Chess is a Draw

[lkaufman]

I once read a study which found that one of the best ways to reduce the draw rate was to allow taking of one's own pieces!

Apparently, the value of being able to do this outweighs the obvious drawback (one loses material).

However, it might be that as the tactic became widely known, it would lose its potency.

There are many such drastic rule changes that would reduce draws rates a lot. One variant (I forget the name) is simply to let the bishops also move like knights (so they are archbishops or januses), I imagine this would drastically reduce draws unless perpetual check becomes too common (in which case it could simply be outlawed). But I'm looking for changes that don't change the nature of games other than games that are currently draws. The draw rules in chess don't seem as sacrosanct as the basic moves of the pieces. If we want to reduce or eliminate draws, why not focus on the draw rules?

Thank you for responding, and that's a very good point. My problem is that engines are becoming unbeatable in the game of chess, and changing the draw rules will only delay the inevitable.

On the other hand, if the solution is to make chess bigger (larger board) or more complicated (new rules which expand the size of the game tree), then it just won't catch on: I've never seen anyone play tic-tac-toe on a grid larger than 3x3, for example (unless "Connect 4" counts as a tic-tac-toe variant).

So maybe delaying the inevitable is the best we can do right now.

Changing the draw rules will not "delay the inevitable". With the rules I proposes, there would be no draws, every game would be decisive. The risk is that one side may have a clear advantage, which at top engine level might translate to one color always winning. But if we get the rules just right, the practical chances should be so close to equal that this won't be an issue until the entire game tree is mapped out, which presumably won't happen before the end of life on earth, though you never know. Basically, we have this situation in shogi now, where it isn't yet clear whether the first player wins or the result is a draw with perfect play, though recently evidence has shifted towards the "first player win" side of the argument. With the right rules, chess could be right on that narrow edge where no one will know the correct result even in a thousand years (probably).

[Chessqueen]

The People from TCEC Knowing about this situation returned back to their entertainment by giving LCZero a bad Opening with Black, here from the 2nd Move on... TCEC chose a bad Opening reply for TCZero 2....c5?? {book} Which is horribe, the best engine reply is 2....e6

[pgn][Event "TCEC Season 26 - Superfinal"]
[Site "https://tcec-chess.com"]
[Date "2024.05.15"]
[Round "2.1"]
[White "Stockfish dev-20240513-e608eab8"]
[Black "LCZero 0.31-dag-5350a2e-BT4-6147500"]
[Result "1-0"]
[BlackElo "3634"]
[ECO "A56"]
[GameDuration "04:40:46"]
[GameEndTime "2024-05-15T02:00:58.400 UTC"]
[GameStartTime "2024-05-14T21:20:11.714 UTC"]
[Opening "Czech Benoni defence"]
[PlyCount "241"]
[Termination "adjudication"]
[TerminationDetails "SyzygyTB"]
[TimeControl "7200+12"]
[WhiteElo "3662"]
[Annotator "archive"]

{WhiteEngineOptions: Hash=98304; Move Overhead=1000; OwnBook=false; Ponder=false; Protocol=uci; SyzygyPath=/home/syzygy7; Threads=101; UCI_ShowWDL=true;, BlackEngineOptions: Backend=demux; BackendOptions=backendcuda-fp16,(gpu0,policy_headoptimistic,value_headwinner),(gpu1,policy_headoptimistic,value_headwinner); CPuct=2.897; CPuctBase=45669; CPuctFactor=3.973; CommandLineOptions=--show-hidden; Contempt=Stockfish0; ContemptMaxValue=600; FpuValue=0.98416; LogFile=lc0.log; MinibatchSize=160; MoveOverheadMs=1000; MoveRuleBucketing=true; MovesLeftSlope=0.0027; MovesLeftThreshold=0.8; PolicyTemperature=1.4; Ponder=false; Protocol=uci; RamLimitMb=94500; ScoreType=WDL_mu; SmartPruningFactor=2; SmartPruningMinimumBatches=300; StrictTiming=true; SyzygyPath=/home/syzygy7; TaskWorkers=3; Threads=2; TimeManager=legacy(book-ply-bonus0.4,first-move-bonus3.5); UCI_ShowMovesLeft=true; UCI_ShowWDL=true; UseUncertaintyWeighting=true; VerboseMoveStats=true; WDLCalibrationElo=3600; WDLContemptAttenuation=0.55; WDLDrawRateReference=0.64; WDLEvalObjectivity=0; WeightsFile=BT4-1024x15x32h-swa-6147500.pb.gz;}
1. d4 {book}
Nf6 {book}
2. c4 {book}
c5 {book}
3. d5 {book}
e5 {book}
4. Nc3 {book}
d6 {book}
5. e4 {book}
Be7 {book}
6. Bd3 {book}
O-O {book}
7. Nge2 {book}
Nh5 {book}
8. Be3 {book}
Bg5 {book}[/pgn]

[Chessqueen]

The People from TCEC Knowing about this situation returned back to their entertainment by giving LCZero a bad Opening with Black, here from the 2nd Move on... TCEC chose a bad Opening reply for LCZero 2....c5?? {book} Which is horribe, the best engine reply is 2....e6

[pgn][Event "TCEC Season 26 - Superfinal"]
[Site "https://tcec-chess.com"]
[Date "2024.05.15"]
[Round "2.1"]
[White "Stockfish dev-20240513-e608eab8"]
[Black "LCZero 0.31-dag-5350a2e-BT4-6147500"]
[Result "1-0"]
[BlackElo "3634"]
[ECO "A56"]
[Opening "Benoni"]
[Variation "Czech, 5.e4 Be7 6.Bd3"]
[WhiteElo "3662"]
[TimeControl "7200+12"]
[Termination "adjudication"]
[PlyCount "241"]
[WhiteType "human"]
[BlackType "human"]

1. d4 Nf6 2. c4 c5 3. d5 e5 4. Nc3 d6 5. e4 Be7 6. Bd3 O-O 7. Nge2 Nh5 8.
Be3 Bg5 9. Qc1 Bxe3 10. Qxe3 Nd7 11. g3 g6 12. Qh6 Kh8 13. h3 a6 14. a4
Nhf6 15. Qe3 b6 16. Kd2 Ne8 17. g4 Qh4 18. Rag1 Ra7 19. f3 Rg8 20. Kc2 h6
21. a5 bxa5 22. Ra1 Kg7 23. Nc1 a4 24. Rxa4 h5 25. Ra1 hxg4 26. fxg4 Nf8
27. Nb3 Rh8 28. Nd2 Rb7 29. Be2 Qd8 30. Nb3 Qh4 31. Nd2 Qd8 32. b3 Kg8 33.
Nd1 Rc7 34. Nf2 g5 35. Nf3 f6 36. Qc3 Rb7 37. Rxa6 Rxb3 38. Qa5 Qxa5 39.
Rxa5 Rb7 40. Rha1 Rhh7 41. Nd2 Nd7 42. Nb1 Kf8 43. Nc3 Rb8 44. Bf1 Rb4 45.
Ra7 Rg7 46. Nd3 Rb8 47. Ne1 Nb6 48. R7a2 Rh7 49. Kd2 Bd7 50. Nc2 Nc8 51.
Ke3 Rb3 52. Ra3 Rb2 53. R1a2 Rb4 54. Kf3 Nc7 55. Kg3 Ke7 56. Ra1 Be8 57.
Bd3 Kd8 58. Rf1 Rf7 59. Ne3 Rf8 60. Rb1 Bg6 61. Rh1 Rf7 62. Rf1 Ne8 63. Bc2
Rb2 64. Ra8 Nc7 65. Raa1 Rb4 66. Nb1 Ne8 67. Nd2 Nb6 68. h4 gxh4+ 69. Kxh4
Rf8 70. Kg3 Rh8 71. Ra2 Rh7 72. Rfa1 Nc7 73. Rd1 Ne8 74. Rf1 Rh8 75. Bd3
Rh7 76. Rfa1 Nc8 77. Bc2 Rhb7 78. Ra3 Nb6 79. Rh1 Rh7 80. Rxh7 Bxh7 81. Kh4
Bg6 82. g5 Bf7 83. Bb3 Nc8 84. Bd1 Nb6 85. Ra7 fxg5+ 86. Kxg5 Nc7 87. Nf5
Nc8 88. Ra3 Rb2 89. Nf1 Rb1 90. N1e3 Be8 91. Ng7 Bd7 92. Bh5 Rg1+ 93. Kf6
Rg3 94. Rb3 Rh3 95. Bg6 Rf3+ 96. Bf5 Ne8+ 97. Nxe8 Kxe8 98. Rb7 Ne7 99. Ra7
Rf4 100. Kg5 Bc8 101. Ra8 Rf3 102. Bxc8 Nxc8 103. Rxc8+ Kd7 104. Rh8 Rxe3
105. Rh7+ Ke8 106. Kf6 Rg3 107. Re7+ Kd8 108. Rf7 Rg4 109. Ke6 Rg6+ 110.
Rf6 Rg8 111. Kxd6 Kc8 112. Kc6 Kb8 113. d6 Ka7 114. d7 Rb8 115. Rd6 Rb6+
116. Kd5 Rb8 117. d8=Q Rxd8 118. Rxd8 Kb7 119. Kd6 Kb6 120. Rd7 Ka5 121.
Kxc5 1-0[/pgn]

[towforce]

If chess is now so clearly drawn with proper play, or even with "good" play, the question arises as to what changes to the draw rules, without changing other rules of the game, would eliminate most or all draws while leaving some doubt as to which side should win. The obvious changes, making stalemate a loss for the stalemated side and making bare king a loss (restoring the rules of around year 1400), would of course reduce draws but not sufficiently to cast any doubt on the drawn result with proper play. So what to do about repetitions and fifty move rule draws? If we assume that it's easier to avoid repetitions than fifty move rule draws, what would happen if all repetitions were called wins for White, while all fifty move rule draws were called wins for Black (plus the above stalemate/bare king rules)? It's not at all obvious to me in this case whether White or Black should win with perfect play. Perhaps someone could modify Stockfish or some other strong engine to use these draw rules and test it, although of course the play wouldn't be optimized for those rules.

OK - let's take a look at these, remembering that we're talking about strong players (the weaker the players, the less likely they are to draw):

* stalemate a loss for the stalemated side

A good idea


* making bare king a loss

I could live with this, but I think that in many situations, it just means that neither side will let the other exchange off their last piece, so still a lot of draws (but see the new 50 move rule!)


* all repetitions were called wins for White

Seems a little arbitrary: how about... repetitions are a loss for the player whose move created a 3rd repetition?


* all fifty move rule draws were called wins for Black

Again, arbitrary: how about... the first player to play their 50th move loses? This would be the first player to move after a capture or pawn move. Hence if there's an exchange in the endgame, then you'd have to be the player taking back, and not the player initiating the exchange. Under Larry's version of the rule ("all fifty move rule draws were called wins for Black"), white would have to try to avoid a drawing endgame, which is a very unfair burden when the premise is that "chess is a draw".

[jefk]

how about... repetitions are a loss for the player whose move created a 3rd repetition?

sounds quite reasonable (similar with the 50 move), but like mr LK
wrote, it remains to be seen how large the White advantage will be
if there are no more draws (and the rules are 'neutral'). Maybe in some
way with some assymetric (b/w) rules this can be compensated.

As for the stalemate (and bare king), stronger human players (and engine users)
may not like it because it might considerably change some endgame play.
To check this, testing with -modified- engines would quickly give
an impression how such games would look like. In addition to the rule changes
in such a 'Kaufman variant' (with bare king = loss), I can think of other
endgame (rule) changes on the basis of the remaining material e.g. pawn
+ piece vs pawn being a win for the side with more material, depending on
specific situations to be specified in detail (also keeping in mind some suggestions
Arno Nickel made in the past).
https://en.chessbase.com/post/correspon ... aw-problem
Such rule modifications also would have a large impact on endgame play ofcourse
(which some strong human otb players may not like); again such changes could be 'tuned'
to ensure equal winning chances for both sides; but then keeping it simple
initially also obviously will be (most) important.

The interesting part of such an anti-draw variant is that we still
have the usual initial board position (and thus opening theory). This
can make it straightaway already more attractive than Chess960.

PS as for testing (with modified engine or optional changes)
the search may have to go back to handcrafted evals, at least after
the opening stage but for testing i don't see that as huge objection.

[lkaufman]

If chess is now so clearly drawn with proper play, or even with "good" play, the question arises as to what changes to the draw rules, without changing other rules of the game, would eliminate most or all draws while leaving some doubt as to which side should win. The obvious changes, making stalemate a loss for the stalemated side and making bare king a loss (restoring the rules of around year 1400), would of course reduce draws but not sufficiently to cast any doubt on the drawn result with proper play. So what to do about repetitions and fifty move rule draws? If we assume that it's easier to avoid repetitions than fifty move rule draws, what would happen if all repetitions were called wins for White, while all fifty move rule draws were called wins for Black (plus the above stalemate/bare king rules)? It's not at all obvious to me in this case whether White or Black should win with perfect play. Perhaps someone could modify Stockfish or some other strong engine to use these draw rules and test it, although of course the play wouldn't be optimized for those rules.

OK - let's take a look at these, remembering that we're talking about strong players (the weaker the players, the less likely they are to draw):

* stalemate a loss for the stalemated side

A good idea


* making bare king a loss

I could live with this, but I think that in many situations, it just means that neither side will let the other exchange off their last piece, so still a lot of draws (but see the new 50 move rule!)


* all repetitions were called wins for White

Seems a little arbitrary: how about... repetitions are a loss for the player whose move created a 3rd repetition?


* all fifty move rule draws were called wins for Black

Again, arbitrary: how about... the first player to play their 50th move loses? This would be the first player to move after a capture or pawn move. Hence if there's an exchange in the endgame, then you'd have to be the player taking back, and not the player initiating the exchange. Under Larry's version of the rule ("all fifty move rule draws were called wins for Black"), white would have to try to avoid a drawing endgame, which is a very unfair burden when the premise is that "chess is a draw".

Of course such symmetrical (color blind) rules are more desirable in principle, and I have proposed those rules in the past, but the problem is that it is obvious that this variant is a win for White. Every strong player or student of opening theory knows that White can normally force a "cosmetic" advantage, such as an extra pawn in an obviously drawn bishops of opposite color ending or a rook and 3 vs rook and two or some similarly unequal but dead drawn ending. Any proposal that eliminates all (or almost all) draws in a colorblind way will leave the game won for White, and most likely at top engine level easily won for White. My proposal at least leaves some doubt as to which side should win; perhaps White can force a win if Black has to avoid all repetitions, who knows? I agree that it would be better to minimize the cases where the rules consider color. Perhaps if repetitions are losses for the repeater, and some subset of 50 move rule positions are awarded by some color blind rule (examples: first player to check during the fifty moves wins, or any material advantage (based on some point count) wins, or some point scoring of pawns based on advancement), with Black winning when no rule awards the win, would be so close to fair that no one would be able to say who should win even with heavy engine analysis. But someone would have to program all these ideas into Stockfish (or another top engine) to allow for testing. Maybe starting with Stockfish 11 (pre-NNUE) would be best for this purpose.

[Chessqueen]

If chess is now so clearly drawn with proper play, or even with "good" play, the question arises as to what changes to the draw rules, without changing other rules of the game, would eliminate most or all draws while leaving some doubt as to which side should win. The obvious changes, making stalemate a loss for the stalemated side and making bare king a loss (restoring the rules of around year 1400), would of course reduce draws but not sufficiently to cast any doubt on the drawn result with proper play. So what to do about repetitions and fifty move rule draws? If we assume that it's easier to avoid repetitions than fifty move rule draws, what would happen if all repetitions were called wins for White, while all fifty move rule draws were called wins for Black (plus the above stalemate/bare king rules)? It's not at all obvious to me in this case whether White or Black should win with perfect play. Perhaps someone could modify Stockfish or some other strong engine to use these draw rules and test it, although of course the play wouldn't be optimized for those rules.

OK - let's take a look at these, remembering that we're talking about strong players (the weaker the players, the less likely they are to draw):

* stalemate a loss for the stalemated side

A good idea


* making bare king a loss

I could live with this, but I think that in many situations, it just means that neither side will let the other exchange off their last piece, so still a lot of draws (but see the new 50 move rule!)

They should make chess Like Checker, that if you can’t move in checkers you lose the game! There’s no passing and it’s not a draw. There are only two ways to win a game of checkers: Capture all your opponent’s pieces Render your opponent unable to move.And a great idea in Chess would be that if a player is losing in material but forces his opponent to a three move repetions, he or she should lose the game, and if you lose on time but you are winning in material it should be a draw, since you worked so hard to obtain that material advantage

NOTE: It is ridiculous to lose a chess game in time when you are beating your opponent with a Queen vs your opponent having either only a Rook or a Bishop +Knight, in such case the game should be delacred a Draw.

[lkaufman]

If chess is now so clearly drawn with proper play, or even with "good" play, the question arises as to what changes to the draw rules, without changing other rules of the game, would eliminate most or all draws while leaving some doubt as to which side should win. The obvious changes, making stalemate a loss for the stalemated side and making bare king a loss (restoring the rules of around year 1400), would of course reduce draws but not sufficiently to cast any doubt on the drawn result with proper play. So what to do about repetitions and fifty move rule draws? If we assume that it's easier to avoid repetitions than fifty move rule draws, what would happen if all repetitions were called wins for White, while all fifty move rule draws were called wins for Black (plus the above stalemate/bare king rules)? It's not at all obvious to me in this case whether White or Black should win with perfect play. Perhaps someone could modify Stockfish or some other strong engine to use these draw rules and test it, although of course the play wouldn't be optimized for those rules.

OK - let's take a look at these, remembering that we're talking about strong players (the weaker the players, the less likely they are to draw):

* stalemate a loss for the stalemated side

A good idea


* making bare king a loss

I could live with this, but I think that in many situations, it just means that neither side will let the other exchange off their last piece, so still a lot of draws (but see the new 50 move rule!)

They should make chess Like Checker, that if you can’t move in checkers you lose the game! There’s no passing and it’s not a draw. There are only two ways to win a game of checkers: Capture all your opponent’s pieces Render your opponent unable to move.And a great idea in Chess would be that if a player is losing in material but forces his opponent to a three move repetions, he or she should lose the game, and if you lose on time but you are winning in material it should be a draw, since you worked so hard to obtain that material advantage

NOTE: It is ridiculous to lose a chess game in time when you are beating your opponent with a Queen vs your opponent having either only a Rook or a Bishop +Knight, in such case the game should be delacred a Draw.

Making stalemate a loss for the stalemated side, although a good idea, won't reduce draws by enough to be worth the fight to make just that one change; forbidding all repetitions (together with the stalemate change) would reduce them drastically and would indeed be worth the fight. Basing it on material count seems unnecessary. As for losing on time with material advantage, no need for a rule change, just refuse to play without increment!

[jefk]

LK

someone would have to program all these ideas into Stockfish (or another top engine) to allow for testing. Maybe starting with Stockfish 11 (pre-NNUE) would be best for this purpose.

or maybe FairStockfish:
https://fairy-stockfish.github.io/
https://github.com/fairy-stockfish/Fairy-Stockfish

Many contributors, latest release 2021,
but (because of various board layouts) no UCI interface/GUI
but GUI's as liground or winboard. For the time being (testing) no problem I think.

[hgm]

100 games and 100 draws. It's over now folks! SF and Lc0 plays perfect chess from opening position.

I had already proven that long ago: I had this engine that due to an error in the sign of the Piece-Square tables very strongly discouraged any forward moves. To my surprise that made it play perfect Chess! I had it play 100 games from the opening position against itself, and they all ended in a draw!