Wilkinson Ratings

OK in an effort to kickstart a standard warrior rating system I have adopted the Wilkinson Benchmark to rate all my warriors, and this has now been extended to a more general selection of warriors culled from Planar's archive. Also I'm going to rate warriors which I haven't released yet so that people can see how they compare here.

After originally creating this page I changed my mind about the inclusion of self fights in the standard so this page is being converted to the new rating system, for comparison you may want to check out the old page.

Also Steve Bailey and Beppe Bezzi have published a load of ratings presumably based on the same benchmark so I'm going to include those here in brackets. Similarly if anyone has any ratings they have compiled then please send them to me, no cheating though, I will be watching ;-) Send the warriors along with your scores and they'll definetely be included

  1. 180.1 Wind Up Toy v 0.7 - Ian Oversby
  2. (170.3) Frontwards v2
  3. (168.7) Naked Dancer
  4. 165.8 Gem of the Ocean - P. Kline
  5. 164.9 Wind Up Toy v0.4 - Ian Oversby
  6. 163.4 Porch Swing - Randy Graham (168.0)
  7. 164.8 Derision - M. R. Bremer
  8. 160.3 Endpoint . - M. R. Bremer
  9. 159.9 Mirage 1.5 - Anton Marsden
  10. 159.2 Quiz - Schitzo (159.6)
  11. 157.2 Jack in the Box - Beppe Bezzi (155.8)
  12. (156.3) Tornado 3.0
  13. 155.6 La Bomba - Beppe Bezzi
  14. 151.6 C I A - Anders Ivner
  15. 151.3 Impfinity v4g1 - Planar (149.4)
  16. 150.0 Marcia Trionfale 1.3 - Beppe Bezzi (150.4)
  17. 147.2 Nobody Special - Mike Nonemacher (145.8)
  18. 147.8 Armoury A5 - J. K. Wilkinson
  19. 147.1 Timescape 1.0 - J. Pohjalainen (149.4)
  20. 147.0 Hector 2 - Kurt Franke
  21. 146.4 Persistence - Kurt Franke
  22. 143.9 Torch t18 - P. Kline (140.3)
  23. 143.2 Blizzard - Anton Marsden
  24. 142.6 Night Train - Karl Lewin
  25. 142.5 Phq - Maurizio Vittuari
  26. 142.4 Door Mat v0.1 - K Lewin
  27. 142.3 Harmony - P. Kline
  28. 142.0 Seventy Five - Anders Ivner
  29. 141.8 Blue Funk 3 - Steven Morrell (142.2)
  30. 141.8 Pretentious v0.2 - Ian Oversby
  31. 141.6 Memories - Beppe Bezzi (146.9)
  32. 141.5 Time Lapse v0.1 - David Boeren
  33. 141.5 Lithium - John K. Wilkinson
  34. 140.6 Clisson Lite - P. Kline.
  35. 141.7 myConfuser - Magnus Paulsson
  36. 140.7 SETI - John K. Wilkinson
  37. 139.9 Koolaid II: Wogg v2.2 - David Boeren
  38. 138.5 Die Hard - P. Kline
  39. 137.5 Juliet and Paper - M.R. Bremer & Beppe Bezzi
  40. 137.3 Paper One - Beppe Bezzi (138.9)
  41. 137.0 Leprechaun on Speed - Anders Ivner
  42. 136.8 anything box - schitzo
  43. 135.8 Aeka - T. Hsu (139.9)
  44. 135.3 Impfinity v 3i - Planar
  45. 134.8 Agony II - Stefan Strack
  46. 134.8 Harmony II - P. Kline
  47. 134.7 Time Lapse v0.8 - David Boeren
  48. 134.7 Crow - Karl Lewin
  49. 134.8 Babbo Natale - Maurizio Vittuari
  50. 134.4 myVamp v3.7 - Paulsson
  51. 134.0 Thermite - Robert Macrae (139.5)
  52. 131.2 Iron Gate - Wayne Sheppard (132.5)
  53. 129.2 The Mystery - Magnus Paulsson
  54. 126.8 Uvavu II revisited
  55. 126.7 Hyakutake Perihelion (Exclusive- Soon to be unleashed!)
  56. 126.6 Hyakutake C/1996 B2 +
  57. 125.8 Mason 2.0 - Robert Macrae
  58. 125.8 Qwiksand - Wayne Sheppard
  59. 125.5 Hyakutake C/1996 B2
  60. 124.3 Hyakutake Zenith
  61. 123.3 Cannonade - Paul Kline (123.5)
  62. 123.0 Extreme Prejudice
  63. 121.4 Rave - Stefan Strack(120.1)
  64. 121.0 You Wouldn't Let It Lie!
  65. 119.7 Judge Nutmeg
  66. 119.7 You Wouldn't Let It Lie! 1.01
  67. 118.4 127 point Imp Spiral - A. MacAulay
  68. 115.8 Uvavu II
  69. 113.2 Hyakutake Approaches
  70. 111.6 Fire Storm v1.1 - W. Mintardjo (107.6)
  71. 110.8 Hyakutake Rising
  72. 109.8 Flashpaper - Matt Hastings (110.4)
  73. 105.7 Uvavu P!
  74. 103.7 Pensive v0.3 - Ian Oversby
  75. 101.8 Hint Test V.4 - Beppe Bezzi
  76. 99.6 Aleph 0 - Jay Han (104.7)
  77. (90.0) Tonto 3 - Steve Bailey
  78. 88.5 Cyclone
  79. (87.2) Tonto 1 - Steve Bailey
  80. (87.1) Mice
  81. 85.5 Tornado - Beppe Bezzi (85.4)
  82. 83.5 Alien Kiss V1.1 - Bjoern Guenzel
  83. 81.5 Pensive v0.1 - Ian Oversby
  84. 81.3 3 Point Imp Spiral - MFCWB
  85. 77.5 Hyakutake Engine
  86. 72.7 Pensive v0.2 - Ian Oversby
  87. 71.5 3 point Imp ring - P. Kline or A. Ivner?
  88. 71.3 Uvavu V1.01
  89. 60.8 Uvavu
  90. (60.5) Cleaner
  91. 58.1 Enlightenment II
  92. 56.7 Enlightenment
  93. (56.0) Dr. Frog
  94. 52.1 Worm - MFCWB
  95. 50.3 Dwarf - A. K. Dewdney (49.4)
  96. 49.4 Deathwalker
  97. 48.7 Cleaver/75 - Wayne Sheppard
  98. 48.5 Imp!!!!! Yes the Original 1-point imp. - A. K. Dewdney
  99. 47.1 Eranu v1.03
  100. 46.6 Eranu V1.02
  101. (45.4) Chang1 (filename from Planar's archive)
  102. 44.9 Simple Scanner (test component for Hyakutake)
  103. 44.4 Deathwalker II
  104. 44.3 Uvavu 40(TNG)
  105. 42.8 Uvavu 20(TNG)
  106. 42.5 Uvavu 60(TNG)
  107. 39.3 Mutagen
  108. 38.0 Uvavu 76(TNG)
  109. 36.9 Uvavu (original)
  110. (34.5) Cancer
  111. 32.8 Ape - Calvin Loh
  112. 1.0 The original Imp Gate - B. Thomsen
  113. 0.8 Wait - MFCWB

Some warriors are credited to MFCWB which of course refers to Steven Morrell's book, where an author has been named in the book then that name is used instead. I don't have the details for all the warriors so if there are uncredited authors then please contact me.

The benchmark warriors are available from Planar's archive, and the rating process I have used is simply 200 battles with each warrior, the scores are then scaled to 100 battles (i.e 0-300).

Only 2400 battles is probably too small to justify the precision I have used in the ratings, but when I get time I will run more tests to increase the accuracy.

So now the question is how the benchmark can be improved, there are only 12 warriors, probably too small for an accurate benchmark, also should the number of rounds be altered.

My opinion is that since some PSpace algorithms are designed for the hills then the number of rounds should reflect the Hills. If more rounds are to be run for increased precision then the rounds should be split into groups of 200 but if extra rounds are going to be run then I'd prefer them to be against other opponents to decrease the dependance on strategy.

Also the standard warriors appear to be pretty poor at dealing with imps, I didn't think my early programs were so weak after all they did make it to the beginners hill, most of them even made the top 10. There's probably too many strong papers and not enough good scanners, so at the moment I'm going to stop running benchmarks until a more balanced standard appears. I'd welcome any suggested additions.

BACK

Scott Manley/spm@star.arm.ac.uk