WCCC 2022

[lkaufman]

Larry - will the Dragon version which played in WCC be the next release and when is the next version due for release.

Yes, the plan is to release the WCC version (with minor mods, mostly related to LimitStrength and Time Management) soon, as soon as we can manage with conflicting vacation schedules etc.

Excellent Larry, looking forward to it. Will the new version be Dragon 3.x or Dragon 4.

Dragon 3.1. Because it fixes several bugs or issues with Dragon 3 (as well as improving strength), it will be a free upgrade to those who bought Dragon 3.

[lkaufman]

Congratulations to Ginkgo. Regardless of the result, I must say that using Armageddon to break ties between the top engines is just ridiculous, the time odds mean very little whereas draw odds is overwhelming. Whoever wins the toss for Black is almost sure to win, might as well skip the game. Much better to keep shortening the time limit in pairs of games, even if you end up with a string of bullet games to decide. At least it would be fair.

Larry, I agree with that. At such a high and similar level of chess engines play, a draw in Armageddon seems much more likely than Armageddon in a game played by humans.
Certainly, this Championship in the next edition should change its playing formula.

Totally agree that Armageddon is a bad idea (even for humans...).

I presume the tiebreak was known before the tourney and one could say it's 'not good form' to criticize the method of tiebreak 'after the result (didn't go your way). Perhaps though you let it be known before the tourney started. But...again, I agree.

At least for human GMs, Armageddon as usually used is roughly "fair" though it is apt to favor players with particular styles or opening preferences. But with top engines, it should be obvious to anyone who follows results of engine tournaments or rating lists that the double time Armageddon used probably gives Black a 90%+ probability of winning if both engines are equal and top-notch. Anyway, I want to be on record as saying that it will be a bad idea for the overall WCC this week even if we end up winning it this way. Basically, it's almost like a coin-flip, whoever gets Black wins.

Yes, horrible for engines.

For humans, I actually like (but have never had the opportunity to play) the idea where each side 'bids' for how much time advantage you will give your opponent and get to pick the color you play if yours is the lower bid. That over...some coin flip (or whatever) where one side if forced to play black with draw odd or white and he must win. I wonder what kind of time odds a programmer would give to another programmer under such a tie-break scenario.

I actually had this experience, in 2009 or 2010. I tied with another (lower-rated) player for the U.S. over-60 championship, with the winner getting the slot (with most expenses paid) to the World Senior. So we had to decide the winner by an Armageddon game, with White getting 30 minutes and the players bidding for Black with draw odds. I think I bid somewhere around 20 minutes, while my opponent, who was clearly clueless, make the ridiculous bid of 5 minutes (!!) and had to give me 30 to 5 time odds playing Black, with only draw odds to compensate! Needless to say, I won quite easily and went on to the World Senior. This does show the downside of bidding, namely that it penalizes the player with less (or no) experience with Armageddon play. It is "fair", but it rewards a skill that is far removed from chess strength (judging time odds). As for what is fair with top engines playing each other, the longer the tc, the greater the ratio must be. Probably if Black gets some Rapid time (say 10' + 5") on good hardware, even 1 year per move won't be enough for White to be favored to win. Only if Black is playing bullet chess (maybe 1' + 1") is Armageddon even feasible with top engines on good hardware, and even then the time odds ratio needed would be huge.

[Harvey Williamson]

Congratulations to Ginkgo. Regardless of the result, I must say that using Armageddon to break ties between the top engines is just ridiculous, the time odds mean very little whereas draw odds is overwhelming. Whoever wins the toss for Black is almost sure to win, might as well skip the game. Much better to keep shortening the time limit in pairs of games, even if you end up with a string of bullet games to decide. At least it would be fair.

Larry, I agree with that. At such a high and similar level of chess engines play, a draw in Armageddon seems much more likely than Armageddon in a game played by humans.
Certainly, this Championship in the next edition should change its playing formula.

I agree Armageddon is not a good way to break ties. Using shorter and shorter time controls to break a tie could become a test of who is the fastest operator, so not ideal either.

[Eduard]

I just want to say that I don't like Armageddon either. For computer chess this is the wrong way.

[Eduard]

Live:
https://icga.org/?page_id=3542

Round 1 (Tuesday 26 July, 08:30)

Ginkgo – Leela Zero Chess ½–½
Shredder – Komodo Dragon ½–½
The Baron bye

Round 2 (Tuesday 26 July, 11:00)

Leela Zero Chess – Shredder 1-0
Ginkgo – The Baron 1-0
Komodo Dragon bye

WCCC Round 3:

Komodo Dragon vs Lc0

The resulting position after move 12 is played hundreds of times a day on PlayChess, thousands of times in a week. White used to be able to win occasionally, and only when there was time trouble on stronger hardware. In the end I only saw a draw. The strongest line for Black is Bg4 (as in the game) and then exchange pieces.

[Eduard]

Round 2, Lc0 vs Shredder (My short comment):

[Event "World Computer Chess Championship"]
[Site "Vienna, Austria"]
[Date "2022.07.26"]
[Round "2.1"]
[White "Leela Chess Zero"]
[Black "Shredder"]
[Result "1-0"]
[ECO "D45"]
[PlyCount "129"]
[EventDate "2022.07.26"]

1. d4 {4} d5 {3} 2. c4 {91} c6 {3} 3. Nc3 {81} Nf6 {3} 4. Nf3 {452} e6 {1} 5.
e3 {4} Nbd7 {3} 6. Qc2 {5} Bd6 {5} 7. Bd3 {12} O-O {5} 8. O-O {78} dxc4 {3} 9.
Bxc4 {66} b5 10. Bd3 {100} Bb7 11. a3 {180} a5 {This variant leads to sharp
play and an open position.} 12. e4 {66} e5 {124} 13. Nxe5 {106} Nxe5 {100} 14.
dxe5 {4} Bxe5 {235} 15. h3 {15 In my PlayChess Database I have more then 3000
games only since 2019. Almost all games ended in draws. 192 times white won
and black only 16 times.} a4 $2 {110 No games in my database. Here b4 and Ba6
are mostly played.} 16. Be3 {190 Stockfish is here +1! I heard that lc0 plays
on 8 A100 graphics units. That should win here too.} Re8 {40} 17. f4 {4} Bxc3 {
132} 18. bxc3 {7} c5 {118} 19. e5 c4 {180} 20. Be2 {5} Be4 {95} 21. Qb2 {6} Nd5
{98} 22. Bd4 {6} Qh4 {88} 23. f5 {6} Rad8 {81} 24. Bf3 {254} Bd3 {35} 25. Rf2 {
15} Qg3 {120} 26. Bxd5 {10} Rxd5 {96} 27. f6 gxf6 {64} 28. Rxf6 {11} Re6 {129}
29. Rf3 Qh4 {121} 30. Kh2 {5} Be4 {76} 31. Qf2 Qxf2 {62} 32. Rxf2 {10} Rd8 {68}
33. g4 {7} Rg6 {92} 34. Rd1 {7} Bc6 {151} 35. Rf5 {29} Rd7 {79} 36. Kg3 {40}
Bd5 {32} 37. Be3 {24} Be6 {43} 38. Rxd7 {309} Bxd7 {2} 39. Rf1 {196} Bc6 {121}
40. Re1 {194} h5 {98} 41. g5 {132} Re6 {5} 42. Bc5 {94} Kh7 {68} 43. Kf4 {88}
Kg6 {96} 44. h4 {25} Re8 {74} 45. Bd6 {79} Bd7 {55} 46. Ke4 {105} Bf5+ {103}
47. Kd4 {66} Rc8 {207} 48. Re2 {36} Be6 {12} 49. Rg2 {155} Rc6 {213} 50. Rf2 {
138} Bf5 {8} 51. Rb2 {212} Bd7 {124} 52. Rf2 {8} Bf5 {48} 53. Rb2 {34} Bd7 {28}
54. Kd5 {123} Rc8 {33} 55. Rf2 {7} Bf5 {70} 56. Rf1 {4} Be6+ {64} 57. Ke4 {6}
Re8 {60} 58. Kd4 {34} Bd7 {39} 59. Kc5 {19} Re6 {104} 60. Kb6 {5} Re8 {46} 61.
Rf6+ {25} Kg7 {36} 62. Kc7 {5} Be6 {50} 63. Kc6 {29} Kg8 {14} 64. Kxb5 {45} Bd5
65. Kxa4 {26} 1-0

[Eduard]

Round 3 (Tuesday 26 July, 14:00)

Shredder – The Baron 1-0
Komodo Dragon – Leela Zero Chess 1/2
Ginkgo bye

[CornfedForever]

[

I actually had this experience, in 2009 or 2010. I tied with another (lower-rated) player for the U.S. over-60 championship, with the winner getting the slot (with most expenses paid) to the World Senior. So we had to decide the winner by an Armageddon game, with White getting 30 minutes and the players bidding for Black with draw odds. I think I bid somewhere around 20 minutes, while my opponent, who was clearly clueless, make the ridiculous bid of 5 minutes (!!) and had to give me 30 to 5 time odds playing Black, with only draw odds to compensate! Needless to say, I won quite easily and went on to the World Senior. This does show the downside of bidding, namely that it penalizes the player with less (or no) experience with Armageddon play. It is "fair", but it rewards a skill that is far removed from chess strength (judging time odds). As for what is fair with top engines playing each other, the longer the tc, the greater the ratio must be. Probably if Black gets some Rapid time (say 10' + 5") on good hardware, even 1 year per move won't be enough for White to be favored to win. Only if Black is playing bullet chess (maybe 1' + 1") is Armageddon even feasible with top engines on good hardware, and even then the time odds ratio needed would be huge.

True, there is not allowance for 'stupid'...but (in theory) it keeps things more in the hands of the players and what they think they are capable of (ir what they think will be most difficult for their opponent) instead of the way used here.

Would Tal be better on the White/win vs Black/draw of the way used here? I think he would prefer White. To force him to play black (or the opponent to have to defend!) arbitrarily just seems...wrong.

[Werewolf]

Any ideas as to Shredder’s hardware?
Arm 64…

[Eduard]

Posted on Discord (Lc0 events-discuss):



Round 4 (Wednesday 27 July, 08:30)

The Baron – Komodo Dragon 0-1
Ginkgo – Shredder ½–½
Leela Zero Chess bye

Round 5 (Wednesday 27 July, 11:00)

Komodo Dragon – Ginkgo 1/2
Leela Zero Chess – The Baron 1-0
Shredder bye

Round 6 (Wednesday 27 July, 14:30)

Leela Zero Chess – Ginkgo
Komodo Dragon – Shredder
The Baron bye