Two Pawn Handicap

[lkaufman]

Yet Komodo won fairly easily playing Black and giving the f7 pawn only, which is the worst pawn for Black to remove...

May be the f7 + a time handicap (time divide by 10 or 50 on the 24 cores).
This will make the games more lively to follow.

f7 is already a huge handicap, way more than just a random pawn, In 94 self-play games at 2' + 1" on 4 threads, Komodo scored 89 wins and 5 draws, so 91.5-2.5 for plus 625 elo. If White prepares the opening it is even worse for Black. We may try giving f7 to an Elite GM but I think Komodo will be the underdog. I agree that we could give some time handicap as well as f7 to a 2500 GM, but nothing like 10 or 50 to 1. Cutting the computer's time in half would make the games more lively for the spectators; reducing it much below that wouldn't improve the experience in my opinion. Also it's not like normal chess if the human player is always on move.

[lkaufman]

Yet Komodo won fairly easily playing Black and giving the f7 pawn only, which is the worst pawn for Black to remove...

May be the f7 + a time handicap (time divide by 10 or 50 on the 24 cores).
This will make the games more lively to follow.

I think larry does not like weakening komodo strength wise that way, as we want to see komodo at its best. it is like saying play komodo on one core.
...

Nevertheless, that would a good publicity for Komodo : "At 1 second by move, Komodo with a pawn down is able to defeat a 2650 GM !!"

Komodo could give a 2650 60 to 1 time odds and White, or the f7 pawn and maybe 2 to 1 time odds, but certainly not both! Perhaps Komodo could give 60 to 1 plus the a2 pawn, since that's not so much worse than just playing Black in normal chess, but calling that a pawn handicap would be misleading. It's more like a dubious gambit.

[chessico]

Yet Komodo won fairly easily playing Black and giving the f7 pawn only, which is the worst pawn for Black to remove.

That game should not have been played, I certainly count the sixth game as a resignation without playing. I did not like the idea of these handicap matches at all, I must admit, but if you make big claims in order to sell more copies of your (fine) program, then at least stick with the rules when things don't go as expected.

The problem with handicaps in general is of course that the computer cannot come up with a strategy how to optimize its chances, depending on the situation and the handicap. Some small opening book here, some fiddling with the contempt settings there: Ironically, the whole event, instead of showing how strong engines are (and the players do not dispute that any longer, so why try to prove it?) only shows that computers cannot make a plan and are not intelligent at all.

Edit: An idea that seems to be untested: The human player can analyze by moving the pieces on the board, like he would in a home analysis, a correspondence game or in the old days when games were interrupted at move 40 and played the next day. I think it would help a lot, has it been done?

[whereagles]

That game should not have been played, I certainly count the sixth game as a resignation without playing. I did not like the idea of these handicap matches at all, I must admit, but if you make big claims in order to sell more copies of your (fine) program, then at least stick with the rules when things don't go as expected.

No one charged you to see the match. It was provided to all as free entertainment, and after Neuman found the winning strategy it lost interest. So a new idea was needed. In fact, Neuman was the one suggesting it, despite knowing he might very well lose 100 bucks on the process (which he did LOL).

[lkaufman]

Yet Komodo won fairly easily playing Black and giving the f7 pawn only, which is the worst pawn for Black to remove.

That game should not have been played, I certainly count the sixth game as a resignation without playing. I did not like the idea of these handicap matches at all, I must admit, but if you make big claims in order to sell more copies of your (fine) program, then at least stick with the rules when things don't go as expected.

The problem with handicaps in general is of course that the computer cannot come up with a strategy how to optimize its chances, depending on the situation and the handicap. Some small opening book here, some fiddling with the contempt settings there: Ironically, the whole event, instead of showing how strong engines are (and the players do not dispute that any longer, so why try to prove it?) only shows that computers cannot make a plan and are not intelligent at all.

Edit: An idea that seems to be untested: The human player can analyze by moving the pieces on the board, like he would in a home analysis, a correspondence game or in the old days when games were interrupted at move 40 and played the next day. I think it would help a lot, has it been done?

The knight sacrifice that earned Komodo one of its draws was quite spectacular, the gm said afterward that no human gm would have even seriously considered it. But that was only one game. The GM wanted to try the f7 handicap in the last game since the c and f handicap was clearly too large. Of course we wouldn't change the conditions unilaterally. I have never heard of any event (with or without computers) in which players could move the pieces around during the game (presumably on a separate pocket size board for a human vs human event). It would be a big handicap for average amateurs, not so much at master level, and very little at Elite level. A 2800 gm could probably play something like 2600 level blindfolded; I doubt that using a pocket set to analyze during the game would add more than 20-30 elo points. They have very good visualization powers. That's always been my weak point in chess; for me it would probably add a hundred elo.

[chessico]

I have never heard of any event (with or without computers) in which players could move the pieces around during the game (presumably on a separate pocket size board for a human vs human event). It would be a big handicap for average amateurs, not so much at master level, and very little at Elite level. A 2800 gm could probably play something like 2600 level blindfolded; I doubt that using a pocket set to analyze during the game would add more than 20-30 elo points. They have very good visualization powers. That's always been my weak point in chess; for me it would probably add a hundred elo.

Well, as an expert you are of course aware of Lev Polugayevskis book, for example, where he shows his analytical skills in adjourned games , i don't remember the title now, but it is a fascinating book, which shows how much can be gained from the above mentioned method in comparison to standard on the board game. Of course the mental limitations of a human vs a computer, whose board representation is always accurate by definition, no matter of five or twenty moves ahead, are significant, even on the highest level. I can play blindfolded myself, so I am aware that also people who can do that are in a disadvantage against a machine, of course.
In practice it would certainly take some adjustment and some training, but I don't see why such an event would be less interesting than the kind of games that have been played recently in these handicap matches.

[Uri Blass]

Yet Komodo won fairly easily playing Black and giving the f7 pawn only, which is the worst pawn for Black to remove.

That game should not have been played, I certainly count the sixth game as a resignation without playing. I did not like the idea of these handicap matches at all, I must admit, but if you make big claims in order to sell more copies of your (fine) program, then at least stick with the rules when things don't go as expected.

The problem with handicaps in general is of course that the computer cannot come up with a strategy how to optimize its chances, depending on the situation and the handicap. Some small opening book here, some fiddling with the contempt settings there: Ironically, the whole event, instead of showing how strong engines are (and the players do not dispute that any longer, so why try to prove it?) only shows that computers cannot make a plan and are not intelligent at all.

Edit: An idea that seems to be untested: The human player can analyze by moving the pieces on the board, like he would in a home analysis, a correspondence game or in the old days when games were interrupted at move 40 and played the next day. I think it would help a lot, has it been done?

The knight sacrifice that earned Komodo one of its draws was quite spectacular, the gm said afterward that no human gm would have even seriously considered it. But that was only one game. The GM wanted to try the f7 handicap in the last game since the c and f handicap was clearly too large. Of course we wouldn't change the conditions unilaterally. I have never heard of any event (with or without computers) in which players could move the pieces around during the game (presumably on a separate pocket size board for a human vs human event). It would be a big handicap for average amateurs, not so much at master level, and very little at Elite level. A 2800 gm could probably play something like 2600 level blindfolded; I doubt that using a pocket set to analyze during the game would add more than 20-30 elo points. They have very good visualization powers. That's always been my weak point in chess; for me it would probably add a hundred elo.

What about allowing the human to make notes during the game by using a computer not to analyze the game by chess engines but to make moves forward and backward(you can give the human also a lot of time and decide that the human has 9 hours for all the game when Komodo has only 1 hour for all the game).

Even if the human can visualize many moves forward with no problem it does not mean that he can memorize all his calculations at long time control and watching the lines that the human analyze may be interesting for the spectators.

[lkaufman]

I have never heard of any event (with or without computers) in which players could move the pieces around during the game (presumably on a separate pocket size board for a human vs human event). It would be a big handicap for average amateurs, not so much at master level, and very little at Elite level. A 2800 gm could probably play something like 2600 level blindfolded; I doubt that using a pocket set to analyze during the game would add more than 20-30 elo points. They have very good visualization powers. That's always been my weak point in chess; for me it would probably add a hundred elo.

Well, as an expert you are of course aware of Lev Polugayevskis book, for example, where he shows his analytical skills in adjourned games , i don't remember the title now, but it is a fascinating book, which shows how much can be gained from the above mentioned method in comparison to standard on the board game. Of course the mental limitations of a human vs a computer, whose board representation is always accurate by definition, no matter of five or twenty moves ahead, are significant, even on the highest level. I can play blindfolded myself, so I am aware that also people who can do that are in a disadvantage against a machine, of course.
In practice it would certainly take some adjustment and some training, but I don't see why such an event would be less interesting than the kind of games that have been played recently in these handicap matches.

The reason analysis in books (pre-computer) was so much better than actual play is not primarily due to being able to move the pieces, but to spending hours rather than minutes on the key moves. Letting a 2800 player analyze for say one hour per move with a board for a game played over a week or so (with computer taking say one hour total) would indeed be an interesting handicap, but too expensive for us to sponsor.

[chessico]

The reason analysis in books (pre-computer) was so much better than actual play is not primarily due to being able to move the pieces, but to spending hours rather than minutes on the key moves. Letting a 2800 player analyze for say one hour per move with a board for a game played over a week or so (with computer taking say one hour total) would indeed be an interesting handicap, but too expensive for us to sponsor.

More time used per move is a factor, clearly, but else I disagree. Isn't it quite common that both players have used a lot of time during the game analyzing a wrong continuation, and post mortem it is a matter of seconds rather than minutes that they see what is now obvious? Making the first 5 moves of a line and then starting the analysis, moving the pieces, should not give better results than having to calculate it all in the mind? I would be very surprised if this were the case.

[lkaufman]

The reason analysis in books (pre-computer) was so much better than actual play is not primarily due to being able to move the pieces, but to spending hours rather than minutes on the key moves. Letting a 2800 player analyze for say one hour per move with a board for a game played over a week or so (with computer taking say one hour total) would indeed be an interesting handicap, but too expensive for us to sponsor.

More time used per move is a factor, clearly, but else I disagree. Isn't it quite common that both players have used a lot of time during the game analyzing a wrong continuation, and post mortem it is a matter of seconds rather than minutes that they see what is now obvious? Making the first 5 moves of a line and then starting the analysis, moving the pieces, should not give better results than having to calculate it all in the mind? I would be very surprised if this were the case.

I'm just saying that it's a much smaller factor for 2800 players than for 2200 players. Often they find better moves in postmortem because they learned something from the actual game. Or they find better moves deep on some line, that don't change the root move.