Friday, April 2, 2021

Jerome Gambit: Artificial Wha?



Sometimes a chess computer program can act as an interesting model of a human chess player.

Lately, in part because of my interest in the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+) and my interest in the notion of "errors in thinking", I have been exploring ways to make a chess computer program dumber

I have done this before: see "Artificial Stupidity" and "Artificial Ignorance (Part 1)".

I mean, I have already learned from chess programs some ideas on how to play chess smarter, like

-give myself a reasonable amount of thinking time per move;

-increase the number of choices that I consider each time I get ready to move, not just look at the "obvious" one; and

-force myself to look one or two ply deeper than I would normally, when evaluating alternatives    

Conversely, if I want my chess computer program to be less smart, I could decrease the available "thinking" time per move, decrease the number of search nodes for each move, or decrease the search ply depth.

One of the more interesting ideas I found in an internet discussion was to have the computer chess engine vary how often it plays what it has calculated as the "best" move, versus playing alternatives it saw as less strong. For example, it could consider any move it evaluated as anywhere near (a pawn's difference, a piece's difference, etc.) the top move as equally playable as the "best". Or, the engine could be set to play the "best" move a certain percentage of the time, and a random move the rest of the time.

Much of this thinking was spurred by the following game.

At the time, I was bored, but not too bored; I wanted a challenge, but not too much of a challenge; I was ready to go to sleep, but not quite ready. 

So I turned to the Chess Titans chess program (for an earlier example see "Contempt?!"), which was set to the Beginner level.


Rick - Chess Titans

beginner level, casual game, 2021

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


Of course, the Jerome Gambit. I probably would have just turned the thing off if it had wanted to play the Nadjorf Sicilian.

4...Kxf7 5.Nxe5+ Nxe5 6.Qh5+ 


6...Ke6 7. Qf5+ Kd6 8. f4 Qh4+ 


Wow! The computer is interested in playing the very sharp "Nibs" variation. That has to be part of its "book".

Funny thing, this has happed before, but I forgot about it. See "Jerome Gambit: Against the Titans (Part 1)"

9.g3 c6 10.Qxe5 checkmate


Well... uh... okay.

I suspect that many chess computer programs would have found 9...Nf3+, instead, and that others would have opted for 9...Qf6.

Quite possibly Chess Titans went for a random suggestion, instead of the best, on move 9.

(The Database points out that 3 of my earlier games - blitz, against human opponents - continued 9...Nf3+ 10.Kd1 Qe7 11.Qd5#. Ahem.)

No comments:

Post a Comment