Computer Chess / Handicap Chess
with Robert Hess (FIDE 2580). This handicap gives White a real attack, and is roughly a two pawn handicap. Black really has to grovel in the opening and may have trouble castling. “Pawn and three moves” was also popular long ago, where White gets to move first after already playing 1. e4 and 2. d4 and removing f7. It’s about a 2½ pawn handicap. Williams beat Komodo once with this handicap. The next step above this is normally knight odds, roughly a 3½ pawn handicap. Another handicap of roughly comparable
magnitude to pawn and move that Staunton gave in many games is the Exchange (a1-rook for b8- knight). We have also used this in the matches against Movsesian, Lenderman, and Nakamura with a slight twist: The black rook starts on b8 instead of on a8. The reason for this is that since White cannot castle queenside without his rook, it seems unfair to let Black have that option. This handicap is quite different from pawn and move, because while the f7 handicap gives White an automatic attack or at least initiative, the Exchange handicap does not give any initiative. On the other hand, the Exchange is more material than a pawn, so the recipient needs only to swap down to an ending to win. But apparently it’s not so easy to do, as in five games against these very strong grandmasters, Komodo won three and drew two, including a draw against superstar Nakamura. I think the Exchange is at least as big a handicap as the “f” pawn, but the first move matters a lot in handicap chess. A handicap of two pawns was never given
in the classical period, although Kasparov introduced it. The question is of course which pawns to give. If one of the pawns is the a2- pawn, it’s not a huge handicap, as Komodo demonstrated by beating GM Martin Petr (2520 FIDE) with three wins and three draws rotating between a2, b2 and a2, d2 and a2, e2. But if we have to pick pawns that favor the receiver, it’s pretty tough; Komodo gave up three losses and two draws to GM Petr Neuman (2455 FIDE) giving c2 and f2. Komodo just had no active play without these pawns, and almost all endgames with two extra pawns are easy wins. Komodo did manage to win one game giving f7 and g7 pawns (!) to FM Larry Gilden (USCF 2220),when after 1. e4 White threatens mate and Black can’t even play 1. ... Nf6 due to 2. e5. Mikhalevski drew at c2 + d2 but lost at b2 + c2. Komodo has played a total of 10 games giving
knight odds to FIDE masters, scoring just 25 percent (one win, three draws). However surprisingly Komodo defeated FM John Meyer by 31
2⁄ to 1 2⁄ at knight for pawn odds (c7, d7, e7,
and f7). A knight is worth about four pawns in the opening, so it’s not surprising that it is just too much to give to a master. Whether any future engine will ever be able to win a knight odds match from a FIDE master is open to
32 April 2016 | Chess Life
debate. Maybe a good definition of a master is a player who can win a serious knight odds match from anyone or anything, updating Lasker’s rule. Of course in blitz it’s another story entirely; at 3’ and 2” Komodo came out ahead 41
⁄2 to 31 ⁄2 against GM Lenderman and beat
Mikhalevski by 7 to 6 at knight odds. Based on these matches, I would estimate
that knight odds is about right at our standard time control for a player around the 2000 USCF level, knight for pawn for 2300, and two pawns for 2400-2700 depending on the pawns chosen. Pawn and two moves seems about right for US Chess 2700, Exchange and move for 2750, f7 and move for 2800, Exchange for 2850, and f2 for 2900 (i.e. Hikaru). Based on handicap games between engines, adding the estimated handicap Elo values to the performance ratings of Komodo gives something in the ballpark of 3300 FIDE or 3400 US Chess for Komodo, pretty close to ratings showing on engine versus engine rating lists. Now let’s take a look at the Komodo versus
Nakamura match, featuring four different handicaps, which Komodo won by 21
2⁄ to 11 2⁄ . All
the games were pretty interesting. Here are three, omitting just the drawn f2 handicap game.
REMOVE F7-PAWN GM Hikaru Nakamura (2787) Komodo
Handicap match,
Chess.com (1), 01.06.2016
1. e4 e6 Perhaps the best choice at f7 handicap.
2. d4 d5 3. exd5 This option gets the queens off but forfeits
any initiative or attack. If White hopes to win I think this is a poor choice. Both 3. Nc3 and 3. Nd2 are objectively better here.
3. ... exd5 4. Qh5+ g6 5. Qe5+ Qe7 6. Qxe7+ Nxe7 7. Bf4 Bg7 8. c3 c5!
I think this surprised Nakamura. It led to a
very exciting game despite the early queen trade. 9. dxc5 Nd7 10. Bd6 Safer was 10. Nd2 Nxc5 11. Ngf3, returning
the second pawn to keep pace in development. 10. ... Nf5
(see diagram top of next column) 11. Bb5?
The continuation 11. Na3 a6 12. Nc2 keeps a safe advantage.
11. ... Nxd6 12. cxd6 a6 13. Ba4 0-0 14. Bb3 Nc5! 15. Bxd5+ Kh8 So Black has given up two more pawns (besides
24. Nxg5!? The continuation 24. Nb3 g4 25. Ng5 Rxd6
26. Nxe6 Rxe6 27. Bd3 g3 28. f3 Rd8 29. Be4 Bb2 30. Re1 Bc3 31. Rc1 with draw by repetition was likely the best line, but Nakamura’s decision to sacrifice a piece for some pawns was practical; he probably had confidence that he could hold the draw.
24. ... Bxd2 25. Nxe6 Bxc1 26. Nxf8 Rxf8 27. Bc6 Nc5 28. Rh3 Bd2 29. d7 Ba5 30. Re3 Rf6 31. Re8+ Kg7 32. Bf3 Nxd7 33. Re7+ Rf7 34. h6+ Kf8 35. Rxf7+ Kxf7 36. Be4 Bd2 37. Bxh7 Bxh6 38. g3 a5 39. f4 Nc5 40. Ke2 Bg7 41. Bd3 Nxd3?!
This was rightly criticized as making the
draw too easy, but probably White would hold anyway. Computers assume the opponent will see what they see, a poor assumption for a handicap game.
42. Kxd3 Bb2 43. Kc4 Bf6 44. Kb5 Bc3 45. a4 Kf6 46. Ka6 Kf5 47. Kb5 Bb4 48. Ka6 Ke4 49. Kb5 Be1 50. Ka6 Kf5 51. Kb5 Bc3 52. Kc4 Bb4 53. Kb5 Ke4 54. Ka6 Kf5, Draw agreed.
Although Komodo claims a 3/4 pawn plus
based on being up bishop for two pawns, it can do nothing without giving up its last pawn. Engines are still not good at recognizing “fortress” draws like this one, although this game motivated us to work on this problem. Nakamura made only one clear mistake this game, and yet had to play well to draw.
the handicap) for a big initiative/attack. So much for computers being materialistic!
16. Nf3 Nd3+ 17. Kf1 Bf5 18. Nbd2 Rad8 19. Bxb7 Nxb2 20. Be4 Be6 21. h4 Bxc3 22. Rc1 Na4 23. h5 g5
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76