 39/ 37/ 23 My 1st try Christian Schmidt 142 99 8 41/ 42/ 17 Hexamorph inversed 141 11 9 41/ 40/ 19 Scan the Can Christian Schmidt 141 48 10 33/ 26/ 41 Triangular Sun inversed 141 8 11 34/ 29/ 37 usual fee, plus expenses John Metcalf 140 6 12 35/ 30/ 35 The Seed Roy van Rijn 139 85 13 35/ 33/ 32 Vampire Knight inversed 138 7 14 27/ 16/ 57 Utterer '88 Christian Schmidt 137 42 15 31/ 26/ 43 m88sib09tst inversed 136 1 16 38/ 39/ 23 Moonwipe Christian Schmidt 136 64 17 41/ 46/ 13 Scan Test C 6 Steve Gunnell 135 106 18 30/ 25/ 45 Guardian Ian Oversby 135 267 19 30/ 25/ 45 test G.Labarga 134 50 20 31/ 28/ 41 SoundOfDarkness Nenad Tomasev 133 31 21 2/ 98/ 0 0 0 7 0 From: "Nenad Tomasev" Subject: Re: optimization equ evil Date: 31 Jul 2006 04:56:26 -0700 Message-ID: <1154346986.524554.247270@b28g2000cwb.googlegroups.com> Hi. I just came back from holidays (had been offline for almost a month), so I wanted to share my thoughts on this subject as well. When I started doing corewar, I spent a couple of months trying to make many warriors implementing different ideas that came to my mind or I had read about them in the tutorials without taking a single look into an existing warrior (even though I had downloaded some warrior archives). Many of them didn't work, but some have and even though they weren't what I would call now good warriors, doing so helped me understand how it all works. After that, I explored strategy by strategy, mostly through trial and error approach, typing all of the code myself, not copying from other files, handpicking the field values... After that, I started using some of the existing components (qscans and imp launchers for instance), but most of the code I still wrote myself. In the end, some of the last warriors I had made and sent to the hills weren't very original and I was just searching for ways to improve some things in older warriors. I had always taken a greater pleasure in making something new, and I even had good success on the hills with warriors that weren't optimaxed at all, like KryneLamiya, Jotunheim, ()()(), DifferentialOperatorWS and some more, and also with those only slightly optimaxed like MoonOfChaos (an hour of optimaxing or less, as I remember). However, I also see nothing wrong in using optimax to improve your existing warriors. If you can find better values, why not do so? For most of the time that I did corewar, I was handpicking the values, looking at the visual display of CoreWin, trying to make conclusions concerning why some steps in stones/scanners work better than others, why are some paper steps good and other not (also peeking at the process count scaled by time, etc.), and it did help me. But it also takes a lot of time if you need to do it for every warrior you make. I still don't optimize oneshots and some other strategies, but isn't it much faster to let optimax find a good stone step in a new CORESIZE than go through a couple of hundred steps before finding one manually? Or maybe run some other auxilary programme like corestep? This game is, among other things, also about having fun. Not about getting totally exhausted... Of course, I did quite the opposite for a while, but back then, I had the time to do so. To spend a couple of days doing almost nothing but corewar, experimenting, implementing new ideas, searching for steps by hand, challenging the hills, etc. Some of the best steps in my warriors, I found that way. The scan step of Chimera, for instance, that was used in LuxAeterna later on... One other thing that deserves to be mentioned is that if you don't have enough knowledge on how things work, you can't make a good warrior with optimax. You have to be very careful about test warriors and warrior groups, thresholds and stuff like that. Also, you need to know which things you should optimize together and which separately, to reduce the size of the problem space. After I started using optimax, it took me some time to gain the experience to do these things the way I think now is good. Optimax is a great tool, if you can use it correctly. I have nothing against other people using it. I also don't expect every beginner should start the way I did, it just wouldn't work that way. I was enthusiastic from the start and had a lot of energy to drive me to keep working, keep testing, etc. Most of the people that start playing corewars are interested, but not that much. And it's not a bad thing. Let them choose their own way to do things. Let them test the existing warriors, see how they score when fighting each other, change something, see how the change reflects on the score, change something else, etc. Of course, out of all these warrior clones, only some might get submitted to the hills. ... I don't think that people submit so many things which aren't, at least partially, original. And being original is getting harder and harder each day... There are many ways of playing corewar, and it's not up to me to say "that one's wrong". It's only a game. As for the evolvers, I think that haveing some around is a great thing. I always wanted to make my own evolver but just didn't find the time to do it. I'm not only interested in corewar as a competitor on the hills, but I also likt to view it from the perspective of ALife and life sciences (I was close to enrolling into biology studies three years ago, but changed my mind in the last minute and decided to study math and computer science instead). I don't think that it's meaningfull or nice to comment on the coding skill of evolver makers, because the part they play in this show is quite a different one. Their goal is not only to make good warriors, but to make something that makes good warriors, as much as I understand, not because they want good warriors as the result, but because they want a good evolver as a result! They seek not the warrior, but the warrior making process, and that's something completely different, and also very interesting. In the end, I don't think that separating people here into groups of those who optimize little or a lot would help the community or any of its members. Making a good warrior takes skill, understanding and a good idea. Optimizing a warrior after you've made a good basic shell is also not a simple thing to do. It requires understanding... and many beautiful warriors have their current form because of the optimization process. A good idea, in a way, deserves to be optimized and to be elevated to its rightful place. But that's just my personal opinion on the matter. :) Sadly, I didn't have time for corewar for quite some time now... and probably won't be active much in the following months, as well. However, I *will* do my best to revive CoreExplorer soon and to make a couple of warriors when I'm taking break from leraning for the exams ahead. Stay well! Regards, Nenad From: dhillismail@netscape.net Subject: Re: AI warrior. Date: 31 Jul 2006 18:13:51 -0700 Message-ID: <1154394831.305318.282520@i3g2000cwc.googlegroups.com> First, since you posted the question at all, it means you would probably enjoy playing Corewars. Even if what you have in mind didn't lead to any successful warriors, similar things would. Cristian210273 wrote: > > > > You probably already know that Redcode is Turing complete, > > > Dave, please help me! What does this mean? I'll say it a different way. Warriors are written in an assembly-like language called 'Redcode.' It is a complete language, so pretty much anything you can write in a "normal" computer language, you can write in Redcode (somehow...). > (Sorry if this question sounds silly...) You are not asking silly questions. For general questions, the best way to answer is the way SimFlyer did: encourage reading the docs. More specific questions tend to get more satisfying answers. > > Whether one of these techniques > > can be fast and compact enough for a warrior is the answer to a more > > detailed question: what kind of neural net? on-line training?, etc. > > My suggestion is to look at P-space warriors and see if you find > > something that interests you. > > > Yes, my idea was to find an AI tecnique compact enough to be > coded in a warrior (I was thinking at fuzzy logic: a membership > function should not be so heavy in terms of number of lines > of code....). I will take a lokk at P-space: can you point me to > a reference? > Thanks! A good way to start is to go to http://koth.org/ look there, scroll down to the Corewars Web Ring and look at the many guides you will find on the various sites. You'll see info on P-Space. It isn't the first thing to learn but I will say a little about it. When 2 warriors compete, they fight a sequence of battles. Each battle starts with the warriors in a clean core (memory space) which they then mess up while trying to kill each other. P-Space is persistent memory each warrior has that only they can read or write into, so they can pass information to themselves from one battle to the next. One way to use this memory is a p-switcher. For instance, a warrior might play one strategy during one battle, if it wins it tries it again, if it loses it tries something else next battle. More elaborate strategies are often encoded into finite state tables. Your fuzzy rules might fit in for something like this. I said that the other warrior can't write into your warrior's P-Space. But it might trick your own warrior into writing harmful information into it. That's called "brain washing." Happy reading! Dave Hillis From: "Kim B�ndergaard Poulsen" Subject: PCRobots variant Date: Mon, 31 Jul 2006 23:23:22 +0200 Message-ID: <44ce72e9$0$67259$157c6196@dreader2.cybercity.dk> Hi Corewar gamers Know this isn't the right group, but hope someone can help me anyway Years ago I had a kind of PCRobots look a like dos program consisting of both an integrated editor, compiler and graphical player engine. The language used was something proprietary but never the less good enough when introducing a newbee to the world of programming (as far as I remember) Anyone out there who know what I'm thinking of? /Kim