To beat a human being a chess program does not have to make a great move every time, it only has to make good move after good move after good move. By way of contrast, if it makes a lesser move followed by a lesser move followed by a lesser move - it risks defeat, as the following Jerome Gambit game shows.
Wall, Bill - AnMon Engine
Palm Bay, FL, 2015
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. Bxf7+
4...Kxf7 5.Nxe5+ Nxe5 6.Qh5+ Ke6 7.f4 d6
Once again the "annoying" or "silicon" defense. (A favorite among computer engines, it works well for humans, too.)
8.fxe5 dxe5 9.Qh3+ Kd6
Black's King goes directly to d6, instead of 9...Ke7 10.Qg3 Kd6 as in Wall, Bill - Comet B50 engine, Palm Bay, FL, 2015 (1-0, 16). The saved tempo improves Black's position.
10.Qg3 Nf6
The right idea: development is key for Black, and taking the g-pawn would be dangerous for White.
11.Nc3 a6 12.Rf1 Be6 13.Ne2 Nh5 14.Qf3 Qh4+ 15.g3 Qh3
White now offers a pawn which Black would do well to decline, but computers are known to be grabby. It is a small slip, but small slips add up...
16.d4 exd4 17.Qd3 Qxh2 18.Nxd4 Bxd4 19.Qxd4+ Kc6
Black has to be careful; there is a way to King safety through this messy position,
but one error can be fatal. This was the whole idea behind White giving up the h-pawn.
20.Qc3+ Kd7 21.Bf4 Ke8 22.O-O-O
Whew, says the human.
22...Nxf4 23.gxf4 Rg8
If White has not equalized, he is pretty close. In the meantime, he has his familiar "Jerome pawns" in play against Black's extra piece. The protected, passed pawn at h7 does not get a chance to play a part in the game.
24.f5 Bf7 25.e5 Qh6+ 26.Kb1 Qc6 27.Qh3 Bd5 28.e6
h6 29.Qh5+ Ke7 30.Rfe1 Raf8
AnMon is not familiar with this blog, or it would be feeling an eerie sense of deja vu.
31.f6+ Rxf6 32.Rxd5 Rxe6 33.Rxe6+ Kxe6
Black's King is too exposed.
34.Qf5+ Ke7 35.Qe5+ Kf7 36.Qf4+ Ke7 37.Qe4+ Kf7 38.Rf5+ Kg6 39.Qd3 Rd8 40.Rd5+ Kf7 41.Rxd8 Qh1+ 42.Qd1 Black resigned
As a followup to the human vs computer game Wall - Akok, 2015 the following game addresses a 12th move alternative for Black. At that point things have gotten critical for the defender (see the notes in "Irrational"), but it would be a shame to hide the current contest in the notes of another game: once again, disaster strikes the computer.
Wall, Bill - Amyan engine
Palm Bay, FL, 2015
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+
4...Kxf7 5.Nxe5+ Nxe5 6.Qh5+ Ke6 7.f4 Qf6
8.Rf1 g6 9.Qh3+ Ke7 10.Nc3 Kd8 11.fxe5 Qxe5
The Black King is not as safe as it would like, and the Black Queen's capture is ill-fated. White's next move shows that the defenders are over-worked.
12.d4 Qxd4
Wall - Akok, 2015 saw 12...Bxd4
13.Bg5+ Ke8
Costing material, despite the eventual Queen exchange, is 13...Ne7, i.e. 14.Bf6 d6 15.Qg3 Qe3+ 16.Qxe3 Bxe3 17.Bxh8.
14.Qf3 Qg7 15.Nd5 Bd6 16.O-O-O
Who wouldn't want to have this position with White against a computer?
16...c6
You know that Black is in trouble when the best line Stockfish recommends is 16...Be5 17.Qc3!? Things would continue dismally: 17...Bf4+ (everything else leads to checkmate) 18.Rxf4 Qxc3 19.bxc3 Ne7 20.Nxc7+ Kd8 21.Nd5 Ke8 22.Bxe7 Rb8 23.Rdf1.
17.e5 Bb8 18.e6 d6
Black, trying to keep the position closed, doesn't have time right now to take the offered Knight. Alas, undeveloped, there is also the problam of a pesky "Jerome pawn".
19.Qf7+ Qxf7 20.exf7+ Kf8 21.fxg8=Q+ Kxg8
22.Rde1 cxd5 23.Re8+ Kg7 24.Re7+ Kg8 25.Bh6 Black resigned
Next is checkmate.
Here we have another human vs computer game (see "Irrational"), one which turns, curiously, on computer "psychology" and a subtle anti-computer strategy. Again, it is the human who applies brutal tactical force to close out the game.
Wall, Bill - Comet B50 engine
Palm Bay, FL, 2015
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+
4...Kxf7 5.Nxe5+ Nxe5 6.Qh5+ Ke6 7.f4 d6
Ah, yes, back to the "annoying" or "silicon" defense.
Black will soon be challenged to decide which pawn(s) it wants to protect, and which one(s) it wants to let go. This is something White can take advantage of, if he pushes it.
8.fxe5 dxe5 9.Qh3+ Ke7
Ten years ago this position proved uneasy for the computer playing the Jerome Gambit, and it quickly decided to save half a point: 9...Kd6 10.Qd3+ Ke7 11.Qg3 Kd6 12.Qd3+ Ke7 13.Qg3 Kd6 14.Qd3+ drawn, Crafty 19.19 - RevvedUp, blitz 2 12, 2006.
10.Qg3 Kd6 11.Qd3+ Bd4
Again, here, in the human - computer supermatch played a decade ago, the computer, with the Jerome Gambit, decided to bail out: 11...Ke7 12.Qg3 Ke6 13.Qh3+ Kf7 14.Qh5+ Ke6 15.Qh3+ drawn, Hiarcs 8 - RevvedUp, blitz 2 12, 2006).
Upon reflection, the g7 pawn is probably more valuable than the e5 pawn, despite the Comet B50's evaluations and calculations, and ...Kd6 is not Black's strongest continuation. Here it will cost a piece.
(On the other hand, most computers opening with the Jerome Gambit as White would probably love to offer and receive a draw after four moves!)
12.c3 Qg5
Comet B50 goes for wild tactics. It is interesting to recall two historical games that showed the computer (in this case, an early version of Fritz) solidly surrendering the piece: 12...c5 13.cxd4 cxd4 14.b3 Kc7 (14...Nf6 15.Ba3+ Kc7 16.Qg3 Re8 17.Qxg7+ Kb8 18.d3 Qa5+ 19.b4 Qb6 20.O-O Re6 21.Nd2 Qd8 22.Nc4 Qg8 23.Qxg8 Nxg8 24.Rf5 Ne7 25.Rxe5 Rxe5 26.Nxe5 Ng6 27.Nf3 Nf4 28.b5 Kc7 29.Ne5 Ng6 30.Nxg6 hxg6 31.Bc5 Bd7 32.a4 Re8 33.Bxd4 a6 34.bxa6 bxa6 35.a5 Kd6 36.Bb6 Bc6 37.Ba7 Bb5 38.Rd1 Ke5 39.Kf2 Ra8 40.Bb6 Ba4 41.Ra1 Bc6 42.Ke3 Re8 43.d4+ Kd6 44.e5+ Kd7 45.g3 Rf8 46.Rd1 Ke6 47.Rd3 Rf1 48.Rc3 Rf3+ 49.Kd2 Rxc3 50.Kxc3 Kd5 51.h4 Ke4 52.Kc4 Bb5+ 53.Kc5 Kf3 54.d5 Kxg3 55.e6 Kxh4 56.d6 Kg5 57.d7 Kf5 58.d8=Q Kxe6 59.Qg8+ Kf5 60.Qd5+ Kf6 61.Kd6 Kg7 62.Qxb5 axb5 63.a6 b4 64.a7 g5 65.a8=Q g4 66.Qe4 g3 67.Qxb4 Kf7 68.Qf4+ Kg6 69.Qg4+ Kh6 70.Qg8 Kh5 71.Bd8 Kh6 72.Qg5+ Kh7 73.Bf6 g2 74.Qg7 checkmate, Fisher-Kirshner,M - Knight Stalker, Fremont, CA, 1993) 15.Qc4+ Kb8 16.Ba3 Qh4+ 17.Kd1 Qh6 18.Qd5 Bg4+ 19.Ke1 Qh4+ 20.g3 Qg5 21.Bd6+ Kc8 22.Qf7 Bd7 23.Na3 Kd8 24.Rc1 Ne7 25.Rf1 Rc8 26.Rxc8+ Nxc8 27.Nc4 Re8 28.Bb4 Qh6 29.Na5 b6 30.Nc4 Bc6 31.d3 Qc1+ 32.Kf2 Qc2+ 33.Kg1 Qxd3 34.Nd6 Qe3+ 35.Rf2 Nxd6 36.Bxd6 Bd7 37.Kg2 Qxe4+ 38.Kf1 Bh3+ 39.Rg2 Qxg2+ 40.Ke1 Qh1+ 41.Kd2 Qxh2+ 42.Ke1 Qxg3+ 43.Kd2 Qc3+ 44.Ke2 d3+ 45.Kf2 Qb2+ 46.Kg3 Qg2+ 47.Kh4 Qg4 checkmate, Fisher-Kirshner,M - Knight Stalker, Fremont, CA, 1993.
13.cxd4 Qxg2 14.dxe5+
14...Kxe5
The King would be relatively safer on e7.
15.Qd5+ Kf6
The Queen is now lost, but otherwise Black loses her and his King: 15...Kf4 16.d4+ Kf3 17.Nd2+ Kg4 18.h3+ Qxh3 19.Rxh3 Kxh3 20.Qh5+ Kg2 21.Qf3+ Kh2 22.Nf1+ Kg1 23.Be3 checkmate.
analysis diagram
16.e5+ Black resigned
The idea of playing the Jerome Gambit against a computer chess engine seems almost as irrational as the Jerome Gambit itself. How is it possible to give the silicon beast a couple of pieces and expect to survive? One doesn't give the computer "Jerome Gambit odds"!
Still, four years ago, I noted on this blog
It was not long after I began looking into the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+) that I noticed one chess match kept turning up in the databases: in 1993, between the human Micah Fischer-Kirshner and the chess-playing program KnightStalker, an early version of Fritz. As luck would have it, I was able to interview Micah about his experience for this blog.
The Jerome Gambit seemed a natural for matches, especially ones involving computers. Jeroen_61 of the Netherlands ran one with Hiarcs, Junior 7, Shredder Paderdorn (6.02) and Fritz 7.
I tried a few myself, notably a Fritz8 vs Fritz5 encounter and a Delphi vs Wealk Delphi contest. Each attempt had its shortcomings. (Perhaps you read about them here.)
The mysterious "perfesser" played an introductory 4-game match with the Talking LCD Chess Gadget. Like the Jerome Gambit itself, it was good for some chuckles.
Topping all efforts so far, "RevvedUp" and his trusted companions Hiarcs 8, Shredder 8, Yace Paderborn, Crafty 19.19 and Fritz 8 explored the Jerome Gambit in a 30-game encounter. It was simply war.
There is also the earlier summary of Randy Tipton's computer vs computer games, and the more recent adventures of "Ionman vs the Bots".
So, when I noticed that the new group of Bill Wall games had some of him playing the Jerome Gambit against some engines, I had to check them out.
The following game is very interesting. The human's flurry of tactics to finish off the game is impressive.
Wall, Bill - Abrok chess engine
Palm Bay, Florida, 2015
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+
4...Kxf7 5.Nxe5+ Nxe5 6.Qh5+ Ke6 7.f4 Qf6
A mild surprise. Computers tend to delight in "the annoying defense" 7...d6.
8.Rf1 g6 9.Qh3+
Bill has a couple of cautions here: 9.fxe5? Qxf1+ 10.Kxf1 gxh5 and 9.Qd1? Qh4+ 10.g3 Qxh2 11.fxe5 Qxg3+ 12.Rf2 Qxf2 checkmate.
9...Ke7
Bill has been here before:
9...Ng4 10.Qxg4+ Ke7 11.Nc3 d5 (11...d6 12.Nd5+ Kd8 13.Nxf6 Bxg4 14.Nxg4 h5 15.Nf2 Kd7 16.Nh3 Re8 17.d3 Nf6 18.Ng5 Ng4 19.h3 Nh2 20.Rh1 Black resigned, Wall,B - Guest3164644, PlayChess.com, 2013) 12.Nxd5+ Black resigned, Wall,B - Betarsolta, PlayChess.com, 2015; also
9...Kf7 10.fxe5 Qxf1+ 11.Kxf1 d6 12.Qc3 Black resigned, Wall,B - Guest1690223, PlayChess.com, 2012; and
9...Ke7 10.Nc3 c6 d6 (10...c6 11.fxe5 Qxe5 12.d3 Nf6 13.Qh4 Bd4 14.Bg5 Bxc3+ 15.bxc3 Qxc3+ 16.Ke2 Qxc2+ 17.Bd2 Qb2 18.Rxf6 Qxf6 19.Bg5 Rf8 20.Bxf6+ Rxf6 21.e5 Black resigned, Wall,B - XCCY, FICS, 2011) 11.Nd5+ Kd8 12.Qg3 Qe6 13.fxe5 dxe5 14.d4 Ne7 15.dxc5 Nxd5 16.Bg5+ Ke8 17.exd5 Qxd5 18.Rf2 Be6 19.Rd2 Qxc5 20.0-0-0 Rf8 21.Qh4 h5 22.Bh6 Rf5 23.Rd8+ Rxd8 24.Rxd8+ Kf7 25.Rf8+ Qxf8 26.Bxf8 Kxf8 27.Qd8+ Kg7 28.Qxc7+ Kh6 29.h4 Rf1+ 30.Kd2 Rf2+ 31.Ke3 Rf5 32.Qd8 a6 33.Qh8 checkmate, Wall,B - Aburasian, Chess.com, 2010.
10.Nc3 Kd8
This is a small slip.
It must be noted that in a Jerome Gambit computer vs computer tournament over a decade ago, Hiarcs 9 chose 10...Kd8 three times:
Junior 7 - Hiarcs 9
Jerome Gambit, The Jeroen Experience, 2003
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.Qh5+ Ke6 7.f4 Qf6 8.Rf1 g6 9.Qh3+ Ke7 10.Nc3 Kd8 11.fxe5 (11.Nd5 d6 12.Qc3 Qh4+ 13.g3 Qxh2 14.fxe5 Bg4 15.Qd3 Ne7 16.Nf4 Nc6 17.Ne6+ Bxe6 18.exd6 Ne5 19.Qe2 Qxg3+ White resigned, Fritz 8 - Hiarcs 9, The Jeroen Experience, 2003) 11...Qxe5 12.d4 (12.d3 Bb4 13.Qh4+ Ne7 14.Bd2 d5 15.0-0-0 d4 16.Bf4 Qe6 17.Nd5 Bd6 18.Bxd6 cxd6 19.Rf6 Qe5 20.Nxe7 g5 21.Qh6 Qxe7 22.Rxd6+ Bd7 23.Rxd4 Kc7 24.Rc4+ Bc6 25.Qh3 b5 26.Rc3 Kb7 27.d4 Bxe4 28.Re1 Rad8 29.Qd3 Rhe8 30.Qxb5+ Ka8 31.Rce3 Qd6 32.Rxe4 Rxe4 33.Rxe4 Qxh2 34.Qf1 h5 35.c3 g4 36.Qe2 Rc8 37.Re8 Rxe8 38.Qxe8+ Kb7 39.Qe4+ Kb6 40.c4 Qg1+ 41.Kd2 Qf2+ 42.Kc3 Qg3+ 43.Kb4 Qf2 44.Qe6+ Kc7 45.Qe7+ Kc8 46.Kc5 Qxb2 47.Qe8+ Kc7 48.Qc6+ Kb8 49.a4 Qb6+ 50.Qxb6+ axb6+ 51.Kd5 h4 52.Ke4 Kc7 53.Kf4 h3 54.gxh3 gxh3 55.Kg3 h2 56.Kxh2 Kd7 57.Kg3 Kc7 58.Kf4 Kc6 59.Ke5 Kc7 60.Ke6 Kb7 61.d5 Kc8 62.Ke7 Kc7 63.d6+ Black resigned, Fritz 5.32 - Hiarcs 9, The Jeroen Experience, 2003) 12...Bxd4 13.Rf8+ Ke7 14.Qf3 g5 15.Rf5 Bxc3+ 16.bxc3 Qg7 17.Rxg5 Qf6 18.Bf4 d6 19.e5 dxe5 20.Qe3 h6 21.Bxe5 Qxg5 22.Bf4+ Kf7 23.Bxg5 hxg5 24.Qxg5 Nf6 25.0-0-0 Rh5 26.Qf4 Rc5 27.Rf1 Rf5 28.Qxc7+ Kg6 29.Re1 b6 30.h4 Rc5 31.Qg3+ Kf7 32.Qd6 Bg4 33.Re7+ Kg6 34.Qd3+ Rf5 35.c4 Rh8 36.Qg3 Rc5 37.Re6 Kg7 38.Rxf6 Kxf6 39.Qxg4 Rhc8 40.Qd4+ Ke6 41.Qe4+ Kf6 42.Qf3+ Ke6 43.g4 Rxc4 44.Qf5+ Ke7 45.g5 R4c5 46.Qe4+ Kd6 47.Kb2 b5 48.g6 Rc4 49.Qf5 R4c5 50.Qf6+ Kd7 51.g7 Rxc2+ 52.Kb3 R2c4 53.g8R Rxg8 54.Qf7+ Kd6 55.Qxg8 a6 56.Qxc4 bxc4+ 57.Kxc4 Ke5 58.Kb4 Kf4 59.a4 Kg4 60.Ka5 Kxh4 61.Kxa6 Kg3 62.Kb5 Kf4 63.a5 Ke5 64.a6 Kd6 65.a7 Ke5 66.Kc5 Ke4 67.a8Q+ Ke3 68.Qd5 Ke2 69.Kd4 Kf2 70.Qe4 Kf1 71.Ke3 Kg1 72.Kf3 Kh1 73.Qc2 Kg1 74.Qg2 checkmate
11.fxe5 Qxe5
This mistake costs the game. Black had to overprotect f8 with 11...Qg7.
12.d4
The move that the computer either overlooked or incorrectly evaluated.
12...Bxd4 13.Rf8+ Ke7 14.Qf3 Qc5 15.Nd5+ Kd6 16.c3 Bg7 17.Be3 Qc4
18.e5+
If the computer could groan, it would.
18...Bxe5 19.Rd1 Bg7 20.Nb6+ Black resigned
