Round 2 Author Strategy Lines and Comments

;name Genetic Beef
;author M Joonas Pihlaja
;strategy clear with strip bombing - sort of.

Early on, a quickish forward 0,0 clear cripples the enemy in front of the
warrior, and pretty soon it's joined by a backward clear that finishes
covering the core.  Works better than head on random scrambling.

Adds or subs make for a better attack than movs, since they create better
clear bombs of the increments and decrements.  This also means that if 
the body is continuously overwritten by some constant (e.g.  by a clear or
HSA), then the warriors only hope is changing it's clear pointers radically
before we are overwritten again.  With mov.f/x, the pointers are simply
reset with something, whereas with add/sub, they are incremented => 
better coverage.

I think subs have only one advantage over adds, and that is the possibility
of subtracting a freshly overwritten (by the enemy) piece of the code
by whatever it was overwritten by, thus undoing the enemy's work.  It's
not as foolproof as I would like it to be, but it does help some.

A genetic algorithm bred the proper mix of spls, subs, and addressing modes
to get the most punch out of each line.  Each candidate was benchmarked
against a test suite of 16 warriors consisting of 2 backward clears and
oneshots, 4 forward (d-)clears and oneshots, four scanner variants (HSA,
Neverland, jmz.f blur -> d-clear, and The Machine), three stone variants
(Rosebud, Newt, .66 bomber->clear), and finally an all 20 spl 0,0 
warrior.  The two remaining positions were filled by the top two warriors of
the previous generation, which hopefully gives the proper amount of
resistance to other similar warriors that may come up.

At *least* it doesn't suicide. :)


;name Stabbing Forward
;author Christian Schmidt
;strategy no .i, no #, mmmhhh...
;strategy the most effective strategy for this round seems to be a coreclear.
;strategy Ok, let's programm an a-field incrementing d-clear with a maximum
;strategy increment/decrement protection. I know that this isn't the only
;strategy clear in the round, so I thought about how can I get an advantage.
;strategy And the answer is: make it as fast as possible :-)

;name Waxer
;author David Moore
;strategy dirty clear

;name backslider
;author Steve Gunnell
;strategy 1.33C linear scanner not triggered by djn streams.
;strategy Limited resistance to increment/decrement streams.
;strategy Tried scanning fowards as well. This was better.
;strategy Tried various decoy generators etc. Bleagh...
;strategy Should beat core clears and get eaten by cmp/add scanners.

;name Seeing Eye
;author Robert Hale
;strategy p-space scanner and duct tape.
;strategy scanner is 66% due to space limitations.
;strategy duct tape is a self repairing core clear.
;strategy Did you know you can use duct tape to repair duct tape?
;strategy A 80% oneshot scanner worked better against my test suite 
;strategy but I hope this is a better all around warrior.

;name Cryton
;author Simon Duff
;strategy Simple Bomber except ADD to A and B fields rather than
;strategy drop DAT or SPL bombs

;name Nibble 
;author Robert Macrae
;strategy Blur-like scanner
; Steps forward at 2.5c linear with spacing of 5.
; Based on NiBl, but with DClear carpet. Vuln to 1,1 bombs.

;name H-Bomb
;author Josh Yeager
;strategy Two cc's running in parallel.

;name 2 Fast 
;assert 1
;author Paul-Virak Khuong
;This is the pattern--^
;I've been thinking of a 67% scanner/bomber. 
;It wouldn't have been that bad.
;A bit like leprechaun. But, why use a scanner part? ;After all,
there's no paper.......
;Wish i haven't just made an error. . . . . . .

;name Censorship
;author Philip Kendall
;strategy A CLP variant

for 0

Ian Oversby's Tournament Round 2: Tiny hill, but 16000 cycles and
no .i or # allowed.

The idea behind this warrior is that the main form of attack used by
many warriors will be a forward clear, probably at 1c. Therefore sit
around and wait until a location (spy) is changed, then boot the my
clear in behind the opponent's clear, then proceed to wipe core and
kill him... hopefully. Also, build a decoy just behind myself at
start-up to trigger as many one-shots as possible, and stick a
dat.f 1,1 at the end so any one-shots with djn streams (say like
my One Shot 'T') will switch to a clear when they hit that, and
then fall into my trap :-)

What it's good against: one-shots and clears
What it's very bad against: stones (and backward clears)


;name Rostov
;author WFB
;strategy For the Oversby Tournament Round 2.
;strategy Ahh!  What to do?  Scanning, q-scan, maybe...
;strategy But how do you kill them?
;strategy Hmm...well, this is a worthless dwarf, which jumps down to a
;strategy core clear (kinda...) if it's hit (sometimes...).

;name Misplaced Optimism
;author A. S. Mehlos
;strategy Submission for round 2 of Ian's tournament:
;strategy I've got _no_ idea how this guy's gonna do, since I 
;strategy really only had my own other warriors to test it 
;strategy with. As I see it, with no "mov.i"'s to deal with,
;strategy there's no good reason not to make every one of the 
;strategy 20 instructions valid. This should be interesting, though.
;strategy (Though I fear I'll do worse this round than last.)