What will Komodo score giving two pawns to GM?

[APassionForCriminalJustic]

Easy draw for the GM but very difficult to win.

You think that this will be an easy draw for a GM? I don't think so man... this isn't Magnus that he will be playing here.

[whereagles]

high hopes for the human contender, after Victor's 4/2 winning performance

[Jesse Gersenson]

I just played a game against Larry's 24-core machine at f2/c2 odds to see the broadcast was working. In a 2 minuet blitz game I lasted 40 moves. Komodo equalized at move 15 and was better by move 19 and the game was all but over by move 23.

[Vinvin]

I just played a game against Larry's 24-core machine at f2/c2 odds to see the broadcast was working. In a 2 minuet blitz game I lasted 40 moves. Komodo equalized at move 15 and was better by move 19 and the game was all but over by move 23.

And what's your rating ?

[Vinvin]

I played 3 blitz (5m+3s) against Amadan (a quite strong player in blitz/bullet, 2150 FIDE) on Lichess : 3-0
I set Komodo to 2min+2sec, manual operation to copy moves.
Here are the games :

[pgn]
[Event "Casual game"]
[Site "http://lichess.org/axDKCqe6"]
[Date "2015.08.01"]
[White "Vinvin"]
[Black "AMADAN"]
[Result "1-0"]
[WhiteElo "1986"]
[BlackElo "2282"]
[PlyCount "71"]
[Variant "From Position"]
[TimeControl "300+3"]
[ECO "?"]
[Opening "?"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PP1PP1PP/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 d5 2. Bf4 Nf6 3. Nf3 e6 4. e3 c5 5. Nc3 a6 6. Be2 Nc6 7. Rc1 cxd4 8. Nxd4 Bd6 9. Nxc6 bxc6 10. O-O O-O 11. Na4 Bxf4 12. Rxf4 e5 13. Rf1 Qd6 14. Rc2 Be6 15. Qd2 Ne4 16. Qc1 c5 17. Nxc5 Nxc5 18. Rxc5 Rfc8 19. b4 a5 20. a3 axb4 21. axb4 d4 22. Rd1 Rxc5 23. bxc5 Qd5 24. Bf3 e4 25. Rxd4 Qa2 26. Bxe4 Rc8 27. Qc3 g6 28. h3 Qa7 29. Rd6 Qc7 30. Qe5 Qe7 31. Bc6 Rd8 32. Bd5 Rxd6 33. cxd6 Qd7 34. Bxe6 fxe6 35. Qc5 Kf7 36. e4 { Black resigns } 1-0

[Event "Casual game"]
[Site "http://lichess.org/Wx4gNdgO"]
[Date "2015.08.01"]
[White "Vinvin"]
[Black "AMADAN"]
[Result "1-0"]
[WhiteElo "1986"]
[BlackElo "2282"]
[PlyCount "49"]
[Variant "From Position"]
[TimeControl "300+3"]
[ECO "?"]
[Opening "?"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PP1PP1PP/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. Nf3 Nf6 2. e3 g6 3. b4 Bg7 4. Bb2 O-O 5. Be2 d6 6. Nc3 e5 7. O-O Nbd7 8. Qc2 c6 9. Rab1 d5 10. Na4 Re8 11. Rbc1 e4 12. Nd4 Ne5 13. b5 Bd7 14. bxc6 Nxc6 15. Nc5 Nxd4 16. Bxd4 Bc6 17. Qb2 Nh5 18. Bxh5 Bxd4 19. Qxd4 gxh5 20. Rf6 Rc8 21. Rcf1 Qe7 22. R1f5 b6 23. Rg5+ Kf8 24. Ne6+ Qxe6 25. Rxe6 { Black resigns } 1-0

[Event "Casual game"]
[Site "http://lichess.org/5mlCSHTE"]
[Date "2015.08.01"]
[White "Vinvin"]
[Black "AMADAN"]
[Result "1-0"]
[WhiteElo "1986"]
[BlackElo "2282"]
[PlyCount "41"]
[Variant "From Position"]
[TimeControl "300+3"]
[ECO "?"]
[Opening "?"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PP1PP1PP/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 g6 2. Bf4 Bg7 3. Nf3 c5 4. e3 cxd4 5. exd4 Qb6 6. Nc3 d6 7. Bb5+ Bd7 8. Nd5 Qa5+ 9. b4 Qd8 10. Rc1 Nc6 11. Ba4 Nf6 12. Nxf6+ Bxf6 13. b5 Nb8 14. O-O O-O 15. Bh6 Bg7 16. Bxg7 Kxg7 17. Ng5 h6 18. Rxf7+ Rxf7 19. Nxf7 Kxf7 20. Qf3+ Kg7 21. Bb3 { Black resigns } 1-0
[/pgn]

[Laskos]

My guess is it will be close, with many draws, say 3-4 draws out of 6. In self-play, Komodo needs 1/100 time odds to make for handicap at this time control on one core, meaning 600 ELO points. The estimated value of the handicap increases first in self play, second, with time control, from 400 points at ultra-fast to 600 points at 45'+15''. If Komodo on a strong hardware is reasonably assumed to be 3200 FIDE ELO rating, that would mean a Komodo performance with c2 and f2 pawns handicap to be around 2600 FIDE ELO, a bit higher than the opponent's 2500. I would choose a range 2.5-3.5 points for Komodo in this match, but I voted for 3.5 because it seems I am the only one predicting that score .

[lkaufman]

My guess is it will be close, with many draws, say 3-4 draws out of 6. In self-play, Komodo needs 1/100 time odds to make for handicap at this time control on one core, meaning 600 ELO points. The estimated value of the handicap increases first in self play, second, with time control, from 400 points at ultra-fast to 600 points at 45'+15''. If Komodo on a strong hardware is reasonably assumed to be 3200 FIDE ELO rating, that would mean a Komodo performance with c2 and f2 pawns handicap to be around 2600 FIDE ELO, a bit higher than the opponent's 2500. I would choose a range 2.5-3.5 points for Komodo in this match, but I voted for 3.5 because it seems I am the only one predicting that score .

I played 100 games between latest Komodo versions with one giving the other the two pawn handicap. Four threads, 2' + 1". White won one game, six draws, 93 losses. +552 elo for Black. This would of course increase with longer time control. So my estimate of the rating value at this tc is at least 700 elo without even considering that White can modify his strategy for the handicap. Not too far from your estimate. I'm a bit concerned that it seems that there is pretty much only one reasonable way for White to play the opening, and therefore he will be able to prepare effectively and/or learn from each game. Because of this, I'll be happy with a drawn match.

[duncan]

My guess is it will be close, with many draws, say 3-4 draws out of 6. In self-play, Komodo needs 1/100 time odds to make for handicap at this time control on one core, meaning 600 ELO points. The estimated value of the handicap increases first in self play, second, with time control, from 400 points at ultra-fast to 600 points at 45'+15''. If Komodo on a strong hardware is reasonably assumed to be 3200 FIDE ELO rating, that would mean a Komodo performance with c2 and f2 pawns handicap to be around 2600 FIDE ELO, a bit higher than the opponent's 2500. I would choose a range 2.5-3.5 points for Komodo in this match, but I voted for 3.5 because it seems I am the only one predicting that score .

anything you would lime to add now that the results are in.?

[Laskos]

My guess is it will be close, with many draws, say 3-4 draws out of 6. In self-play, Komodo needs 1/100 time odds to make for handicap at this time control on one core, meaning 600 ELO points. The estimated value of the handicap increases first in self play, second, with time control, from 400 points at ultra-fast to 600 points at 45'+15''. If Komodo on a strong hardware is reasonably assumed to be 3200 FIDE ELO rating, that would mean a Komodo performance with c2 and f2 pawns handicap to be around 2600 FIDE ELO, a bit higher than the opponent's 2500. I would choose a range 2.5-3.5 points for Komodo in this match, but I voted for 3.5 because it seems I am the only one predicting that score .

anything you would lime to add now that the results are in.?

That an opening book was necessary and that the black advantage of c2 and f2 handicap is too much positional too, besides material. The computer in self-play doesn't learn an overwhelming opening with forced exchanges. My guess is that varied 2 pawns handicap, not c2 and f2, with an opening book, can stand against a 2500 ELO human in 6 games. But I might well be wrong again, who knows what these GMs are up to .

[Jesse Gersenson]

...My guess is that varied 2 pawns handicap, not c2 and f2, with an opening book, can stand against a 2500 ELO human in 6 games. But I might well be wrong again, who knows what these GMs are up to.

Every GM I talked with about a 2-pawn handicap said, after a few seconds thought, "Two pawns, that's a lot." They didn't need clarification about which two pawns to make their assessment. They didn't care what program, or hardware, they'd be playing against. They didn't want a copy of the program to practice against.

The final match score at two pawns was 1 out of 5. So 90% of you miscalculated something every GM agreed upon, two pawns is a lot.

For an interesting match we should defer to GM's to figure out interesting odds, not engine fans.