First knight odds match with IM Saturday

[lkaufman]

Game 4. Komodo played the wing gambit deferred, which seemed sensible for a knight odds game but didn't work out very well. Ray traded in the extra knight to reach a three pawns up endgame, but Komodo won one of them and had enough pressure to reach a rare TB-drawn rook vs rook and two pawns endgame.

[pgn][Event "Live Chess - Knight Odds"]
[Site "Chess.com"]
[Date "2020.Jan.4"]
[Round "4"]
[White "PlayKomodo"]
[Black "Mimsychess (IM Raymond Kaufman)"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[TimeControl "900+10"]

1. Nf3 c5 2. e4 d6 3. b4 cxb4 4. a3 bxa3 5. d4 g6 6. Rxa3 Bg7 7. Bb5+ Bd7 8. Qe2
Bxb5 9. Qxb5+ Qd7 10. Qe2 Nc6 11. Bb2 Nf6 12. O-O O-O 13. c4 Nh5 14. g3 Qg4 15.
Bc1 f5 16. exf5 Rxf5 17. Qe6+ Kh8 18. h3 Qxh3 19. Ng5 Nd8 20. Nxh3 Nxe6 21. g4
Rf7 22. gxh5 Bxd4 23. Re1 Nc5 24. Rd1 e5 25. Ng5 Rf5 26. Nf3 Ne6 27. Nxd4 Nxd4
28. h6 Raf8 29. Be3 Nc2 30. Rxa7 Nxe3 31. fxe3 Rg5+ 32. Kh2 Rh5+ 33. Kg3 Rxh6
34. Rxb7 g5 35. Rd7 Rc8 36. Ra1 Rg6 37. Raa7 h5 38. Rh7+ Kg8 39. Rxh5 Rxc4 40.
Rhh7 g4 41. Rhd7 Rc8 42. Ra6 Rf8 43. Raxd6 Rf3+ 44. Kg2 Rxd6 45. Rxd6 Kf7 46.
Ra6 Rxe3 0-1[/pgn]

Komodo headed deliberately for this two pawns down rook endgame knowing it was a tablebase draw, then upon reaching it inexplicably failed to move and timed out.

[AdminX]

For some reason game one and three really stood out for me. Game one because of how aggressive an the amount of contempt it how for it's human opponent while on the attack. Game three also for it's aggressive nature and the piece mobility. The very active nature of white's two bishops and rook on the open file was really something to watch. I was surprised that your son was able to hold on and reach a draw. He really played well for that time control.

Thanks for the games and the commentary.

[AdminX]

Game one, that moment that is still etched in my brain. White to move, and I just loved that in for a penny in for a pound attitude that it seemed to display. Click image for larger view.

[duncan]

For the first time ever, an engine (Komodo, probably the MCTS nn dev. version) will play a rapid (15' + 10") knight odds match on chess.com with an I.M., namely my son Raymond Kaufman, this Saturday Jan 4 at 4 pm Eastern time. Four games are planned, alternating g1/b1 knight odds. Look for games of "playkomodo". We'll post games here whenever time permits. As some of you may recall, 77 year old FM Larry Gilden lost 6 to zero under these conditions, but with 30' + 10" scored 1.5 to 3.5. But Ray is a full class stronger and less than half of Gilden's age, so I would expect a close match. While Ray is a rather low rated IM, he tied for first in the Canadian Open in 2018 so he's a strong player, and also rather good at blitz and rapid. If Komodo wins the match we can make a case for a higher-profile knight odds rapid match against a more prominent IM or even a GM. If not we'll have to improve Komodo a bit before trying that.

Komodo played very well. Thanks for doing this. Any thoughts/ conclusions on the knight handicap for Komodo and what lies ahead?

[duncan]

Hoping to see some live broadcast with commentary (I know I sound like a broken record, but I expect that one day the platform where one of such matches will be broadcast will be announced.)

Chess.com is only interested in providing live commentary for matches with "big names", which cost real money (we do give small prizes for wins and draws (triple for wins) in all matches). Hopefully we'll have such a match before too long. Saturday's match can be regarded as preparation for such matches, in terms of determining fair time limits for knight odds, finding flaws in play at knight odds or in the online play, etc. We've already pretty much proven that we can give two pawns (as White) or f7 and two moves or knight for f7 pawn or the equivalents to strong GMs in rapid play, but in the opening a knight is worth nearly four pawns so we need to establish that giving knight odds to a GM is not ridiculous, so an IM is the obvious step. If knight odds proves unreasonable for GMs, there is still f7 and three moves as an in-between step, which we tried twice at slow rapid, losing to GM Simon Williams but beating IM Danny Rensch. But beating a GM in a knight odds match is the "holy grail" of chess, the ultimate challenge.

What timing are you hoping to beat a GM at knight odds ?

[Nordlandia]

Ted: i wish we could turn the knights 360°

[AdminX]

Ted: i wish we could turn the knights 360°

Yeah, I hope they (Chessbase) will fix this.

[lkaufman]

For the first time ever, an engine (Komodo, probably the MCTS nn dev. version) will play a rapid (15' + 10") knight odds match on chess.com with an I.M., namely my son Raymond Kaufman, this Saturday Jan 4 at 4 pm Eastern time. Four games are planned, alternating g1/b1 knight odds. Look for games of "playkomodo". We'll post games here whenever time permits. As some of you may recall, 77 year old FM Larry Gilden lost 6 to zero under these conditions, but with 30' + 10" scored 1.5 to 3.5. But Ray is a full class stronger and less than half of Gilden's age, so I would expect a close match. While Ray is a rather low rated IM, he tied for first in the Canadian Open in 2018 so he's a strong player, and also rather good at blitz and rapid. If Komodo wins the match we can make a case for a higher-profile knight odds rapid match against a more prominent IM or even a GM. If not we'll have to improve Komodo a bit before trying that.

Komodo played very well. Thanks for doing this. Any thoughts/ conclusions on the knight handicap for Komodo and what lies ahead?

It looks like in general knight handicap is the suitable one for such matches, although both the rating of the opponent and the exact time limit are factors. I was pleased to see that Komodo did well with both knight handicaps; previously I was worried that b1 off is too large a handicap because 1.e4 is dubious due to 1...d5!, but it seems that White can play 1.Nf3 (or 1.d4) and still get enough action to pose problems. At the time limit of this match, I estimate we would be evenly matched at knight odds against a strong IM (Danny Rensch perhaps) or a weak GM (perhaps Simon Williams would like a second match?), but to claim we can give knight odds to a GM it should really be someone over 2500 FIDE so that it is clear that he would still qualify as a GM based on current level of play, and I don't think we're quite there yet. In blitz we could surely give knight odds to fairly strong GMs. At fast blitz (3' + 1") we might even be able to do it with Elite GMs, probably not with Nakamura (or Carlsen) although I'd be willing to try. At standard tournament time limits (40/90' + 30' + 30") a fair opponent is probably not more than 2200 FIDE, but this isn't practical for a public event (I could set it up for an informal online match on request from a suitable interested player). No definite plans yet, but I would expect that knight odds will be the prime handicap in future Komodo vs human matches, with time limits suitable for the opponent in question. We may also try knight odds in FRC (chess960); we beat FM John Meyer 4 to 0 that way a couple years ago at 45' + 15", but his strength was just about 2100 FIDE so we need to try it vs. a much stronger opponent.

[Paloma]

Ted: i wish we could turn the knights 360°

360° is the same as now.

90 or 270° are the best!

[Uri Blass]

For the first time ever, an engine (Komodo, probably the MCTS nn dev. version) will play a rapid (15' + 10") knight odds match on chess.com with an I.M., namely my son Raymond Kaufman, this Saturday Jan 4 at 4 pm Eastern time. Four games are planned, alternating g1/b1 knight odds. Look for games of "playkomodo". We'll post games here whenever time permits. As some of you may recall, 77 year old FM Larry Gilden lost 6 to zero under these conditions, but with 30' + 10" scored 1.5 to 3.5. But Ray is a full class stronger and less than half of Gilden's age, so I would expect a close match. While Ray is a rather low rated IM, he tied for first in the Canadian Open in 2018 so he's a strong player, and also rather good at blitz and rapid. If Komodo wins the match we can make a case for a higher-profile knight odds rapid match against a more prominent IM or even a GM. If not we'll have to improve Komodo a bit before trying that.

Komodo played very well. Thanks for doing this. Any thoughts/ conclusions on the knight handicap for Komodo and what lies ahead?

It looks like in general knight handicap is the suitable one for such matches, although both the rating of the opponent and the exact time limit are factors. I was pleased to see that Komodo did well with both knight handicaps; previously I was worried that b1 off is too large a handicap because 1.e4 is dubious due to 1...d5!, but it seems that White can play 1.Nf3 (or 1.d4) and still get enough action to pose problems. At the time limit of this match, I estimate we would be evenly matched at knight odds against a strong IM (Danny Rensch perhaps) or a weak GM (perhaps Simon Williams would like a second match?), but to claim we can give knight odds to a GM it should really be someone over 2500 FIDE so that it is clear that he would still qualify as a GM based on current level of play, and I don't think we're quite there yet. In blitz we could surely give knight odds to fairly strong GMs. At fast blitz (3' + 1") we might even be able to do it with Elite GMs, probably not with Nakamura (or Carlsen) although I'd be willing to try. At standard tournament time limits (40/90' + 30' + 30") a fair opponent is probably not more than 2200 FIDE, but this isn't practical for a public event (I could set it up for an informal online match on request from a suitable interested player). No definite plans yet, but I would expect that knight odds will be the prime handicap in future Komodo vs human matches, with time limits suitable for the opponent in question. We may also try knight odds in FRC (chess960); we beat FM John Meyer 4 to 0 that way a couple years ago at 45' + 15", but his strength was just about 2100 FIDE so we need to try it vs. a much stronger opponent.

I would like to see komodo without a knight play for fide rating at standard tournament time limit.
Fide only need to change the rules to allow computers in these conditions.

There are swiss system tournaments that include different levels when it is clear that komodo is not going to get 100% or 0%