Next Komodo vs GM match - 2 pawns

[lkaufman]

The next Komodo vs GM handicap match will again be against GM Sergey Erenburg (FIDE 2585, USCF 2677, born 1983), who lost 3.5 to 0.5 last weekend in normal chess playing standard 90' + 30" vs Komodo on one core only playing blitz (3' + 1"), with no ponder for komodo, no tablebases, no opening book past move 3, and White pieces every game to the GM.
So this time (July 15 and 16, noon and 5 pm Eastern time each day, games on chess.com) we are reverting to material handicaps, with our traditional 45' + 15" time control for each side and no restrictions on Komodo, which will use 24 cores. Since Komodo has yet to lose a single game at odds of the f7 pawn, even with two moves for White (Simon Williams did win with f7 and THREE moves), we are again going to try two pawn handicap as we have in three earlier matches. Komodo always plays White,removes the e2 pawn, and also the b2,c2,d2, and f2 pawns (once each, four games total). None of these handicaps have ever been tried in any game on record to the best of my knowledge.
Previous two pawn handicap results were a 4 to 1 loss to GM Petr Neuman (2455 FIDE) at c2 + f2, a 4.5 to 1.5 win against GM Martin Petr (2515) at a2 plus b2,d2, and e2 rotating, and a 1.5 to 0.5 win against GM Victor Mikhalevski (2534 FIDE) at c2 plus b2 once, d2 once. The Petr match showed that the edge pawns are pretty small handicaps for obvious reasons, but the overall score in the matches not involving the edge pawns is 2.5 for Komodo to 4.5 for GMs. Erenburg will be the highest rated GM we have ever given two pawns to, and these are non-edge pawns. In our favor is the fact that the handicap will vary each time, so no repeating successful openings. I think we are underdogs this time, what do all of you think?

[Nay Lin Tun]

In the previous 1 core game, I already expected 3.5 or 4 in favour of Komodo as I know how strong was the software even on my first smart phone HTC HD2 years ago. This result will not much depend on human rating but depend how he have prepared with this odd! I mean even 2200 player who analysed hundreds of games for this position will get a good positional advantage enough to win than some with 2500 who havent trained in this position.
Guess- 2.5-1.5 in favour of Komodo.

[MikeB]

Definitely beatable

[pgn][Event "Computer Chess Game"]
[Site "Mac-Pro.local"]
[Date "2016.06.30"]
[Round "-"]
[White "Komodo 10 64-bit"]
[Black "Mikhail 0.04 OSx 64"]
[Result "0-1"]
[TimeControl "300+5"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P1PP1PPP/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]


1. Nf3 {-1.26/17} Nf6 {+1.12/18 3} 2. Ba3 {-1.25/18 3} d5 {+1.22/18 3} 3.
d4 {-1.20/18 3} Bg4 {+1.30/17 5} 4. Nbd2 {-1.18/17 2.2} Ne4 {+1.31/16 0.5}
5. Qb1 {-0.93/18 2.3} Nxd2 {+1.12/15 0.7} 6. Nxd2 {-0.99/19 1.5} Qc8
{+1.09/17 0.7} 7. Bd3 {-0.97/19 2.5} Qe6+ {+1.29/18 4} 8. Kf1 {-1.04/19 4}
Qb6 {+1.24/18 2.2} 9. c3 {-1.08/19 6} Qxb1+ {+1.28/18 4} 10. Rxb1
{-1.09/19 2.4} b6 {+1.32/18 1.9} 11. h4 {-1.10/19 6} g6 {+1.38/18 4} 12. g3
{-1.21/18 6} Bf5 {+1.58/18 2.0} 13. Bxf5 {-1.39/18 2.4} gxf5 {+1.58/16 0.2}
14. Kg2 {-1.55/18 4} Nd7 {+1.69/17 2.0} 15. Rb3 {-1.49/19 4} Nf6
{+1.86/17 2.4} 16. Nf3 {-1.75/16 2.3} Rg8 {+1.93/18 2.7} 17. Ne5
{-1.93/18 2.1} f4 {+2.03/18 2.1} 18. c4 {-1.97/18 2.6} dxc4 {+2.17/18 2.2}
19. Nxc4 {-2.13/19 2.9} fxg3 {+2.20/17 1.2} 20. fxg3 {-2.07/18 1.1} Rd8
{+2.23/18 2.5} 21. Rd1 {-2.14/19 6} Ne4 {+2.23/17 1.3} 22. Rbd3
{-2.15/17 0.8} f5 {+2.23/16 0.3} 23. d5 {-2.15/19 2.1} f4 {+2.54/17 2.8}
24. Kf3 {-2.31/18 3} Nc3 {+2.59/16 0.4} 25. Rg1 {-2.29/19 1.1} Nxd5
{+2.59/14 0.1} 26. g4 {-2.28/19 1.6} Bg7 {+2.65/15 0.4} 27. Bb2
{-2.20/18 2.5} Bxb2 {+2.68/16 1.8} 28. Nxb2 {-2.30/18 0.7} c5
{+2.67/16 0.4} 29. a4 {-2.28/19 2.9} Kf7 {+2.75/18 1.9} 30. Rgd1
{-2.37/19 0.9} e6 {+2.64/17 0.5} 31. Nc4 {-2.30/19 0.7} Ke7 {+2.55/18 4}
32. Ne5 {-2.00/17 0.9} Rd6 {+2.45/18 4} 33. g5 {-2.14/19 2.4} Rf8
{+2.37/18 4} 34. Nc4 {-2.11/19 1.0} Rdd8 {+2.30/18 2.4} 35. R3d2
{-2.14/19 1.3} a6 {+2.60/18 2.2} 36. Rb2 {-2.13/18 1.8} Nb4 {+2.68/18 1.8}
37. Rbd2 {-2.40/17 0.9} Rxd2 {+2.49/17 0.3} 38. Rxd2 {-2.61/18 0.7} Nd5
{+2.61/18 0.3} 39. Rb2 {-1.63/16 0.4} Rb8 {+2.61/16 0.1} 40. h5
{-1.80/18 1.4} b5 {+2.26/18 3} 41. Ne5 {-1.42/17 1.3} Kd6 {+2.63/19 2.2}
42. Nf7+ {-1.46/15 0.4} Kc7 {+2.63/17 0.1} 43. g6 {-2.39/18 3} hxg6
{+3.12/17 2.0} 44. hxg6 {-2.96/15 0.6} b4 {+3.12/16 0.1} 45. Rc2
{-3.28/16 1.1} b3 {+4.80/19 2.8} 46. Rxc5+ {-4.19/18 0.3} Kd7
{+4.80/18 0.1} 47. Rc1 {-5.10/20 0.7} b2 {+5.21/19 2.3} 48. Ne5+
{-5.97/21 2.1} Ke7 {+5.45/19 2.2} 49. Rb1 {-6.36/20 1.4} Rb3+
{+5.45/18 0.1} 50. Kg4 {-6.71/19 1.0} Nc3 {+7.56/17 2.2} 51. Rg1
{-6.94/19 0.8} b1=Q {+7.56/16 0.3} 52. Rxb1 {-7.03/17 0.2} Rxb1
{+7.56/14 0.1} 53. Kxf4 {-7.45/19 0.9} Kf6 {+7.74/16 1.0} 54. g7
{-7.56/17 0.8} Kxg7 {+8.06/16 1.6} 55. Nc6 {-11.41/20 0.9} Nxa4
{+8.12/16 1.4} 56. Kf3 {-250.00/24 0.4} Kf6 {+8.69/16 1.7} 57. Na5
{-250.00/28 1.2} Ke5 {+10.38/16 1.3} 58. Nc4+ {-11.52/15 0.3} Kd4
{+12.54/17 2.5} 59. Nd6 {-12.30/18 0.8} Nc3 {+12.54/15 0.1} 60. Kg2
{-250.00/19 1.1} a5 {+14.10/16 4} 61. Nc8 {-250.00/27 0.8} a4
{+16.31/13 0.7} 62. Ne7 {-250.00/26 0.4} a3 {+16.55/12 0.2} 63. Nc6+
{-250.00/30 0.7} Kd5 {+16.55/10 0.1} 64. Ne7+ {-250.00/31 0.6} Kd6
{+1000.11/18 1.3} 65. Nc8+ {-250.00/32 0.5} Ke5 {+1000.09/18 0.6} 66. Ne7
{-250.00/26 0.8} a2 {+1000.08/18 0.7} 67. Ng6+ {-250.00/27 2.9} Ke4
{+1000.06/18 0.6} 68. Kg3 {-1000.17/29 4} Rg1+ {+1000.05/16 0.3} 69. Kh4
{-1000.09/21 0.2} Rxg6 {+1000.04/17 0.5} 70. Kh3 {-1000.03/99 0.2} a1=Q
{+1000.03/16 0.7} 71. Kh2 {-1000.02/99 0.1} Qg1+ {+1000.02/15 0.5} 72. Kh3
{-1000.01/5 0.1} Qh1# {+1000.01/128 0.1}
{Black mates} 0-1
[/pgn]

[Laskos]

The next Komodo vs GM handicap match will again be against GM Sergey Erenburg (FIDE 2585, USCF 2677, born 1983), who lost 3.5 to 0.5 last weekend in normal chess playing standard 90' + 30" vs Komodo on one core only playing blitz (3' + 1"), with no ponder for komodo, no tablebases, no opening book past move 3, and White pieces every game to the GM.
So this time (July 15 and 16, noon and 5 pm Eastern time each day, games on chess.com) we are reverting to material handicaps, with our traditional 45' + 15" time control for each side and no restrictions on Komodo, which will use 24 cores. Since Komodo has yet to lose a single game at odds of the f7 pawn, even with two moves for White (Simon Williams did win with f7 and THREE moves), we are again going to try two pawn handicap as we have in three earlier matches. Komodo always plays White,removes the e2 pawn, and also the b2,c2,d2, and f2 pawns (once each, four games total). None of these handicaps have ever been tried in any game on record to the best of my knowledge.
Previous two pawn handicap results were a 4 to 1 loss to GM Petr Neuman (2455 FIDE) at c2 + f2, a 4.5 to 1.5 win against GM Martin Petr (2515) at a2 plus b2,d2, and e2 rotating, and a 1.5 to 0.5 win against GM Victor Mikhalevski (2534 FIDE) at c2 plus b2 once, d2 once. The Petr match showed that the edge pawns are pretty small handicaps for obvious reasons, but the overall score in the matches not involving the edge pawns is 2.5 for Komodo to 4.5 for GMs. Erenburg will be the highest rated GM we have ever given two pawns to, and these are non-edge pawns. In our favor is the fact that the handicap will vary each time, so no repeating successful openings. I think we are underdogs this time, what do all of you think?

IIRC GM Neuman with c2+f2 first drew several games until he learned how to use the advantage in order to massively exchange pieces and go into won endgames. The advantage is both material and positional, White has difficulty developing without these 2 pawns and goes to exchanges even with high value of Contempt. ELO-wise it's probably 700 ELO points handicap, but one can learn how to handle the opening to beat White.

Here we have a bit smaller 600-650 ELO points handicaps, but similarly important is that they are played only once each. In case GM doesn't home prepare overwhelming openings leading to massive exchanges, it will be tough for him to win, so the result would be many draws. Will Komodo have any opening book? Did "Contempt" improve since GM Neuman match? It's crucial that Komodo avoids exchanges and goes to complications.

My bet would be, if GM doesn't home cook overwhelming openings, 2/4 for Komodo.

[lkaufman]

In the previous 1 core game, I already expected 3.5 or 4 in favour of Komodo as I know how strong was the software even on my first smart phone HTC HD2 years ago. This result will not much depend on human rating but depend how he have prepared with this odd! I mean even 2200 player who analysed hundreds of games for this position will get a good positional advantage enough to win than some with 2500 who havent trained in this position.
Guess- 2.5-1.5 in favour of Komodo.

Since he will get $100 per draw and $300 per win (on top of base fee), and since he was clearly motivated to play this match by the prize money, I do expect him to prepare a fair amount. This didn't help any of the GMs who played at f7 and move or two moves or Exchange odds to win even a single game, but perhaps with two pawns it will be different. GM Lenderman shared his full analysis of the f7 handicap with me after the game; he clearly put some effort into preparing. I don't think a 2200 player would win any game at these handicaps no matter how much he prepared, but a nearly 2600 player should win some. The GMs often get big advantages out of the opening in these handicap games, but can't win them. But I think that winning with two extra pawns with just slightly less development is easier for a human than winning with one extra pawn and a great position. Most pawn plus endgames are just too hard for a human to win against Komodo, but with two extra pawns almost any endgame should be easy to win. In general, Komodo will be "happy" to reach an endgame only one pawn down which it can probably draw.

[Ozymandias]

Back to those odds already? Too bad.

[lkaufman]

The next Komodo vs GM handicap match will again be against GM Sergey Erenburg (FIDE 2585, USCF 2677, born 1983), who lost 3.5 to 0.5 last weekend in normal chess playing standard 90' + 30" vs Komodo on one core only playing blitz (3' + 1"), with no ponder for komodo, no tablebases, no opening book past move 3, and White pieces every game to the GM.
So this time (July 15 and 16, noon and 5 pm Eastern time each day, games on chess.com) we are reverting to material handicaps, with our traditional 45' + 15" time control for each side and no restrictions on Komodo, which will use 24 cores. Since Komodo has yet to lose a single game at odds of the f7 pawn, even with two moves for White (Simon Williams did win with f7 and THREE moves), we are again going to try two pawn handicap as we have in three earlier matches. Komodo always plays White,removes the e2 pawn, and also the b2,c2,d2, and f2 pawns (once each, four games total). None of these handicaps have ever been tried in any game on record to the best of my knowledge.
Previous two pawn handicap results were a 4 to 1 loss to GM Petr Neuman (2455 FIDE) at c2 + f2, a 4.5 to 1.5 win against GM Martin Petr (2515) at a2 plus b2,d2, and e2 rotating, and a 1.5 to 0.5 win against GM Victor Mikhalevski (2534 FIDE) at c2 plus b2 once, d2 once. The Petr match showed that the edge pawns are pretty small handicaps for obvious reasons, but the overall score in the matches not involving the edge pawns is 2.5 for Komodo to 4.5 for GMs. Erenburg will be the highest rated GM we have ever given two pawns to, and these are non-edge pawns. In our favor is the fact that the handicap will vary each time, so no repeating successful openings. I think we are underdogs this time, what do all of you think?

IIRC GM Neuman with c2+f2 first drew several games until he learned how to use the advantage in order to massively exchange pieces and go into won endgames. The advantage is both material and positional, White has difficulty developing without these 2 pawns and goes to exchanges even with high value of Contempt. ELO-wise it's probably 700 ELO points handicap, but one can learn how to handle the opening to beat White.

Here we have a bit smaller 600-650 ELO points handicaps, but similarly important is that they are played only once each. In case GM doesn't home prepare overwhelming openings leading to massive exchanges, it will be tough for him to win, so the result would be many draws. Will Komodo have any opening book? Did "Contempt" improve since GM Neuman match? It's crucial that Komodo avoids exchanges and goes to complications.

My bet would be, if GM doesn't home cook overwhelming openings, 2/4 for Komodo.

Contempt has improved a bit since Neuman match, and perhaps I have a better idea of how to set it now. I may use an opening book, but it will probably be only 3 or 4 moves deep and will mostly just serve to save time on these moves, unless I notice something special about any of the handicaps. Komodo will "want" to avoid exchanges, but whether that will be possible without excessive concessions isn't clear. I don't know how predictable Komodo will be; at the c2 + f2 handicap there seemed to be only one reasonable setup for White, so he could prepare easily.

[lkaufman]

Back to those odds already? Too bad.

I'm guessing you would have preferred another match like the last one but with larger time odds? I didn't like this so much because making Komodo play much faster than 3' + 1" seems pointless, since hardly anyone uses it with just one or two seconds thinking time I imagine. Or did you have another idea in mind?

[JJJ]

The next Komodo vs GM handicap match will again be against GM Sergey Erenburg (FIDE 2585, USCF 2677, born 1983), who lost 3.5 to 0.5 last weekend in normal chess playing standard 90' + 30" vs Komodo on one core only playing blitz (3' + 1"), with no ponder for komodo, no tablebases, no opening book past move 3, and White pieces every game to the GM.
So this time (July 15 and 16, noon and 5 pm Eastern time each day, games on chess.com) we are reverting to material handicaps, with our traditional 45' + 15" time control for each side and no restrictions on Komodo, which will use 24 cores. Since Komodo has yet to lose a single game at odds of the f7 pawn, even with two moves for White (Simon Williams did win with f7 and THREE moves), we are again going to try two pawn handicap as we have in three earlier matches. Komodo always plays White,removes the e2 pawn, and also the b2,c2,d2, and f2 pawns (once each, four games total). None of these handicaps have ever been tried in any game on record to the best of my knowledge.
Previous two pawn handicap results were a 4 to 1 loss to GM Petr Neuman (2455 FIDE) at c2 + f2, a 4.5 to 1.5 win against GM Martin Petr (2515) at a2 plus b2,d2, and e2 rotating, and a 1.5 to 0.5 win against GM Victor Mikhalevski (2534 FIDE) at c2 plus b2 once, d2 once. The Petr match showed that the edge pawns are pretty small handicaps for obvious reasons, but the overall score in the matches not involving the edge pawns is 2.5 for Komodo to 4.5 for GMs. Erenburg will be the highest rated GM we have ever given two pawns to, and these are non-edge pawns. In our favor is the fact that the handicap will vary each time, so no repeating successful openings. I think we are underdogs this time, what do all of you think?

IIRC GM Neuman with c2+f2 first drew several games until he learned how to use the advantage in order to massively exchange pieces and go into won endgames. The advantage is both material and positional, White has difficulty developing without these 2 pawns and goes to exchanges even with high value of Contempt. ELO-wise it's probably 700 ELO points handicap, but one can learn how to handle the opening to beat White.

Here we have a bit smaller 600-650 ELO points handicaps, but similarly important is that they are played only once each. In case GM doesn't home prepare overwhelming openings leading to massive exchanges, it will be tough for him to win, so the result would be many draws. Will Komodo have any opening book? Did "Contempt" improve since GM Neuman match? It's crucial that Komodo avoids exchanges and goes to complications.

My bet would be, if GM doesn't home cook overwhelming openings, 2/4 for Komodo.

I think the same, but I think Komodo is now 50 elo better, that's matter a lot to me and will play better with contempt also and with full power. Komodo is not underdog to me, the human won't be well prepared as usual, so I think Komodo will win this 2,5/4.

[Ozymandias]

Back to those odds already? Too bad.

I'm guessing you would have preferred another match like the last one but with larger time odds? I didn't like this so much because making Komodo play much faster than 3' + 1" seems pointless, since hardly anyone uses it with just one or two seconds thinking time I imagine. Or did you have another idea in mind?

A faster TC for Komodo, a longer TC for the GM or a stronger human opponent. Any of those variations, or a viable combination, would give us a more even match.

There's also the option of using a popular phone, instead of a laptop.