Tuesday, November 4, 2008

Jerome Gambit and The Perfesser (Part I)


One way a chess program's strength can be increased or decreased is by controlling the number of plys (half-moves) it analyzes before making its move. Five years ago, a strong, creative chess player -- who I will refer to as The Perfesser -- ran a simple experiment with a simple computer and a simple opening: the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+).

We will look at 1-ply, 2-ply, 3-ply and 4-ply games with notes by my learned friend.

The Perfesser - Talking LCD Chess (1 ply)
casual game 2003

After my online debut with the Jerome this morning, I played four more Jeromes just for kicks against a hand-held chess computer, Talking LCD Chess by Excalibur, setting up the first three moves by hand so that I wouldn't have to wade through a pile of Sicilians and a Two Knights Defense before getting the desired position.

(Note: Talking LCD Chess has a nasty quirk: after moving its King and returning it to e8, it will sometimes castle! Other than that it's a fine little machine to kick around.)

To simulate cluelessness, I set it first on "one ply" – and it promptly fell into

1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+ Kxf7



5.Nxe5+ Nxe5 6.Qh5+ Kf6??


The game is uninteresting and I will not bore you with it, though I recorded it.
[After 7.Qf5+ Ke7 8.Qxe5+ Kf8 9.Qxc5+ White has recovered his two pieces and is two pawns up - Rick]
Stray thought, though: at one ply the evaluation parameters must come out more clearly than with deeper lookahead. Do we see a combination of defending the piece with a penalty for "center tropism" with the King? The center of the f6 square is, geometrically, a little further from the absolute center point of the board than the center of the e6 square is, and hence presumably f6 is evaluated as "safer"
... 1-0
graphic by Jeff Bucchino, "The Wizard of Draws"

No comments: