In summer/autumn 1998 Benjamin Good from Pittsburgh played 2 e-mail games against Dabbaba (programmed by Jens Bęk Nielsen, Denmark). Dabbaba was allowed up to an hour pr. move on a Pentium II 233 Mhz. A chess variant called nightrider chess was played. The knights make long moves in this variant. Benjamin Goods uscf chess rating is somewhere in the range of 1500-1600, and he had a hard time in this difficult variant against the computer. You have to be aware right from the beginning: 1.e4?,Nxe4! Here are the games. (a positive score means Dabbaba thinks white has the best position) Game 1. White: Benjamin Good ------------- information on Dabbabas move --------------- whites seconds Dabbabas score depth Dabbabas expected move used move (next move) line ---------------------------------------------------------------------- 1. d2-d4 2119 d7-d5 8 8,1 e2e3 c8d7 g1e2 2. e2-e3 3328 Nb8-d7 3 7,23 g1d7 c8d7 Dabbaba now forces exchange of a couple of nightriders now because of the threats 3.-,Nxa1 and 3.-,Nxg1. 3. Ng1xd7 3470 Bc8xd7 24 8,10 b1h4 a8b8 h4d2 White thinks he gains a pawn now, but misses blacks reply. (look at Dabbabas expected line: 4.Nh4,Rb8 - attack and defense of b7!) 4. Nb1xd5 3291 Bd7-g4! -7 8,17 d5c7 d8c7 d1g4 White should now capture on c7 to get a pawn more for the nightrider. 5. Qd1xg4? 1608 Qd8xd5 -166 8,1 g4d1 e7e5 After 5 moves there is only one nightrider left. But that is enough to make this game quite special. After this game I have adjusted the value of the nightrider from 550 to 625, so the position-evaluation in this game would have been 75 lower. 6. c2-c4 3284 Qd5-a5+ -128 8,23 e1d1 g8f6 If now 7.Bd2???,Qxd2# - Ng8 covers d2! White did not see that when he played 6.c4. The position of the white king becomes bad now. 7. Ke1-d1 3287 e7-e5 -191 7,23 f1d3 g8f6 g4e2 8. Qg4-f3 1670 Ng8-c6 -146 7,1 f1d3 e5d4 e3d4 9. Bc1-d2 1467 Bf8-b4! -269 8,1 d2c1 a8d8 d4d5 Dabbaba considers 10.Bc1 as the best move, but who wants to play that? After 10.Qe2 black captures twice on d2 and wins a pawn on d4. After 10.d5,Ne7 11.Bxb4,Qxb4 black threats on both b2 and h1. 10. Bd2xb4 2179 Qa5xb4 -294 8,1 d1c2 e5d4 f1d3 White could consider 11.Qe2 to cover b2, and if black plays 11.-,e5xd4 white can recapture with check. But instead black plays 11.-,0-0-0 with big pressure on d4 (12.d5??,Nxe2). White can try 12.Qg4+,Kb8 (if now 13.Qxg7?,Qxb2 14.Rc1,Nxd4! with mate in 6 according to Dabbaba; fx 15.Bd3,Nf3!), but if Dabbaba plays against itself from here this line occurs: 13.Kc2,exd4 14.Qxg7,Rhg8 15.a3,Rxg7 16.axb4,Nxb4+ 17.Kb3,Nxf2 18.Rg1,dxe3 and white can resign. 11. b2-b3 1221 e5xd4 -319 7,1 e3d4 c6d4 Dabbaba also considered 11.-,Ng8 with matethreat on d2... 12. Qf3-e4+ 1683 Nc6-e7 -657 8,1 g2g3 e8c8 White could now try 13.g3,0-0-0 (threats -,Nc3+) 14.Bh3+,Kb8 15.Kc2,Qc3+ 16.Kb1,Qxe3+! (discovered check by Ne7!) Other lines are not much better. 13. e3xd4? 3303 0-0 -730 7,22 f1d3 e7c3 Unpins Ne7 that threats on c3 and h1 now. Game over. Dabbaba could also have played 13.-,0-0-0 14.Qg4+,Kb8 15.g3,Qc3! (better than 15.-,Rxd4 16.Qxd4,Nf5+ 17.Qe3) 14. Bf1-d3 3277 Ne7-c3+ -677 7,35 ---- White could mate if black captured Rh1. 15.Resigns. max nodes pr. move: 18728 at move 11. min nodes pr. move: 17145 at move 12. Game 2. Black: Benjamin Good --------- information on Dabbabas move -------- Dabbabas blacks score seconds depth Dabbabas expected move move used (next move) line -------------------------------------------------------------------- 1. d2-d4 d7-d5 67 1326 8,1 d7d5 g2g3 c8f5 2. Ng1-e5 e7-e6 43 3347 7,18 e7e6 b1f3 g8e4 3. Nb1-f3 Bf8-d6 3 1395 7,1 g8e4 g2g3 4. Ne5-d3 Ng8-e4 29 1994 8,1 g8e4 c1d2 considered Ng4+ 5. Bc1-d2 0-0 11 3342 7,7 b7b6 e2e3 b8d7 6. g2-g3 b7-b6 16 3312 7,22 b7b6 f1g2 7. Bf1-g2 Bc8-a6 39 3346 7,8 c8a6 d3c1 d8e8 <<< dabbaba improved >>> From here Dabbaba plays with 'positional selektivity'. This increases the searchspeed with 40%. Unfortunately I was not aware, that I also introduced two errors regarding the hashtables. 8. Nd3-c1 c7-c5 -55 1207 7,1 ---- <<< dabbaba improved >>> Hashtables is turned on again. 9. Nf3-g1 c5xd4 166 2411 8,1 c5d4 g2e4 <<< dabbaba improved >>> Hashtables now at full strength; 2.048 entries instead of 256. 10. Bg2xe4 d5xe4 135 3330 8,23 d5e4 a2a3 d8f6 11. e2-e3 d4-d3 165 1252 8,1 e6e5 e3d4 e5d4 12. a2-a3 Nb8-e2 174 1516 8,1 a6b7 c1a2 d8f6 <<< dabbaba improved >>> Dabbaba now plays with 'nullmove'; observe the depth (incl. nullmoves!) The speed should be increased by further 30% with 'nullmove'. 13. Nc1-a2 Qd8-c7 285 1423 13,1 d8e7 g1e2 d3e2 14. Ng1xe2 d3xe2 330 1972 12,1 d3e2 a2e4 f7f6 15. Na2xe4+ f7-f6 320 1382 17,1 f7f6 d1c1 ms-error.... well, look at this analysis by Dabbaba and me: -,Qf1+! (if now Kd2,e1Q+ mate) Rxf1,exf1Q+ Kd2,Rd8+ (if now Rd3,Rxd3+ cxd3,Qxd3+ Ke1,Qe2+ mate - a mate similar to the one you had seen) Kc3,Rxd1 Qb2 (here I saw -,Rd3+, and Dabbaba confirms it is good; fx -,Rd3+ Kxc4,Rxa3+ Kd4,Qd1+ with mate in a few moves, but Dabbaba found:) -,Bd5 and white must loose his queen and play Qa2 to avoid -,Qe1+ mate. Do you agree? If so there is no reason to continue play and you will loose the game again... And 25.-,Bxb4 seems to be okay. A pity you did not see the end of the pretty combination. Dabbabas 15.Nxe4+ was a programming error, that is corrected now. You should have played 16.-,Bb7!