next chess.com handicap match

[lkaufman]

Are there any plans for future matches with material odds (or something else ) ?

Nothing else has been scheduled yet, but it is a safe bet that there will be more matches.

[JJJ]

So you weren't too generous after all

[Lyudmil Tsvetkov]

Yaacov Norowitz was unable to connect to chess.com at game time (I don't know what the problem was) and after a few minutes of trying, chess.com got Hikaru Nakamura to fill in for him!! So we played Nakamura, one of the world's best players, giving him knight for (c7) pawn odds!! Amazingly, Komodo won the game rather convincingly!! It's true that Nakamura had no advance time to prepare, and he played too rapidly, probably expecting to win easily, but the result was still remarkable. We beat a nearly 2800 FIDE player at knight for pawn, but lost 3 out of 3 to a 2100 FIDE rated player at full knight odds!

nothing surprising, Nakamura is simply too weak.

As I told you, the N for c7 pawn odds is at least twice as difficult as N for b7 pawn odds.

I have also been having some hard time with this against SF.

[Lyudmil Tsvetkov]

Komodo won game 3, the one with even material but the Black king starting on e4 (with ..f6 played). Yaacov played strangely in the opening, and by move 10 his advantage was already gone. So far I can only recall one game in which a human has managed a draw with a non-material handicap (GM Perelshteyn drew with the handicap of 1.e4 e6 2.d4 Ke7?). I've noticed that I can get Komodo evals that seem much more predictive of results vs. humans by setting the Dynamism value way down to something like 50, which makes Komodo evaluate much more like humans do.

then maybe it is altogether better to lower dynamism in Komodo?

for example, I have noticed that Komodo will frequently overvalue dynamic factors like rook on open/semiopen file, some central piece attacks, etc., to the detriment of more positional factors.

that is, however, probably irreproducible in STC games.

[duncan]

So we played Nakamura, one of the world's best players, giving him knight for (c7) pawn odds!! Amazingly, Komodo won the game rather convincingly!! It's true that Nakamura had no advance time to prepare, and he played too rapidly, probably expecting to win easily, but the result was still remarkable. We beat a nearly 2800 FIDE player at knight for pawn, but lost 3 out of 3 to a 2100 FIDE rated player at full knight odds!

would you be prepared to offer Nakamura knight for (c7) pawn odds again but this time with more time to prepare to recover his honour ?

[lkaufman]

So we played Nakamura, one of the world's best players, giving him knight for (c7) pawn odds!! Amazingly, Komodo won the game rather convincingly!! It's true that Nakamura had no advance time to prepare, and he played too rapidly, probably expecting to win easily, but the result was still remarkable. We beat a nearly 2800 FIDE player at knight for pawn, but lost 3 out of 3 to a 2100 FIDE rated player at full knight odds!

would you be prepared to offer Nakamura knight for (c7) pawn odds again but this time with more time to prepare to recover his honour ?

Sure. I'll be the first to admit that we would be underdogs in a rematch. Chess.com likes variety, so if we did a rematch at knight for pawn (or a match with some other GM that way), we might do one game with each knight for c7 pawn and one with each knight for b7 pawn. These two pawns seem the most appropriate for this sort of handicap game, because the edge pawns are not very valuable due to the immediately active rook, the "f" and "g" pawns expose the king, and the center pawns distort the openings more and also open too many lines for the bishops and queen when removed. Since full knight odds was too much even for players rated around 2100 FIDE, knight for pawn should be winnable by good GMs. But this single victory at knight for pawn over Nakamura does make me wonder about whether the prediction that no engine would ever win a serious match from a top GM at full knight odds is correct.

[Uri Blass]

So we played Nakamura, one of the world's best players, giving him knight for (c7) pawn odds!! Amazingly, Komodo won the game rather convincingly!! It's true that Nakamura had no advance time to prepare, and he played too rapidly, probably expecting to win easily, but the result was still remarkable. We beat a nearly 2800 FIDE player at knight for pawn, but lost 3 out of 3 to a 2100 FIDE rated player at full knight odds!

would you be prepared to offer Nakamura knight for (c7) pawn odds again but this time with more time to prepare to recover his honour ?

Sure. I'll be the first to admit that we would be underdogs in a rematch. Chess.com likes variety, so if we did a rematch at knight for pawn (or a match with some other GM that way), we might do one game with each knight for c7 pawn and one with each knight for b7 pawn. These two pawns seem the most appropriate for this sort of handicap game, because the edge pawns are not very valuable due to the immediately active rook, the "f" and "g" pawns expose the king, and the center pawns distort the openings more and also open too many lines for the bishops and queen when removed. Since full knight odds was too much even for players rated around 2100 FIDE, knight for pawn should be winnable by good GMs. But this single victory at knight for pawn over Nakamura does make me wonder about whether the prediction that no engine would ever win a serious match from a top GM at full knight odds is correct.

If you want to test better what is possible to do with knight odds then I suggest the following harder match for humans.

Play with full knight odd but in every move the computer suggest 3 possible moves when the human need to reply everyone of the 3 possible moves.
Only after the human reply to all moves the computer choose the best move to play based on the new knowledge.

You can give the human more time to compensate for needing to think about 3 different replies in every move but I think that even if you give the human 135+45 instead of 45+15 in normal knight odd match then it is going to be relatively harder for the human and I still expect top GM's to win.

[lkaufman]

So we played Nakamura, one of the world's best players, giving him knight for (c7) pawn odds!! Amazingly, Komodo won the game rather convincingly!! It's true that Nakamura had no advance time to prepare, and he played too rapidly, probably expecting to win easily, but the result was still remarkable. We beat a nearly 2800 FIDE player at knight for pawn, but lost 3 out of 3 to a 2100 FIDE rated player at full knight odds!

would you be prepared to offer Nakamura knight for (c7) pawn odds again but this time with more time to prepare to recover his honour ?

Sure. I'll be the first to admit that we would be underdogs in a rematch. Chess.com likes variety, so if we did a rematch at knight for pawn (or a match with some other GM that way), we might do one game with each knight for c7 pawn and one with each knight for b7 pawn. These two pawns seem the most appropriate for this sort of handicap game, because the edge pawns are not very valuable due to the immediately active rook, the "f" and "g" pawns expose the king, and the center pawns distort the openings more and also open too many lines for the bishops and queen when removed. Since full knight odds was too much even for players rated around 2100 FIDE, knight for pawn should be winnable by good GMs. But this single victory at knight for pawn over Nakamura does make me wonder about whether the prediction that no engine would ever win a serious match from a top GM at full knight odds is correct.

If you want to test better what is possible to do with knight odds then I suggest the following harder match for humans.

Play with full knight odd but in every move the computer suggest 3 possible moves when the human need to reply everyone of the 3 possible moves.
Only after the human reply to all moves the computer choose the best move to play based on the new knowledge.

You can give the human more time to compensate for needing to think about 3 different replies in every move but I think that even if you give the human 135+45 instead of 45+15 in normal knight odd match then it is going to be relatively harder for the human and I still expect top GM's to win.

That might work in theory, but it is not a practical idea for a match on chess.com. We need to stick with simple things that have a defined start position and adhere to normal chess rules. We can vary times, start positions, incentives to win or draw, and perhaps allow takebacks, but not much else. It can't be anything complicated.