Underpromotion Statistics

I did a statistical analysis of chess games. I took all games played over the course of one month and selected only games between stronger players, where both players had ratings above 2000 — a total of 14,043,483 games.

A pawn promotion occurred in 2,093,618 games (14.9%). Of those promotions, 98.9% were promotions to a queen, while promotions to another piece occurred in 23,948 games (1.1%).

At first glance, that may seem like quite a lot. However, after looking through those games, it turned out that in most cases these underpromotions were merely played for effect. So I used Stockfish to analyze when an underpromotion was actually justified. According to Stockfish, promotion to a piece other than a queen made genuine sense in only 1,081 cases. In all other cases, promotion to a queen was either a better move, an equally good move, or pawn promotion was not the best move in the position at all.

This means that a truly meaningful underpromotion occurs only about once every 13,000 games.

That raises an interesting question: does this rule really make sense, or should the rule simply be that a pawn always promotes to a queen?

Here are a few positions from those 1,081 games.

There is no need to make the rule less general

Those who like promoting to queens only can set the corresponding checkbox in Settings > Gameplay > Gameplay