Stockfish play very decent giving pawn odds.........

[lkaufman]

Here is the game where Komodo defeated GM Izoria (2604 FIDE) in 2018 at the same handicap (c2,g2) with which van Foreest defeated Stockfish. The Izoria game was at a 50% longer time control (45' + 15") than the van Foreest game.

https://lichess.org/6CwiDyrR#48

[Ovyron]

Chessqueen: Please stop removing the code at the beginning of your posts, it's required for quote boxes to work, and such code removal may cause the thread to stop working later (if PGN tags are used and the forum gets stuck trying to process quote boxes' content as chess games, like the other day.)

[Chessqueen]

Here is the game where Komodo defeated GM Izoria (2604 FIDE) in 2018 at the same handicap (c2,g2) with which van Foreest defeated Stockfish. The Izoria game was at a 50% longer time control (45' + 15") than the van Foreest game.

https://lichess.org/6CwiDyrR#48

I am starting to believe that because GM Foreest trained for three weeks according to what he said minutes after he beat Stockfish that even if he used a much weaker hardware that experience of training with stockfish gave hima deep understanding of how to beat Stockfish, he really did not needed a super fast computer, he could just have used an intel i7 single or quad cores, all that he head to do was to play game in standard time control or game in 120 minutes instead of TC 30 + 10. I really do not know for how long GM Izora (2604 FIDE) trained against Komodo, but it is very important and it makes a hell of a difference to train for 2 or three weeks versus NOT training enough. There are a lot of GM that are low rated in the Mid 2450 like Jorge Sammour but since he trained for a few games + 50 games that makes him more dangerous than a GM who is rated FIDE 2600 who has not played enough odds games.

[lkaufman]

Here is the game where Komodo defeated GM Izoria (2604 FIDE) in 2018 at the same handicap (c2,g2) with which van Foreest defeated Stockfish. The Izoria game was at a 50% longer time control (45' + 15") than the van Foreest game.

https://lichess.org/6CwiDyrR#48

I am starting to believe that because GM Foreest trained for three weeks according to what he said minutes after he beat Stockfish that even if he used a much weaker hardware that experience of training with stockfish gave hima deep understanding of how to beat Stockfish, he really did not needed a super fast computer, he could just have used an intel i7 single or quad cores, all that he head to do was to play game in standard time control or game in 120 minutes instead of TC 30 + 10. I really do not know for how long GM Izora (2604 FIDE) trained against Komodo, but it is very important and it makes a hell of a difference to train for 2 or three weeks versus NOT training enough. There are a lot of GM that are low rated in the Mid 2450 like Jorge Sammour but since he trained for a few games + 50 games that makes him more dangerous than a GM who is rated FIDE 2600 who has not played enough odds games.

I am sure that you are right about preparation/training being very helpful in these handicap matches, but in the specific case of the comparison of the two games at c2/g2 handicap, I think that the results have more to do with Komodo's decision to castle long vs. Stockfish's decision to castle short. It seems quite clear that Komodo made the better choice on this point. Incidentally, I tried some Rapid knight odds games with the recent Lc0 70/70 series networks, and it played them pretty well; I come out ahead, as I do with Komodo MCTS, but about even if we count draws as losses for me. The Lc0 networks starting with 2 thru 6 probably couldn't beat a 1500 player at knight odds, they just blunder, but these networks play the handicap in the right style, very agressively but not insanely. I think they are a bit too non-materialistic, but I could be wrong. So Komodo MCTS may have a legitimate current rival for best engine at giving handicaps to humans. I'll need more games at more handicaps to have a definite opinion.

[Chessqueen]

Here is the game where Komodo defeated GM Izoria (2604 FIDE) in 2018 at the same handicap (c2,g2) with which van Foreest defeated Stockfish. The Izoria game was at a 50% longer time control (45' + 15") than the van Foreest game.

https://lichess.org/6CwiDyrR#48

I am starting to believe that because GM Foreest trained for three weeks according to what he said minutes after he beat Stockfish that even if he used a much weaker hardware that experience of training with stockfish gave hima deep understanding of how to beat Stockfish, he really did not needed a super fast computer, he could just have used an intel i7 single or quad cores, all that he head to do was to play game in standard time control or game in 120 minutes instead of TC 30 + 10. I really do not know for how long GM Izora (2604 FIDE) trained against Komodo, but it is very important and it makes a hell of a difference to train for 2 or three weeks versus NOT training enough. There are a lot of GM that are low rated in the Mid 2450 like Jorge Sammour but since he trained for a few games + 50 games that makes him more dangerous than a GM who is rated FIDE 2600 who has not played enough odds games.

I am sure that you are right about preparation/training being very helpful in these handicap matches, but in the specific case of the comparison of the two games at c2/g2 handicap, I think that the results have more to do with Komodo's decision to castle long vs. Stockfish's decision to castle short. It seems quite clear that Komodo made the better choice on this point. Incidentally, I tried some Rapid knight odds games with the recent Lc0 70/70 series networks, and it played them pretty well; I come out ahead, as I do with Komodo MCTS, but about even if we count draws as losses for me. The Lc0 networks starting with 2 thru 6 probably couldn't beat a 1500 player at knight odds, they just blunder, but these networks play the handicap in the right style, very agressively but not insanely. I think they are a bit too non-materialistic, but I could be wrong. So Komodo MCTS may have a legitimate current rival for best engine at giving handicaps to humans. I'll need more games at more handicaps to have a definite opinion.

You can still challenge a low FIDE 2575 to 2600 versus Komodo by using a combination of material odds plus time odds for instance those first positions that I posted plus time odds 30+10 to the low GM and 20 + 10 to Komodo

1 st game b2 + f2 odds ( 30 + 10 to the low GM and 20 + 10 to Komodo)
2nd game b2 + g2 odds (30 + 10 to the low GM and 20 + 10 to Komodo)
3rd game c2 + f2 odds (30 + 10 to the low GM and 20 + 10 to Komodo)
4th game c2 + g2 odds (30 + 10 to the low GM and 20 + 10 to Komodo)

[lkaufman]

Here is the game where Komodo defeated GM Izoria (2604 FIDE) in 2018 at the same handicap (c2,g2) with which van Foreest defeated Stockfish. The Izoria game was at a 50% longer time control (45' + 15") than the van Foreest game.

https://lichess.org/6CwiDyrR#48

I am starting to believe that because GM Foreest trained for three weeks according to what he said minutes after he beat Stockfish that even if he used a much weaker hardware that experience of training with stockfish gave hima deep understanding of how to beat Stockfish, he really did not needed a super fast computer, he could just have used an intel i7 single or quad cores, all that he head to do was to play game in standard time control or game in 120 minutes instead of TC 30 + 10. I really do not know for how long GM Izora (2604 FIDE) trained against Komodo, but it is very important and it makes a hell of a difference to train for 2 or three weeks versus NOT training enough. There are a lot of GM that are low rated in the Mid 2450 like Jorge Sammour but since he trained for a few games + 50 games that makes him more dangerous than a GM who is rated FIDE 2600 who has not played enough odds games.

I am sure that you are right about preparation/training being very helpful in these handicap matches, but in the specific case of the comparison of the two games at c2/g2 handicap, I think that the results have more to do with Komodo's decision to castle long vs. Stockfish's decision to castle short. It seems quite clear that Komodo made the better choice on this point. Incidentally, I tried some Rapid knight odds games with the recent Lc0 70/70 series networks, and it played them pretty well; I come out ahead, as I do with Komodo MCTS, but about even if we count draws as losses for me. The Lc0 networks starting with 2 thru 6 probably couldn't beat a 1500 player at knight odds, they just blunder, but these networks play the handicap in the right style, very agressively but not insanely. I think they are a bit too non-materialistic, but I could be wrong. So Komodo MCTS may have a legitimate current rival for best engine at giving handicaps to humans. I'll need more games at more handicaps to have a definite opinion.

You can still challenge a low FIDE 2575 to 2600 versus Komodo by using a combination of material odds plus time odds for instance those first positions that I posted plus time odds 30+10 to the low GM and 15+ 5 to Komodo

Actually I am convinced now that moderate time odds like double or triple do not reduce the chances of Komodo (or Stockfish or Lc0) at all when giving material handicaps to humans, in fact it's probably good strategy. The strength difference is still so huge that the slightly less perfect engine play is easily offset by giving the human less free ponder time. Actually in our test games with Lenderman we used a Komodo version that played about twice as fast as normal, and it scored 2.5 out of 4 at the van Foreest handicaps if you discount the loss of the final dead drawn game due to interface issues. I even advised the Stockfish team for the van Foreest match to set Stockfish slowmover to play fast, but for technical/contractual reasons they didn't do so.

[Chessqueen]

Actually, it looks to me like the T70/71 nets will be just as good as or better than the T10 nets at handicap play, because they apparently stopped resigning their training games early. The PVs and evals for handicap positions look fine now. Although I think that Komodo MCTS is better than Stockfish in handicap play vs. humans, T70/71 might turn out to be even better with a powerful GPU. In any case any future events should be based on who wins the match or on the score, not on winning a single game, which is relatively easy if the handicap is at all fair.

Komodo at the same TC that stockfish played vs GM Foreest should beat GM Forrest if he is willing to take a smaller amount of $Money, even a reasonable amount like $2,500.00, if he wins at least 2 games with either the b2, c2 or d2, plus the f2 and g7 pawns combined removed, but if he Manage to win at least 1 game out of the 6 games, he should win at least $1,500.00. Honestly this is NOT a bad offer to make to a FIDE 2660+ for a couple of hours to play versus an Engines. Thre games to be played on any Saturday followed by the other three games the next Sunday. The 6 positions should be something like this for instance ..........

1 st game b2 + f2 odds
2nd game b2 + g2 odds
3rd game c2 + f2 odds
4th game c2 + g2 odds

Actually we already made him an offer, but we have not reached an agreement. Aside from talking about money, I basically like your idea, except for two points. 1. I have no interest in rewarding wins without penalizing losses. It is the net result, wins minus losses, that must decide any prize. Whether it's just based on winning the match or on the margin of victory is a detail. Of course if the handicap is clearly too large or too small, then it might be appropriate to count draws as wins for the underdog, but we are talking about finding a fair handicap here. At knight odds, the player, even a low rated GM, should have to actually win more than half the games, but at two pawns like this I don't even know which side is the favorite to win the match by normal scoring. 2. I think the center pawns are less suitable for handicaps, not because they make the handicap too big (that's not clear, as their absence opens lines for bishops and queen and affects knights and king), but because they simply change the game more than the removal of the bishop's and knight's pawns, whose removal only affects the mobility of a single piece. So I think two pawns should always mean the first four handicaps you list (keeping balance one per each wing), and "two pawns and move" would be the same handicaps with Black pawns removed (suitable for a 2500 level GM perhaps). Maybe an eight game match repeating the four handicaps once would be ideal.

I believe that GM Foreest got paid too much for the Stockfish match, but if you offer him $500.00 for every win and for him to paid $100.00 for every game that he lose against Komodo that is very reasonable. Even if Komodo draw 5 games, but Draw not counting for either side, since I expect Komodo to draw at least 5 games, which would be a great promotion for Komodo, specially for Komodo to do better than Stockfish. There should be at least one game with the same position that Stockfish lost to GM Foreest, so chess fans and future Komodo customers can use it for comparison reason.

I would agree to that, and yes, any two pawn match we play should include the c2,g2 handicap, which is one of the four "balanced" ones anyway. But having just won (reportedly) $20,000, he is not well motivated to play a similar match for a likely payout of maybe $1200 (if 3 wins each and 2 draws). The prize may have created the false impression that the publicity is worth many thousands of dollars to an engine, whereas in reality I would have my doubts about whether we would actually recoup even one thousand dollars extra from such a match. But we also learn things from these matches, so I don't try to justify every penny spent, and I do like to help promote chess in general and computer chess in particular.
[/quote]


You are right about that, but if GM Foreest does NOT agree you can always get a higher or equal to GM Foreest from India who will be willing to take even $250.00 for each win to probably $50.00 for each Komodo lost since people from India do NOT make too much Money and for a GM from India to give $50.00 it might be too much, but the quality of play would be no less.Simply make the offer via Chessbase India.com to any GM from 2650-2690 and you will find plenty of reply and willing to take your challenge.
[/quote]

Yes, one way or another we'll probably have a two pawn rapid handicap match with a pretty strong GM before too long. In the unlikely event that the GM won no games and lost some, I wouldn't actually ask him to pay anything, the loss money would only be subtracted from winnings. Or he could just play more games with the same conditions. Keep in mind that we have already had six such matches, three with GMs around 2500, and three with GMs around 2600 around 2017 (Erenburg, Lenderman and Izoria), all at 45' + 15". Komodo scored very well with White except at c2 + f2 in those matches. But we haven't tried two pawns against 2700 level players, except for one draw at f2 + g2 vs. MVL in a much faster game (15' + 2" I believe). Playing 30" + 10" vs. a 2700 with two pawns off including f2 or g2 every time would be a challenge.
[/quote]

These are some candidates from India:
Sethuraman S.P Standard Rating 2651
Sasikiran, Krishnan Standard Rating 2678
Adhiban, B Standard Rating 2689
[/quote]


Technically Adhiban, B Rapid rating which is the one that should be used for this type of match instead of their standard rating, is higher than GM Foreest.
https://ratings.fide.com/card.phtml?event=5018471 GM Adhiban, B Rapids Rating = 2624

https://ratings.fide.com/card.phtml?event=1039784 GM Foreest Rapid Rating = 2543
[/quote]

Well, based on that Lenderman and others we've played are already above him in Rapid. But this is misleading, because he only plays about one rapid event per year, so his Rapid rating is unrealistically low. I'm not sure what ratings most accurately reflect all the top level big money online rapid events being played now due to the COVID crisis, someone should rate them all in one system. Regarding the players you mention, I drew with Sasikiran over 20 years ago, and after the game taught him to play shogi. In view of the concerns about cheating, we should only consider players who have played in major online events and have not been suspected of cheating. I don't follow these closely enough to know which players (if any) are under suspicion, even if nothing has been proven. Even with a webcam on you, it is certainly possible to cheat. Fortunately it seems that most of the top players are honest.
[/quote]


Larry, you can create a Playkomodo Donation just like this http://talkchess.com/forum3/viewtopic.php?f=2&t=67083 Talkchess Donation, and I know that there are a lot of people here that want to see GM Foreest Crush, so create a donation from $5.00 up to $50.00 where I can donate and many other who simply like to see Komodo playing Odds and those that Donate at least will get the next version of Komodo sent to their Email
[/quote]

I'd have to get our webmaster to do that, and it wouldn't make much sense to give someone a free copy of new Komodo for a donation of less than the discounted price to buy it, but maybe we'll do something like that. If you want to contribute $48 (the discount price for Komodo) towards such a match just order Komodo 14 (even though you already have it), and I'll send you Komodo 14.1 when it comes out for free (remind me if I forget), same for anyone else who we show as having Komodo 14. Just keep in mind that the actual opponent may be someone else depending on who is interested and willing to play for reasonable rewards.
[/quote]

What would be a reasonable rewards, or should I say for a player very close to 2700 what are your limit as far a rewards ? I know these players wold charge you more than players from India

7 Kamsky, Gata g USA 2674 0 1974
8 Robson, Ray g USA 2673 0 1994
9 Sevian, Samuel g USA 2660 0 2000
10 Swiercz, Dariusz USA 2649 0 1994

[lkaufman]

Actually, it looks to me like the T70/71 nets will be just as good as or better than the T10 nets at handicap play, because they apparently stopped resigning their training games early. The PVs and evals for handicap positions look fine now. Although I think that Komodo MCTS is better than Stockfish in handicap play vs. humans, T70/71 might turn out to be even better with a powerful GPU. In any case any future events should be based on who wins the match or on the score, not on winning a single game, which is relatively easy if the handicap is at all fair.

Komodo at the same TC that stockfish played vs GM Foreest should beat GM Forrest if he is willing to take a smaller amount of $Money, even a reasonable amount like $2,500.00, if he wins at least 2 games with either the b2, c2 or d2, plus the f2 and g7 pawns combined removed, but if he Manage to win at least 1 game out of the 6 games, he should win at least $1,500.00. Honestly this is NOT a bad offer to make to a FIDE 2660+ for a couple of hours to play versus an Engines. Thre games to be played on any Saturday followed by the other three games the next Sunday. The 6 positions should be something like this for instance ..........

1 st game b2 + f2 odds
2nd game b2 + g2 odds
3rd game c2 + f2 odds
4th game c2 + g2 odds

Actually we already made him an offer, but we have not reached an agreement. Aside from talking about money, I basically like your idea, except for two points. 1. I have no interest in rewarding wins without penalizing losses. It is the net result, wins minus losses, that must decide any prize. Whether it's just based on winning the match or on the margin of victory is a detail. Of course if the handicap is clearly too large or too small, then it might be appropriate to count draws as wins for the underdog, but we are talking about finding a fair handicap here. At knight odds, the player, even a low rated GM, should have to actually win more than half the games, but at two pawns like this I don't even know which side is the favorite to win the match by normal scoring. 2. I think the center pawns are less suitable for handicaps, not because they make the handicap too big (that's not clear, as their absence opens lines for bishops and queen and affects knights and king), but because they simply change the game more than the removal of the bishop's and knight's pawns, whose removal only affects the mobility of a single piece. So I think two pawns should always mean the first four handicaps you list (keeping balance one per each wing), and "two pawns and move" would be the same handicaps with Black pawns removed (suitable for a 2500 level GM perhaps). Maybe an eight game match repeating the four handicaps once would be ideal.

I believe that GM Foreest got paid too much for the Stockfish match, but if you offer him $500.00 for every win and for him to paid $100.00 for every game that he lose against Komodo that is very reasonable. Even if Komodo draw 5 games, but Draw not counting for either side, since I expect Komodo to draw at least 5 games, which would be a great promotion for Komodo, specially for Komodo to do better than Stockfish. There should be at least one game with the same position that Stockfish lost to GM Foreest, so chess fans and future Komodo customers can use it for comparison reason.

I would agree to that, and yes, any two pawn match we play should include the c2,g2 handicap, which is one of the four "balanced" ones anyway. But having just won (reportedly) $20,000, he is not well motivated to play a similar match for a likely payout of maybe $1200 (if 3 wins each and 2 draws). The prize may have created the false impression that the publicity is worth many thousands of dollars to an engine, whereas in reality I would have my doubts about whether we would actually recoup even one thousand dollars extra from such a match. But we also learn things from these matches, so I don't try to justify every penny spent, and I do like to help promote chess in general and computer chess in particular.

You are right about that, but if GM Foreest does NOT agree you can always get a higher or equal to GM Foreest from India who will be willing to take even $250.00 for each win to probably $50.00 for each Komodo lost since people from India do NOT make too much Money and for a GM from India to give $50.00 it might be too much, but the quality of play would be no less.Simply make the offer via Chessbase India.com to any GM from 2650-2690 and you will find plenty of reply and willing to take your challenge.
[/quote]

Yes, one way or another we'll probably have a two pawn rapid handicap match with a pretty strong GM before too long. In the unlikely event that the GM won no games and lost some, I wouldn't actually ask him to pay anything, the loss money would only be subtracted from winnings. Or he could just play more games with the same conditions. Keep in mind that we have already had six such matches, three with GMs around 2500, and three with GMs around 2600 around 2017 (Erenburg, Lenderman and Izoria), all at 45' + 15". Komodo scored very well with White except at c2 + f2 in those matches. But we haven't tried two pawns against 2700 level players, except for one draw at f2 + g2 vs. MVL in a much faster game (15' + 2" I believe). Playing 30" + 10" vs. a 2700 with two pawns off including f2 or g2 every time would be a challenge.
[/quote]

These are some candidates from India:
Sethuraman S.P Standard Rating 2651
Sasikiran, Krishnan Standard Rating 2678
Adhiban, B Standard Rating 2689
[/quote]


Technically Adhiban, B Rapid rating which is the one that should be used for this type of match instead of their standard rating, is higher than GM Foreest.
https://ratings.fide.com/card.phtml?event=5018471 GM Adhiban, B Rapids Rating = 2624

https://ratings.fide.com/card.phtml?event=1039784 GM Foreest Rapid Rating = 2543
[/quote]

Well, based on that Lenderman and others we've played are already above him in Rapid. But this is misleading, because he only plays about one rapid event per year, so his Rapid rating is unrealistically low. I'm not sure what ratings most accurately reflect all the top level big money online rapid events being played now due to the COVID crisis, someone should rate them all in one system. Regarding the players you mention, I drew with Sasikiran over 20 years ago, and after the game taught him to play shogi. In view of the concerns about cheating, we should only consider players who have played in major online events and have not been suspected of cheating. I don't follow these closely enough to know which players (if any) are under suspicion, even if nothing has been proven. Even with a webcam on you, it is certainly possible to cheat. Fortunately it seems that most of the top players are honest.
[/quote]


Larry, you can create a Playkomodo Donation just like this http://talkchess.com/forum3/viewtopic.php?f=2&t=67083 Talkchess Donation, and I know that there are a lot of people here that want to see GM Foreest Crush, so create a donation from $5.00 up to $50.00 where I can donate and many other who simply like to see Komodo playing Odds and those that Donate at least will get the next version of Komodo sent to their Email
[/quote]

I'd have to get our webmaster to do that, and it wouldn't make much sense to give someone a free copy of new Komodo for a donation of less than the discounted price to buy it, but maybe we'll do something like that. If you want to contribute $48 (the discount price for Komodo) towards such a match just order Komodo 14 (even though you already have it), and I'll send you Komodo 14.1 when it comes out for free (remind me if I forget), same for anyone else who we show as having Komodo 14. Just keep in mind that the actual opponent may be someone else depending on who is interested and willing to play for reasonable rewards.
[/quote]

What would be a reasonable rewards, or should I say for a player very close to 2700 what are your limit as far a rewards ? I know these players wold charge you more than players from India

7 Kamsky, Gata g USA 2674 0 1974
8 Robson, Ray g USA 2673 0 1994
9 Sevian, Samuel g USA 2660 0 2000
10 Swiercz, Dariusz USA 2649 0 1994
[/quote]

It depends on what coverage we would get. I think that to get significant coverage we would need someone over 2700 FIDE (unless it's van Foreest); the players you mention aren't rated that far above Lenderman and Izoria, with whom we've already done well at equal or greater handicaps and time controls. It would really be worthwhile if we got someone in the world top twenty (rapid or standard).

[Chessqueen]

Actually, it looks to me like the T70/71 nets will be just as good as or better than the T10 nets at handicap play, because they apparently stopped resigning their training games early. The PVs and evals for handicap positions look fine now. Although I think that Komodo MCTS is better than Stockfish in handicap play vs. humans, T70/71 might turn out to be even better with a powerful GPU. In any case any future events should be based on who wins the match or on the score, not on winning a single game, which is relatively easy if the handicap is at all fair.

Komodo at the same TC that stockfish played vs GM Foreest should beat GM Forrest if he is willing to take a smaller amount of $Money, even a reasonable amount like $2,500.00, if he wins at least 2 games with either the b2, c2 or d2, plus the f2 and g7 pawns combined removed, but if he Manage to win at least 1 game out of the 6 games, he should win at least $1,500.00. Honestly this is NOT a bad offer to make to a FIDE 2660+ for a couple of hours to play versus an Engines. Thre games to be played on any Saturday followed by the other three games the next Sunday. The 6 positions should be something like this for instance ..........

1 st game b2 + f2 odds
2nd game b2 + g2 odds
3rd game c2 + f2 odds
4th game c2 + g2 odds

Actually we already made him an offer, but we have not reached an agreement. Aside from talking about money, I basically like your idea, except for two points. 1. I have no interest in rewarding wins without penalizing losses. It is the net result, wins minus losses, that must decide any prize. Whether it's just based on winning the match or on the margin of victory is a detail. Of course if the handicap is clearly too large or too small, then it might be appropriate to count draws as wins for the underdog, but we are talking about finding a fair handicap here. At knight odds, the player, even a low rated GM, should have to actually win more than half the games, but at two pawns like this I don't even know which side is the favorite to win the match by normal scoring. 2. I think the center pawns are less suitable for handicaps, not because they make the handicap too big (that's not clear, as their absence opens lines for bishops and queen and affects knights and king), but because they simply change the game more than the removal of the bishop's and knight's pawns, whose removal only affects the mobility of a single piece. So I think two pawns should always mean the first four handicaps you list (keeping balance one per each wing), and "two pawns and move" would be the same handicaps with Black pawns removed (suitable for a 2500 level GM perhaps). Maybe an eight game match repeating the four handicaps once would be ideal.

I believe that GM Foreest got paid too much for the Stockfish match, but if you offer him $500.00 for every win and for him to paid $100.00 for every game that he lose against Komodo that is very reasonable. Even if Komodo draw 5 games, but Draw not counting for either side, since I expect Komodo to draw at least 5 games, which would be a great promotion for Komodo, specially for Komodo to do better than Stockfish. There should be at least one game with the same position that Stockfish lost to GM Foreest, so chess fans and future Komodo customers can use it for comparison reason.

I would agree to that, and yes, any two pawn match we play should include the c2,g2 handicap, which is one of the four "balanced" ones anyway. But having just won (reportedly) $20,000, he is not well motivated to play a similar match for a likely payout of maybe $1200 (if 3 wins each and 2 draws). The prize may have created the false impression that the publicity is worth many thousands of dollars to an engine, whereas in reality I would have my doubts about whether we would actually recoup even one thousand dollars extra from such a match. But we also learn things from these matches, so I don't try to justify every penny spent, and I do like to help promote chess in general and computer chess in particular.

You are right about that, but if GM Foreest does NOT agree you can always get a higher or equal to GM Foreest from India who will be willing to take even $250.00 for each win to probably $50.00 for each Komodo lost since people from India do NOT make too much Money and for a GM from India to give $50.00 it might be too much, but the quality of play would be no less.Simply make the offer via Chessbase India.com to any GM from 2650-2690 and you will find plenty of reply and willing to take your challenge.

Yes, one way or another we'll probably have a two pawn rapid handicap match with a pretty strong GM before too long. In the unlikely event that the GM won no games and lost some, I wouldn't actually ask him to pay anything, the loss money would only be subtracted from winnings. Or he could just play more games with the same conditions. Keep in mind that we have already had six such matches, three with GMs around 2500, and three with GMs around 2600 around 2017 (Erenburg, Lenderman and Izoria), all at 45' + 15". Komodo scored very well with White except at c2 + f2 in those matches. But we haven't tried two pawns against 2700 level players, except for one draw at f2 + g2 vs. MVL in a much faster game (15' + 2" I believe). Playing 30" + 10" vs. a 2700 with two pawns off including f2 or g2 every time would be a challenge.
[/quote]

These are some candidates from India:
Sethuraman S.P Standard Rating 2651
Sasikiran, Krishnan Standard Rating 2678
Adhiban, B Standard Rating 2689
[/quote]


Technically Adhiban, B Rapid rating which is the one that should be used for this type of match instead of their standard rating, is higher than GM Foreest.
https://ratings.fide.com/card.phtml?event=5018471 GM Adhiban, B Rapids Rating = 2624

https://ratings.fide.com/card.phtml?event=1039784 GM Foreest Rapid Rating = 2543
[/quote]

Well, based on that Lenderman and others we've played are already above him in Rapid. But this is misleading, because he only plays about one rapid event per year, so his Rapid rating is unrealistically low. I'm not sure what ratings most accurately reflect all the top level big money online rapid events being played now due to the COVID crisis, someone should rate them all in one system. Regarding the players you mention, I drew with Sasikiran over 20 years ago, and after the game taught him to play shogi. In view of the concerns about cheating, we should only consider players who have played in major online events and have not been suspected of cheating. I don't follow these closely enough to know which players (if any) are under suspicion, even if nothing has been proven. Even with a webcam on you, it is certainly possible to cheat. Fortunately it seems that most of the top players are honest.
[/quote]


Larry, you can create a Playkomodo Donation just like this http://talkchess.com/forum3/viewtopic.php?f=2&t=67083 Talkchess Donation, and I know that there are a lot of people here that want to see GM Foreest Crush, so create a donation from $5.00 up to $50.00 where I can donate and many other who simply like to see Komodo playing Odds and those that Donate at least will get the next version of Komodo sent to their Email
[/quote]

I'd have to get our webmaster to do that, and it wouldn't make much sense to give someone a free copy of new Komodo for a donation of less than the discounted price to buy it, but maybe we'll do something like that. If you want to contribute $48 (the discount price for Komodo) towards such a match just order Komodo 14 (even though you already have it), and I'll send you Komodo 14.1 when it comes out for free (remind me if I forget), same for anyone else who we show as having Komodo 14. Just keep in mind that the actual opponent may be someone else depending on who is interested and willing to play for reasonable rewards.
[/quote]

What would be a reasonable rewards, or should I say for a player very close to 2700 what are your limit as far a rewards ? I know these players wold charge you more than players from India

7 Kamsky, Gata g USA 2674 0 1974
8 Robson, Ray g USA 2673 0 1994
9 Sevian, Samuel g USA 2660 0 2000
10 Swiercz, Dariusz USA 2649 0 1994
[/quote]

It depends on what coverage we would get. I think that to get significant coverage we would need someone over 2700 FIDE (unless it's van Foreest); the players you mention aren't rated that far above Lenderman and Izoria, with whom we've already done well at equal or greater handicaps and time controls. It would really be worthwhile if we got someone in the world top twenty (rapid or standard).
[/quote]

If you do get somebody between 2700 thru 2730 from any part of the World you believe that Komodo can still give those four positions odds combined with twice the time handicap for instance 30 + 10 for the GM Human player and 15 + 5 to Komodo and still come ahead ? Remember even if Komodo play fast that does not mean that the human GM has to play fast, he can take his sweet time and make the most use of his TC, there is no need to play as fast as Komodo and human GM knows how to take advantage of a time handicap until the very last seconds left

[JJJ]

Just by curiosity Larry, if Carslen would agree to play at some handicap, what would you choose ?