The Dzindzi – Rybka 3 Handicap Match

Dzindzi – Rybka 3 Handicap Match

Report by IM Larry Kaufman

Chess with material handicaps has a long history, with many such games by Philidor, Staunton, Morphy, and Steinitz to be found in the databases. Particularly popular was “pawn and move”, meaning Black plays without his f7 pawn. It is quite a substantial handicap, but not so large that a strong player can win just by just trading pieces, as he could with knight odds.

When I joined the Rybka team at the start of 2007, I made it a goal to see Rybka reach the level where she could successfully give pawn and move to a grandmaster in a serious match. That goal has now been achieved.

We worked our way up to this step by step. First came a match with GM Jaan Ehlvest in March 2007, in which Rybka gave each of the eight White pawns in turn; Rybka won 5.5-2.5. Then we increased the handicap to each of the eight pawns but alternating colors against GM Joel Benjamin, again winning the match (4.5-3.5 officially, 5-3 counting a replay of a game we had to forfeit due to setting an invalid option).

Then came a match with GM Roman Dzindzichashvili (FIDE 2548, but at one time among the world’s top ten and still extremely strong in blitz and rapid chess), where we took Black in each of the eight games, with a 4-4 tie score. The Benjamin match time limit was 60’+30”, the other two were 45’+10”.

GM Roman Dzindzichashvili

We agreed to play a tiebreak match of four games, but because in the interim I got an octal computer (the other matches were played on a quad) I decided that we could increase the handicap to the traditional “pawn and move”, meaning f7. This is a much larger handicap than the other pawn handicaps (except perhaps g7) as it does not aide development, weakens the king, and makes it difficult for Black to develop and castle without making further concessions.

We tried a match this way at a faster time limit (25’+10”), but unfortunately Roman became very ill that day and lost three games without a fight. We decided that this was unfair to Roman, so we scheduled a new match after he had regained his health and won the South Carolina Open. This time the time limit was a more serious 30’+20”, and Roman was in good health.

As it happens, the match took place on July 28, the final day before Rybka 3 was “set in stone”. So I believe that the version we used was indeed “Rykba 3” in final form. In order to provide variety in the openings and to avoid possible opening traps that can easily give Black a hopeless game at this handicap, I prepared a small opening book (itself heavily based on Rybka analysis) for this (and other) handicaps, which is included with the Rybka 3 package.

The match was broadcast on Playchess.com. It took place at my home in Potomac, Maryland USA (near Washington D.C.). “Dzindzi” received a small fee plus $50 per point scored, so he was playing not only for pride. A “contempt” factor of 60 (60% of a pawn) was set so that Dzindzi could not bail out with a draw if he lost the bulk of his initial advantage.

The first surprise happened with Dzindzi’s opening move, 1.d4?!. This is clearly inferior to 1.e4 at this handicap, because the option of Qh5+ severly limits Black’s options after 1.e4. Rybka 3 scores 1.e4 as +1.17 after 18 ply, while 1d4 is “only” scored as +0.91. Therefore my tiny opening book was 90% devoted to 1.e4; I figured Rybka could manage other openings with minimal help. I guess Roman chose 1.d4 (which he stuck with for the whole match) partly to “kill” my book, partly because he prefers it in normal chess, and partly because it avoids playing into Rybka’s strength in open games.

In game 1 Rybka played the Queen’s Indian, reaching a playable position (+1.12 after 11 moves). After some inaccurate moves White’s edge was down to about half a pawn after 16 moves, when on move 21 White had to choose between staying in the middlegame with 21.bxc3 (which would keep the half pawn edge) or heading for a favorable ending with rook and pawn for two knights. He chose the endgame. Perhaps he could have kept an edge, but Roman allowed a forced repetition.

Dzindzichashvili,Roman (2548) - Rybka 3 (3200)
Dzindzi vs Rybka 3 handicap match (1), 28.07.2008
1.d4 Nf6 2.c4 e6 3.Nf3 b6 4.g3 Bb7 5.Bg2 Bb4+ 6.Bd2 Qe7 7.0-0 0-0 8.Bf4 Nc6 9.a3 Bd6 10.e3 Rab8 11.Nc3 Na5 12.Nd2 Bxf4 13.gxf4 Bxg2 14.Kxg2 c5 15.d5 Qf7 16.dxe6 dxe6 17.Qe2 Qg6+ 18.Kh1 Rbd8 19.Nf3 Ne4 20.Ne5 Nxc3 21.Nxg6 Nxe2 22.Nxf8 Kxf8 23.Rad1 Rxd1 24.Rxd1 Nb3 25.Rd7 Na5 26.Rxa7 Nxc4 27.b4 cxb4 28.axb4 Nc3 29.Rc7 b5 30.Rc5 Ke7 31.Kg2 h6 32.h4 Kd6 33.Kg1 Ne2+ 34.Kf1 Nc3 35.Kg1 Kd7 36.Kg2 Kd6 37.Kg1 Ne2+ draw. [Click to replay]

In game two I “chose” (via greenlighting book moves between the games) the Benoni; I had decided to avoid repeat openings for maximum spectator interest and to learn more myself. Rybka’s eval is decent after nine moves (+1.00), but this time Roman played quite well reaching +1.36 after 16 moves, when he chose 17.b3?! rather than the superior Qd3 planning a3 and b4. Still, his plan of kingside attack was also promising as 21.Qb2! would have kept a large plus. As the game went, he returned his handicap pawn to retain a huge advantage in space and some attack. His last chances for some edge were 26.Bg5! and later 31.Bf4. After that time pressure took its toll and in a bad but not hopeless position he made a simple blunder and resigned.

Dzindzichashvili,Roman (2548) - Rybka 3 (3200)
Dzindzi vs Rybka 3 handicap match (2), 28.07.2008
1.d4 Nf6 2.c4 c5 3.d5 g6 4.Nc3 Bg7 5.e4 0-0 6.Be2 d6 7.Nf3 Nh5 8.g3 Nf6 9.h3 Qb6 10.Rb1 Nbd7 11.Be3 Qa5 12.Nd2 Ne8 13.Qc2 Nb6 14.g4 Bd7 15.0-0 Na4 16.Nxa4 Bxa4 17.b3 Bd7 18.Kg2 Nc7 19.f4 Na6 20.e5 Nb4 21.Qe4 Qxa2 22.e6 Be8 23.f5 Rc8 24.Rf2 Kh8 25.Rbf1 Qb2 26.Nf3 gxf5 27.gxf5 Bf6 28.Kh1 Rg8 29.Nh2 Qc2 30.Qxc2 Nxc2 31.Bc1 h5 32.Bxh5 Bxh5 33.Rxc2 Rg3 34.Bb2 Bxb2 35.Rxb2 Rcg8 36.f6 exf6 37.Rbb1 Kg7 38.Rg1 Kf8 39.Rxg3 Rxg3 40.Nf1 Bf3+ 0-1. [Click to replay]

For game three I “chose” the Semi-Slav, a very poor decision. I felt I should try a Queen’s Gambit declined at least once, and chose the Semi-Slav because the move…c6 at least insured that Rybka would not block the c-pawn by an early …Nc6, which the default version (but not the human one) still may do sometimes. However the cure was worse than the disease, as simple moves led to a position where Black was not only a pawn down, but had a seriously weak isolated e-pawn and a bishop that could not develop normally to b7 due to this weakness. After a dozen moves White’s advantage was up to 1.40, and soon reached a decisive 1.60. White had everything: better pawns, better knight, and much better bishop, plus his extra pawn. I suppose Ivanchuk would resign here as Black. Roman played flawlessly and his advantage climbed to two pawns. Rather than try to win quickly, he chose to head to the endgame, still retaining his extra pawn and huge positional edge. He had a sure win by 52.h5, but he waited one move too long, and as played Rybka found a way to “only” lose the exchange while blocking up the position. After this he had to sacrifice a pawn to make progress, and getting short of time and not seeing how to proceed he chose to make waiting moves and offered a draw, which we accepted. I had Rybka play out the final position at 3 seconds per move and White won, but I think it was too difficult for a human in time pressure to try for the win.

Rybka author IM Vasik Rajlich and senior advisor IM Larry Kaufman (2007 in Mexico)

Dzindzichashvili,Roman (2548) - Rybka 3 (3200)
Dzindzi vs Rybka 3 handicap match (3), 28.07.2008
1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.Nc3 c6 5.e3 Nbd7 6.Bd3 Bd6 7.e4 dxe4 8.Nxe4 Nxe4 9.Bxe4 Nf6 10.Bc2 0-0 11.0-0 Qc7 12.c5 Be7 13.Bg5 Nd5 14.Bxe7 Qxe7 15.Ne5 b6 16.Qd3 g6 17.Qa3 Bb7 18.g3 Rad8 19.Rad1 Qc7 20.Be4 b5 21.f4 Nf6 22.Bg2 Nd5 23.Rfe1 a5 24.h4 Bc8 25.Qd3 Kh8 26.Qe2 Rf6 27.Ng4 Rf5 28.Nh6 Rh5 29.Ng4 Rf5 30.Be4 Rf7 31.Bb1 Nf6 32.Qe5 Qxe5 33.Nxe5 Rc7 34.Be4 Nxe4 35.Rxe4 Kg7 36.Kf2 a4 37.Re2 Kf6 38.Ke3 Bd7 39.g4 Be8 40.Rh2 a3 41.b3 Rg7 42.Ke4 Ke7 43.Rdd2 Kf6 44.Rdg2 Ra8 45.Re2 Ke7 46.Rh3 Kf8 47.Reh2 Kg8 48.g5 Kf8 49.Ng4 Ke7 50.Ke5 b4 51.Nf6 Rd8 52.Re3 Rf7 53.h5 Rd5+ 54.Ke4 Rxf6 55.gxf6+ Kxf6 56.h6 Rd8 57.Rg3 Bd7 58.Kd3 Rf8 59.Re3 Ra8 60.Re5 Bc8 61.Rh1 Rb8 62.Rhe1 Rb7 63.d5 cxd5 64.Rc1 Rc7 65.c6 Ke7 66.Kd4 Kd6 67.Rg5 Rf7 68.Ke3 Kc7 69.Re5 Rf6 70.Kf3 Rf8 71.Rg5 draw. [Click to replay]

For game four I “chose” the Budapest, on the advice of spectators during game three. Objectively I should have stuck with the Queen’s Indian of Game 1, but I wanted variety and this seemed no worse than other options. It was not in the handicap book so I added a few key lines in the available ten minutes. However 4.e4 was not among those lines, so Rybka was already on her own. The result was acceptible this time; Rybka was not much worse other than the pawn down from the handicap. Roman then surprised everyone by sacrificing back his extra pawn with 11.e5!, which turns out to be Rybka’s choice as well given enough time. With a few perfect moves Dzindzi increased his advantage to a decisive +1.36, but threw a lot of his edge away by 16.a3? instead of 16.Qb3!. So he decided to head for a favorable but not clearly won endgame which he converted to an extra pawn. However on move 36 he erred with h4? losing back the pawn and the game soon reached a dead drawn king and pawn ending.

Dzindzichashvili,Roman (2548) - Rybka 3 (3200)
Dzindzi vs Rybka 3 handicap match (4), 28.07.2008
1.d4 Nf6 2.c4 e5 3.dxe5 Ng4 4.e4 Nxe5 5.Nf3 Qf6 6.Nc3 Bb4 7.Bd2 0-0 8.Be2 c6 9.0-0 Nxf3+ 10.Bxf3 Na6 11.e5 Qxe5 12.Re1 Qf6 13.Ne4 Qg6 14.Bxb4 Nxb4 15.c5 Kh8 16.a3 Nd5 17.Ng3 Nf4 18.Qd6 Qxd6 19.cxd6 Nd3 20.Re2 b6 21.b4 b5 22.Rd1 Nf4 23.Re5 a6 24.Rde1 a5 25.Rxb5 axb4 26.axb4 Nd3 27.Rf5 Kg8 28.Rb1 Nxb4 29.Rxf8+ Kxf8 30.Be4 Nd5 31.Bxh7 Kf7 32.Be4 Nc3 33.Rc1 Nb5 34.Rd1 Ke6 35.Bf5+ Ke5 36.h4 Ra4 37.h5 Rd4 38.Rxd4 Nxd4 39.Bg6 c5 40.Kf1 Bb7 41.f3 Kxd6 42.Nf5+ Nxf5 43.Bxf5 Bd5 44.Ke2 Bf7 45.g4 Be8 46.f4 Ke7 47.Bg6 Bxg6 48.hxg6 Kf6 49.f5 d5 50.Kd3 d4 51.Ke4 Kg5 52.Kd3 Kf6 53.Ke4 Kg5 54.Kd3 Kf6 draw. [Click to replay]

So Rybka won the match by 2½ to 1½, and my goal was achieved. However the result might have been different at a standard tournament time limit, so we still have to prove that this handicap is appropriate for a strong GM in tournament level chess. Also, there remains the goal of giving this handicap to a 2700+ player. Is it possible that some future Rybka will be able to give pawn and move at 40/2 to any human in the world with success? We are clearly not there yet, but I think this is a realistic goal. Not since Paul Morphy retired has anyone (or anything) offered pawn and move to any human alive, but perhaps it will happen again.

Replay games provided by Holger Lieske, SysOp on Playchess

Links

• Click to download the Dzindzi-Rybka games in PGN
• Not just the strongest chess program in the world
• All you need to know about Rybka
• Rybka defends Chess960 title

Prices

Note that Rybka 3 includes a database of one million games, and that the purchase of the program entitles you to one year of access to the chess server Playchess.com. Rybka 3 is a UCI engine, with 32 and 64-bit versions included in the package. Rybka can be made the default engine in ChessBase 10.

Order Rybka now