Modern chess engines running on very modest computer hardware can easily outclass any grandmaster. But are they actually fun to play against?
Modern chess engines running on very modest computer hardware can easily outclass any grandmaster. But are they actually fun to play against?
No.
In order for a human to play one of these monsters and still have a chance of winning, the engines must be dumbed down. You may be familiar with how they play when this is done. They will play precise, tactically perfect moves for a while, and then blunder away a piece. Or they will do something useless like taking a rook off of an open file and placing it behind an unmoved pawn that isn’t even being attacked. This is obviously not how humans play. It makes the games feel very artificial.
And why do they play like that? Because computers don’t understand abstract concepts. They merely crunch numbers. On lower levels, engines will randomly choose an inferior move—how often and how inferior the move depends on settings. They have no inkling as to whether or not it’s a move a human would make; they only know that the mathematical search algorithm has assigned it a smaller value. Zeros and ones are no substitute for ideas.
In the earlier days of computer chess a program may have a maximum strength well under 2000 USCF when playing at classical time controls. It could be made to play fairly convincing and natural games. The best way to adjust its strength was to simply limit the time it was allowed to calculate each move. It would always play the move it thought best, so move quality was consistent within each setting. This method doesn’t work with modern hardware and software because the calculation speed is exponentially higher, and a grandmaster-level move is usually found within a mere second or two.
You don’t have to buy an ancient IBM to play the old games with the limited system resources for which they were designed. There are several free and open-source software emulators you can install on your system which will mimic old computers. VICE is a very good option for those wishing to play old Commodore 64 games. DOSBox can be used to simulate older IBM and compatible machines.
Most of the games themselves can be purchased at a reasonable price. However, it is important to note that many of them are abandonware. The company that made a particular game my not even exist any more and may not have sold the rights to any other entity. In any case, these are programs that no one is interested in selling anymore, which most of the time means the rights holders (if any exist) don’t care about people redistributing copies. Because of this there are many sites that provide these programs as free downloads. But it goes without saying that you should check your local laws first.
My current preference is Chessmaster 3000 played with DOSBox. I find good challenge by turning off deep thinking (the computer is not allowed to calculate on its opponent’s time) and allowing it to think for two or five seconds per move. There are a handful of preset styles that can be selected. You can modify any of them or create your own. There are settings for things like aggression, defensiveness, etc. The “Marshall” style is very aggressive. “Nimzovitch” is as craven as its namesake.
I have included a game that I recently played against the “Lasker” style. I modified it by turning off deep thinking and limiting its book depth to 6 moves. It was given 2 seconds per move. I played some inaccuracies, but only annotated the moves that were relevant to this article.
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "CBBS"]
[Black "CM3K"]
[Result "1-0"]
1.Nf3 f5 2.g3
{An attack from one flank can be met by a counterattack from the other. My strategy against the Dutch defense is to enter into a Reti setup. An early d4 would only encourage an e4 knight outpost. The queenside flank pawns are the ones to launch. They can work in conjunction with the fianchettoed bishop.}
2...Nf6 3.Bg2 Nc6 4.c4 e5 5.d3 d5 6.cxd5 Bb4+ 7.Bd2 Nxd5 8.Bxb4 Ndxb4
{With its book depth limited, the computer is already drifting a bit from the usual kingside attack in this opening. Again, computers don't understand ideas. Just math. Consequently my strategy must now be adjusted.}
9.Nc3 Be6 10.a3 Nd5 11.Nxd5 Bxd5 12.Qa4 O-O 13.O-O Bxf3
{This move and the next few comprise an errant line that I found to be very human. You'll not find this type of natural, short-term tactical miscalculation with a modern engine.}
14.Bxf3 Nd4 15.Qc4+ Rf7?? 16.Bd5 Nxe2+ 17.Kh1 Qe7 18.Bxf7+ Qxf7 19.Qxf7+ Kxf7 20.Rfe1 Nd4 21.Rxe5 c6??
{The program cannot have a concept of the importance of the seventh rank. Rd8 would have been obvious to a human--although a stronger engine could find the move by brute force alone.}
22.Rae1 Kf6 23.Kg2 Rd8 24.Re7 Rb8 25.Rd7 c5 26.Ree7 b6 27.Rxg7 a6 28.Rxh7 Ke6 29.Rb7
{Yes, it is capable of resigning.}
1-0