Moderator: Ras
Code: Select all
# PLAYER : RATING POINTS PLAYED (%)
1 dorpsgek : 2195.0 976.5 1190 82
2 Embla-5142 : 1937.6 356.0 860 41
3 Embla-5140 : 1924.5 122.5 611 20
4 tscp-1.8.1c : 1627.1 135.0 279 48
5 CP-a090da4 : 1286.7 18.5 248 7 <---
6 CP-d031f3a : 1124.8 0.5 30 2
Looking at our lists you can see that strength is not a main concern for us.
Adjudication is not the same as resigning on own accord. Adjudication is like a referee saying:Gabor Szots wrote: ↑Sat Oct 16, 2021 8:56 am So in the case of Princhess the rating you find at CCRL is not an overestimation owing to adjudicating wins when it may not have been able to win. Either it checkmated the opponent or the opponent resigned of its own accord.
mvanthoor wrote: ↑Sat Oct 16, 2021 12:07 pmAdjudication is not the same as resigning on own accord. Adjudication is like a referee saying:Gabor Szots wrote: ↑Sat Oct 16, 2021 8:56 am So in the case of Princhess the rating you find at CCRL is not an overestimation owing to adjudicating wins when it may not have been able to win. Either it checkmated the opponent or the opponent resigned of its own accord.
Ref: "Hm... White is up 10 points of material here. Black, do you agree with thtat?"
Black: "Yes."
Ref: "Game over, white wins."
If you score games like that, and it's even combined with using Syzygy for mating (which would be a weak player consulting a book to look up K+R vs K mate), you _will_ overestimate the real strength of the engine. It's also obvious why, because the engine gets a lot of outside help: adjudication on material advantage, and book knowledge about endgames through the table bases. Reaching a winning position is enough to either win the game by referee, or by looking up how to actually win it.
At the moment, Princhess is like that generally strong club player that gets a winning position against about anyone else in the club, but then doesn't have the knowledge on how to finish it off. And after that becomes common knowledge, no-one will ever resign against this player. He or she will therefore probably not become the club champion.
So yes, Princhess _would_ be around 1815 - 1865, if the engine could indeed win the winning positions.
1'+1" rating ~1630.flok wrote: ↑Sat Oct 16, 2021 9:32 am For others: https://vanheusden.com/CaffeinatedPawn/ or https://github.com/folkertvanheusden/CaffeinatedPawn/
For me, I have no problem with Princhess requiring Syzygy at this point in time. There is a case to be made that an engine that doesn't require an egtb is stronger than one that does, but there's lots of things that could be improved about Princhess and how she plays without an egtb is not the most interesting thing for me right now. So I probably won't look at that for some time (altho hopefully some of the other improvements have a side effect of improving the no egtb play, even if it's not their goal).using Syzygy for mating (which would be a weak player consulting a book to look up K+R vs K mate)
Code: Select all
all_versions_1 | Rank Name Elo +/- Games Score Draw
all_versions_1 | 1 Rustic Alpha 3.0.0 211 22 1000 77.1% 20.0%
all_versions_1 | 2 Princhess 0.4.0 -34 18 1000 45.1% 27.9%
all_versions_1 | 3 Princhess 0.3.1 -166 19 1000 27.8% 29.1%
Rustic doesn't have pruning (more or less), so the tree it searches is much wider than what Princhess searches, maybe that is it?Lana Samson wrote: ↑Sat Oct 16, 2021 5:08 pmFor me, I have no problem with Princhess requiring Syzygy at this point in time. There is a case to be made that an engine that doesn't require an egtb is stronger than one that does, but there's lots of things that could be improved about Princhess and how she plays without an egtb is not the most interesting thing for me right now. So I probably won't look at that for some time (altho hopefully some of the other improvements have a side effect of improving the no egtb play, even if it's not their goal).using Syzygy for mating (which would be a weak player consulting a book to look up K+R vs K mate)
I did run the following overnight (10+0.1, Princhess using Syzygy):So Princhess 0.4.0 doesn't seem to be a regression, but Rustic definitely has something over her. Perhaps there's something in its play that just hits a Princhess weakness consistently.Code: Select all
all_versions_1 | Rank Name Elo +/- Games Score Draw all_versions_1 | 1 Rustic Alpha 3.0.0 211 22 1000 77.1% 20.0% all_versions_1 | 2 Princhess 0.4.0 -34 18 1000 45.1% 27.9% all_versions_1 | 3 Princhess 0.3.1 -166 19 1000 27.8% 29.1%
That's more than I expected!Brunetti wrote: ↑Sat Oct 16, 2021 2:47 pm1'+1" rating ~1630.flok wrote: ↑Sat Oct 16, 2021 9:32 am For others: https://vanheusden.com/CaffeinatedPawn/ or https://github.com/folkertvanheusden/CaffeinatedPawn/
Alex
Code: Select all
4 tscp-1.8.1c : 1600.8 1807.5 2296 79
5 CP-a8c074d : 1307.8 308.0 1948 16Maybe you used a time control where your engine is less performing, or that limited field of opponents is particularly bad for Caffeinated. My testings are based on a sort of Swiss system, and the resulting ratings look quite accurate (if you ask me, of courseflok wrote: ↑Sat Oct 16, 2021 8:09 pm That's more than I expected!
Code: Select all
4 tscp-1.8.1c : 1600.8 1807.5 2296 79 5 CP-a8c074d : 1307.8 308.0 1948 16
).Code: Select all
Rank Elo ± Engine Score Games Wins Draws Loss Oppo time stal ille
-------------------------------------------------------------------------------------------------------------
1. 1627 33 CaffeinatedPawn 46203a4 64-bit 44% 451 193 15 243 -51 0% 22% 0%
-------------------------------------------------------------------------------------------------------------
Results of CaffeinatedPawn 46203a4 64-bit
-------------------------------------------------------------------------------------------------------------
Rank Elo Engine Score Games Wins Draws Losses
-------------------------------------------------------------------------------------------------------------
1. 2211 Madeleine 0.2 0% 1 0 0 1
2. 2208 Blunder 6.1.0 64-bit 0% 1 0 0 1
3. 2195 Kurt 0.9.2.2 beta 64-bit 0% 1 0 0 1
4. 2192 DrunkenMaster 1.2 0% 1 0 0 1
5. 2181 Tiny Chess 1.4.6 64-bit 0% 1 0 0 1
6. 2067 Bad Chess Engine 0.4.4-ucifix 64-bit 0% 1 0 0 1
7. 2062 ProChess 102 A doctor 0% 1 0 0 1
8. 2043 Sapeli 2.0 64-bit 0% 1 0 0 1
9. 2042 DelphiMax 2.9 0% 1 0 0 1
10. 2032 Etude 0.11 0% 1 0 0 1
11. 1961 Gunborg 1.0 64-bit 0% 1 0 0 1
12. 1936 Lila 3.0.1 64-bit 33% 3 1 0 2
13. 1930 Odonata 0.3.26 64-bit 0% 4 0 0 4
14. 1923 FreeChess 2.17b 25% 4 1 0 3
15. 1921 Roce 0.039 13% 4 0 1 3
16. 1911 Clueless 1.4 0% 5 0 0 5
17. 1905 Rustic Alpha 3.0.0 64-bit 20% 5 1 0 4
18. 1904 Deep Chess r113 64-bit 10% 5 0 1 4
19. 1901 Minace 1.0 20% 5 1 0 4
20. 1898 Snowy 0.1 64-bit 20% 5 1 0 4
21. 1894 GearHeart 0.16 TP 64-bit 0% 5 0 0 5
22. 1892 Deepov 0.4 64-bit 10% 5 0 1 4
23. 1890 Predateur2.2.1 20% 5 1 0 4
24. 1886 CDrill 1800 Build 4 20% 5 1 0 4
25. 1885 HeavyChess 0.13 beta 0% 5 0 0 5
26. 1878 Chareth 2020-06-28 64-bit 0% 5 0 0 5
27. 1877 Heracles 0.5.9 10% 5 0 1 4
28. 1874 Sayuri 2018-05-23 64-bit 20% 5 1 0 4
29. 1869 Ziggy 0.7 64-bit 0% 5 0 0 5
30. 1869 Hopper 20211004 64-bit 20% 5 1 0 4
31. 1866 Pigeon 1.5.1 64-bit 20% 5 1 0 4
32. 1857 Apollo 1.2.1 64-bit 20% 5 1 0 4
33. 1855 Zurichess Appenzeller 64-bit 0% 5 0 0 5
34. 1835 Silke Chess 1.2.1209 17% 6 1 0 5
35. 1827 BeRoChess 1.00.2012.04.10.0112 64-bit 67% 6 4 0 2
36. 1809 Hopper 20211003 64-bit 33% 6 2 0 4
37. 1801 Hedwig rev25 64-bit 29% 7 2 0 5
38. 1774 Rocinante 2.0 64-bit 36% 7 2 1 4
39. 1770 Testina 3c 71% 7 5 0 2
40. 1767 Celestial 1.0 64-bit 7% 7 0 1 6
41. 1760 RapChessCs 2021-07-22 64-bit 14% 7 1 0 6
42. 1759 Shallow Blue 2.0.0 64-bit 0% 7 0 0 7
43. 1738 Pwned 1.3 64-bit 56% 8 4 1 3
44. 1733 Small-C 90.3 13% 8 1 0 7
45. 1712 Crabby 1.00 64-bit 38% 8 3 0 5
46. 1705 FracTal 1.0 64-bit 28% 9 2 1 6
47. 1692 CaSpEr rev4 64-bit 33% 9 2 2 5
48. 1680 Walleye 1.4 64-bit 22% 9 2 0 7
49. 1669 Chancellor 1.00d 33% 9 3 0 6
50. 1657 ArcBishop 1.00d 67% 9 6 0 3
51. 1631 Pigeon 1.37 64-bit 20% 10 2 0 8
52. 1623 Cicada 0.1 64-bit 55% 10 4 3 3
53. 1617 Pulse 1.7.2 64-bit 60% 10 6 0 4
54. 1609 Conqueror 1.2 70% 10 7 0 3
55. 1599 Koedem 1.1 64-bit 44% 9 4 0 5
56. 1586 SebLague Chess-AI 0.1 .Net5 56% 9 5 0 4
57. 1559 ChessEngine (SeminarEngine) 0.1-alpha 64-bit 63% 8 5 0 3
58. 1553 DoctorB 1.2.1 64-bit 50% 8 4 0 4
59. 1549 Piranha 0.5 75% 8 6 0 2
60. 1531 Jacky Chess 0.9.14 64-bit 75% 8 6 0 2
61. 1529 Chessier 0.1.0 64-bit 88% 8 7 0 1
62. 1526 Irina 0.15 88% 8 7 0 1
63. 1508 Superpawn build 110 64-bit 63% 8 5 0 3
64. 1498 Walleye 1.3.0 64-bit 71% 7 5 0 2
65. 1470 Quokka 2.1 64-bit 57% 7 4 0 3
66. 1459 Lynx 5.2 64-bit 100% 6 6 0 0
67. 1459 Madlenka 2015.02.13 DC 64-bit 75% 6 4 1 1
68. 1453 Lynx 5.1 64-bit 67% 6 4 0 2
69. 1452 Hactar 0.9.0 64-bit 100% 6 6 0 0
70. 1451 Lynx 5.3 64-bit 86% 7 6 0 1
71. 1445 Eden 0.0.13 83% 6 5 0 1
72. 1427 Neocortex 1.0 64-bit 83% 6 5 0 1
73. 1426 Endamat 20210217 64-bit 83% 6 5 0 1
74. 1390 Overhead 64-bit 60% 5 3 0 2
75. 1384 Sargon 1978 1.01 100% 5 5 0 0
76. 1372 KillerQueen 2 beta 3 63% 4 2 1 1
77. 1362 AXWChessbot 0.1.0 64-bit 100% 5 5 0 0
78. 1354 Haddock 0.2.1 64-bit 100% 5 5 0 0
79. 1331 Rumney 0.2.1 64-bit 75% 4 3 0 1
80. 1320 Belofte 2.1.3 64-bit 100% 1 1 0 0