Could a modern chess engine on a Pentium have defeated Kasparov in 1996?

In February 1996, IBM's Deep Blue computer program defeated chess champion Garry Kasparov in a historic game, only to lose the rematch a year later. To accomplish this, IBM spent millions on research and developed specially designed hardware, including 30 P2SC processors and 480 custom VLSI chess co-processors. However, the argument can be made that a modern desktop computer running a more advanced chess engine could have defeated Kasparov without the need for Deep Blue.

Using a 200 MHz Intel Pentium Pro from 1995 with 64MB RAM and a custom Stockfish engine rated at an Elo of 3300, it is possible to outperform Kasparov. This engine, analyzing millions of positions per second, demonstrates the potential of advanced programming techniques over raw hardware power. Stockfish, the current top chess engine, rated at 3700 Elo on modern hardware, further emphasizes the impact of improved programming techniques.

Two experiments were conducted to test this hypothesis. The first experiment involved running the 1996 version of the engine Crafty on the Pentium Pro, achieving an Elo of 1999 at 1 second per move and 2383 at 10 seconds per move. When running the same engine on modern hardware, its Elo increased significantly to 2614 at 1 second per move.

By incorporating modern search algorithms, improved quiescence methods, and even neural-network evals, the Elo score of Crafty on the Pentium Pro was increased to 2718 at 1 second per move and 2974 at 10 seconds per move.

In the second experiment, the modern Stockfish engine was adapted for 1996 hardware, showing a rating of 2496 at 1 second per move and 2940 at 10 seconds per move. By adjusting the algorithm and optimizing it for the Pentium Pro, its Elo improved to 2846 at 1 second per move, 3109 at 10 seconds per move, and 3180 at 30 seconds per move.

Comparing the performance of these engines in a match of 100 games, a modern Stockfish engine running on a 1996 Pentium Pro won 78 games, drew 17, and lost 5 against the 1996 Crafty engine running on modern hardware. This result demonstrates that, while hardware power plays a role, the majority of the Elo gains in modern chess engines comes from advancements in programming techniques.