For a very nice and well-presented introduction to the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+) check out the YouTube video "Jerome Gambit Variation of the Italian Game || Chess", presented by Think And Play Chess.
You might want to check out the video and visit the Think And Play Chess website to review and enjoy their 124 videos on a variety of chess topics and openings.
Here we have a final look at a couple of Jerome Gambit games played by maia, a computer chess program designed to play in a human-like manner.
The researchers hope to learn more about the process of human error - arguably one of the backbones of Jerome Gambit play.
drhicks1 - maia1
5 8 blitz, lichess.org, 2021
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.Qh5+ g6 7.Qxe5 d6 8.Qxh8 Qh4
.png)
9.O-O Qxe4 10.Qxh7+ Kf8 11.d3 Qe2 12.Nc3 Qxc2 13.Bh6+ Nxh6 14.Qxh6+ Kf7 15.Rab1 Qxd3 16.Rfd1 Qf5 17.Qh4 Be6 18.Qh7+ Kf6 19.Rf1 Re8 20.Qxc7 Re7 21.Qd8 Kf7 22.Nb5 a6 23.Nxd6+ Bxd6 24.Qxd6 Rd7 25.Qb4 Bxa2 26.Rbd1 Rxd1 27.Rxd1 Qc2 28.Qxb7+ Kf6 29.Qxa6+ Be6 30.Qd3 Qxb2 31.Rd2 Qc1+ 32.Rd1 Qf4 33.Qd4+ Qxd4 34.Rxd4 Ke5 35.Rd8 g5 36.Re8 g4 37.h3 gxh3 38.gxh3 Kf5 39.h4 Kg4 40.Rxe6 Kxh4 41.Rg6 Kh5 42.Rg8 Kh6 43.Kg2 Kh7 44.Rg3 Kh6 45.Kf3 Kh5 46.Kf4 Kh4 47.Rg1 Kh3 48.Rg8 Kh2 49.Kf3 Kh3 50.Rh8 checkmate
.png)
su6ar - maia9
3 2 blitz, lichess.org, 2021
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.Qh5+ Ng6
.png)
7.Qd5+ Ke8 8.Qxc5 d6 9.Qe3 Nf6 10.O-O Qe7 11.Re1 Ng4 12.Qg3 Rf8 13.Re2 Nf4 14.Re1 Nh5 15.Qc3 Nxf2 16.d3 Qh4 White resigns
.png)
We have been looking at the chess computer program Maia Chess, starting with an earlier post, where
I wondered: If its goal is "to play the human move - not necessarily the best move", and if the project wishes to "begin to algorithmically capture what kinds of mistakes players at specific skill levels make" - has it ever attempted the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+)?
After considering play with the White pieces, discussion has turned in the last couple of posts to maia as a defender against the Jerome Gambit.
One handicap that the program seems to have, is its preference for an all-too-human line in defending in the Blackburne defense to the Jerome Gambit.
It is almost as if there were some anxiety in the play.
A couple of examples:
gawrbage - maia1
10 0 rapid, lichess.org, 2021
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.Qh5+ g6 7.Qxe5 d6 8.Qxh8 Qf6
.png)
Black counters the danger along the a1-h8 diagonal.
White will exchange Queens and rely on his extra material to win.
9.Qxh7+ Qg7 10.Qxg7+ Kxg7 11.O-O Nf6 12.c3 Nxe4 13.d4 Bb6 14.Nd2 Nxd2 15.Bxd2 Bf5 16.Rae1 Rh8 17.Re3 c5 18.Rfe1 cxd4 19.Re7+ Kf6 20.h4 dxc3 21.Bg5 checkmate
.png)
BIBI31 - maia1
10 0 blitz, lichess.org, 2021
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.Qh5+ g6 7.Qxe5 d6 8.Qxh8 Qf6
.png)
9.Qxf6+ Nxf6 10.d3 Ng4 11.Be3 Bxe3 12.fxe3 Nxe3 13.Kd2 Nxg2 14.Nc3 Nf4 15.Raf1 g5 16.Nd5 Kg6 17.Nxf4+ gxf4 18.Rxf4 Be6 19.Rg1+ Kh6 20.Rh4 checkmate
.png)
[to be continued]
Looking further as to how the computer program maia fared in defending against the Jerome Gambit
It scored 21 - 43 - 8 (35%) using the Blackburne defense 1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.Qh5+ g6 7.Qxe5 d6. Apparently it was not not able to overcome the loss of the Rook after 8.Qxh8 (a complicated variation, ultimately evaluated as even), overall underperforming its rating by 130 points;
It scored 19 - 18 - 2 (51%) using the 1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.Qh5+ Ng6 defense, outperforming its rating by 50 points
We will look at some representative games in these main Jerome Gambit lines, but first will turn to the fact that maia faced 1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 6.d4 once and 1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 6.0-0 once.
The games
ZeR007 - maia9
10 2 rapid, lichess.org, 2021
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.d4 Bxd4 7.Qxd4 d6 8.f4 Nc6 9.Qc4+ Be6 White resigned
.png)
Stockfish 15.1 rates Black as about 3 1/2 pawns better.
The Database shows White scoring 2 - 10 from this position.
Vquach4310 - maia1
14 17 classical, lichess.org, 2022
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7 5.Nxe5+ Nxe5 6.O-O Passive Nf6 7.c3 d6 8.d4 Nc6 9.dxc5 Nxe4 10.cxd6 Nxd6 11.h4 Qf6 12.Qh5+ g6 13.Qd5+ Be6 14.Qa5 Nxa5 White resigned
.png)
[to be continued]
How - and how well - does maia the computer program defend against the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+)?
Overall, according to The Database, it has won 52 games, lost 65 games, and drawn 13 games - scoring 45%
It is interesting to see how the computer program handles several different challenges.
First, we will look at the infamous "Face Palm Variation", which maia utterly (and appropriately) destroys, yielding only a half point in 8 games, only because it can not figure out a checkmate while being ahead a Rook, a Bishop and 3 pawns.
ZackAddy - maia1
1 0 bullet, lichess.org, 2021
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+
4...Kxf7 5.Ng5+
In a hurry to begin attacking the enemy King, White overlooks something. The Database has over 700 examples, and I am sure that I have missed a lot more.
5...Qxg5
This was the end of the game in GusTheMaster - maia1, 10 0 rapid, lichess.org, 2021 and hariprasath21 - maia1, 0 1 bullet, lichess.org, 2021.
6.d4
.png)
This is the "trap" associated with the previous move: White uncovers an attack on the enemy Queen while also attacking the enemy Bishop.
The problem is that Black's Queen can escape the attack and create a greater threat, with 6...Qxg2, as maia1 shows.
maia1 has faced a number of interesting alternatives:
6.f4 Qxg2 7.Rg1 Qxg1+ 8.Ke2 Qf2+ 9.Kd3 Qd4+ 10.Ke2 Qxe4+ 11.Kf1 Qxf4+ 12.Ke1 Qf2 checkmate, Elsachess000 - maia1, 25 4 classical, lichess.org, 2021;
6.O-O Nf6 7.d4 exd4 8.Bxg5 Nxe4 9.Bd2 Re8 10.Re1 d5 11.f3 d3+ 12.Kh1 Nf2+ 13.Kg1 Nxd1+ 14.Be3 Bxe3+ 15.Rxe3 Nxe3 16.Kf2 Nxc2 17.Nc3 Nxa1 18.Nxd5 d2 19.Nxc7 d1=Q 20.Nxa8 Qe2+ 21.Kg3 Qe5+ White resigned, divyansh91 - maia1, bullet, lichess.org,, 2021;
6.Qf3+ Nf6 7.d3 Qxc1+ White resigned, tajikix17 - maia1, 1 0 bullet, lichess.org,, 2022; and
6.g4 Nf6 7.f4 Qxf4 8.Rf1 Qxe4+ 9.Qe2 Qxe2+ 10.Kxe2 Nd4+ 11.Kd1 d6 12.Rxf6+ gxf6 13.g5 fxg5 14.h3 Bxh3 15.Na3 Bg4+ 16.Ke1 Nf3+ 17.Kd1 Nd4+ 18.Ke1 Nf3+ 19.Kd1 Nd4+ drawn, kian095 - maia1, 10 5 rapid, lichess.org, 2023.
6...Qxg2 7.Qh5+
The text move is not the solution, but the alternative was unsuccessful as well: 7.Rf1 Qxe4+ 8.Be3 exd4 9.Nc3 dxc3 10.bxc3 Bxe3 11.Qh5+ g6 12.fxe3+ Nf6 13.Rxf6+ Kxf6 14.O-O-O gxh5 15.Rf1+ Ke5 16.Re1 d5 17.Kb2 d4 18.exd4+ Nxd4 19.Rxe4+ Kxe4 20.cxd4 Kxd4 21.Kb3 Be6+ 22.Kb2 c5 23.Ka3 b5 24.Kb2 c4 25.Ka3 a5 26.Kb2 b4 27.Kc1 a4 28.Kb2 b3 29.axb3 axb3 30.c3+ Kd3 31.Kb1 Kxc3 32.Kc1 Ra1 checkmate, wakandaforever123 - maia1, 1 0 bullet, lichess.org, 2021
7...g6 8.Rg1 Qxg1+
.png)
Black has a checkmate in 8 moves.
[to be continued]
Thinking about Maia Chess, mentioned in the previous post, I wondered: If its goal is "to play the human move - not necessarily the best move", and if the project wishes to "begin to algorithmically capture what kinds of mistakes players at specific skill levels make" - has it ever attempted the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+)?
Poring through several million (!) of maia games*, I found two Jeromes by the lesser-skilled engine, maia1, and two by the highest-skilled, maia9.
They were all very, very human.
maia1 - yung-m90
10 3 rapid game, lichess.org, 2022
1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+
Another game continued 5.Ng5+ (You can hardly get more "human" than the "Face Palm Variation"! It appears that maia1's opponent is equally skilled) Ke8 6.Qf3 Nh6 7.d3 Nd4 8.Qh5+ Kf8 9.Nc3 c6 10.Be3 Nxc2+ 11.Kd1 Nxa1 12.Bxc5+ d6 13.Be3 Bg4+ 14.f3 Bxh5 15.Kd2 Qb6 16.Rc1 Nf7 17.Nxf7 Bxf7 18.Bxb6 axb6 19.Rxa1 g6 20.b4 Kg7 21.Rg1 c5 22.bxc5 bxc5 23.Ra1 Be8 24.Nd5 Bf7 25.Nc7 Ra5 26.Ke3 Rc8 27.Nd5 Bxd5 28.exd5 Rca8 29.Ke4 Rxa2 30.Rf1 Ra1 31.Rxa1 Rxa1 32.Ke3 Ra2 33.g3 Rxh2 34.f4 Rh3 35.Kf3 exf4 36.Kxf4 b5 37.Kf3 b4 38.Ke2 b3 39.Kd2 Rh2+ 40.Kc3 b2 41.Kb3 b1=Q+ 42.Kc3 Rc2 checkmate, maia1 - CatrinaM, 7 5 rapid, lichess.org, 2022.
5...Nxe5
Or 5...Kf8 6.Nxc6 bxc6 7.d4 Be7 8.Qf3+ Nf6 9.O-O d5 10.Bg5 h6 11.Bxf6 Bxf6 12.e5 Ba6 13.exf6 Bxf1 14.fxg7+ Kg8 15.gxh8=Q+ Kxh8 16.Kxf1 Qe7 17.Nd2 Re8 18.Qe3 Black resigned, maia9 - Jam8sJ0yce, 5 2 blitz, lichess.org, 2023
6.Qh5+
Otherwise 6.d4 Bxd4 7.Qxd4 Qf6 8.O-O Nf3+ 9.gxf3 Qxd4 10.Be3 Qxb2 11.Nd2 Qxc2 12.Rac1 Qxa2 13.Rxc7 Ne7 14.Nc4 Ke6 15.Rd1 d5 16.exd5+ Nxd5 17.Re1 Nxc7 18.Bf4+ Kf5 19.Be5 Qxc4 20.Bxg7 Rg8 21.Re5+ Kg6 22.Rg5+ Kxg5 23.Kg2 Rxg7 24.h4+ Kf6+ 25.Kh2 Black resigned, maia9 - soyunbotenajedrez_20, bullet, lichess.org, 2022
6...g6 7.Qxe5 Nf6
.png)
Readers of this blog might not expect outsiders to know Jerome Gambit theory: best is Whistler's 7...Qe7; historic is Blackburne's 7...d6.
8.Qxc5 Nxe4 9.Qc4+ Kg7 10.Qxe4
.png)
This is so "human" that it hurts.
10...Re8 11.Qxe8 Qxe8+ 12.Kd1 d5 13.Nc3 Bg4+ 14.f3 Bh5 15.Re1 Qc6
.png)
16.d3 g5 17.g4
Better late than never? Not really.
17...Bg6 18.Bxg5 d4 19.Ne4 Bxe4 20.Rxe4 Kg6 21.Be7 Re8 22.Bh4 Rxe4 23.fxe4 Qe8
.png)
In fact, the human now gives a lesson on the dominance of Her Majesty over the enemy pieces.
24.Ke2 Qf7 25.Rf1 Qxa2 26.Rf6+ Kg7 27.Rf3 Qxb2 28.e5 Qxc2+ 29.Kf1 Qd1+ 30.Be1 Qxf3+ 31.Bf2 Qxd3+ 32.Kg2 Qe4+ 33.Kg1 d3 34.Bxa7 d2 35.Bb8 d1=Q+ 36.Kf2 Qef3 checkmate
.png)
* I had it easy. According to the profile of Maia Chess 1100 at lichess.org, "This version was trained by learning from over 10 million Lichess games between 1100s."
[continued from the previous post]
Certainly a blitz game (unknowingly, at the time, against a computer) was a diversion from the events of the day, but it was not the best choice. Tennis, anyone?
perrypawnpusher - guestM
5 1 blitz, FICS, 2023
26.Kh2 Qh6+ 27.Kg1 Rxg3 28.Qe4 Bh3 29.Rff2
.png)
My Jerome Gambit has been a wild ride, but not a successful one.
It is time for my opponent to finish me off.
29...Bxg2
Forcing, but not best.
30.Rxg2 Rxg2+ 31.Rxg2 d5
Those annoying pawns. (Page Dante Alighieri.)
32.Qe5 Rg8
.png)
Odd.
Stockfish 16's analysis starts off with the recommended 32...c4, which seems more to the point.
My opponent seems to have lost its way.
33.Qxd5
What else?
After the suggested improvement 33.Rg5 Rd8 the pawn is no longer available, although after the game I let Stockfish analyze for a while and it struggled to convert Black's advantage.
33...Qxf4 34.Qxc5 Qc1+ 35.Kh2 Qf4+ 36.Kg1 Qc1+
.png)
Or is it?
37.Kh2
Giving Black's Queen choices. Too many choices? Perhaps.
37...Qf4+
Again, Stockfish 16 prefers 37...Qh6+, but it still has a hard time bringing the whole point home quickly.
38.Kg1
.png)
At this point the game was drawn by repetition.
'Tis a puzzlement.
And, for me, an escape. I was happy to collect the half point.
By the way, it turns out that my opponent was Maia Chess, "a human-like neural network chess engine"
Maia’s goal is to play the human move — not necessarily the best move. As a result, Maia has a more human-like style than previous engines, matching moves played by human players in online games over 50% of the time..
Maia is an ongoing research project using chess as a case study for how to design better human-AI interactions. We hope Maia becomes a useful learning tool and is fun to play against. Our research goals include personalizing Maia to individual players, characterizing the kinds of mistakes that are made at each rating level, running Maia on your games and spotting repeated, predictable mistakes, and more...
Maia is particularly good at predicting human mistakes. The move-matching accuracy of any model increases with the quality of the move, since good moves are easier to predict. But even when players make horrific blunders, Maia correctly predicts the exact blunder they make around 25% of the time. This ability to understand how and when people are likely to make mistakes can make Maia a very useful learning tool.
My interest in the Jerome Gambit, as I have pointed out in the past, comes from investigating "errors in thinking", so perhaps Maia and I are not so far apart, after all.
