Skip to content
All posts
For improvers3 min read

What a mistake costs

A mistake of 10 to 30 points of winning chances costs about 5 points in 100 games up to 999 Elo, 7 at 1400–1599 and 9 from 2200. The better the opponents, the more each mistake costs.

Result that a mistake of 10 to 30 points costs

Points per 100 games against what the ratings predict, compared with moves without a noticeable loss.

What the numbers show

  • A mistake of 10 to 30 points costs 5.4 points in 100 games up to 999 Elo, 9.1 from 2200.
  • A blunder (30 points or more) costs 8.1, 11.9 and 22.1, in the groups up to 999, 1400–1599 and 2200+.
  • The place hardly matters: on the move that leaves theory the price is within 0.5 points of the same mistake elsewhere.
  • Weaker players are punished less often. A likely reason is that their opponents use mistakes less often. That is a guess, the numbers only show the difference.
Price of a mistake in points per 100 games
Rating6 to 1010 to 3030 and up10 to 30 on the exit
up to 9993.95.48.15.9
1000–11994.06.010.16.3
1200–13993.96.110.76.4
1400–15994.16.611.97.1
1600–17993.96.512.66.7
1800–19994.17.014.77.1
2000–21994.67.817.87.7
2200+5.39.122.19.3

Moves in half-moves 8 to 24.

Questions

How is the price measured?
For every move we compare the result of the player who moved with what the ratings predict. The price is the gap to moves that lost less than 6 points. It shows what goes along with a mistake, not its cause alone.
Why 10 to 30?
In the study a move from 10 points of loss is a mistake, from 30 a blunder. The band 6 to 10 holds small inaccuracies.

Sources

  1. [1]E. B. Wilson (1927): Probable inference, the law of succession, and statistical inference. Journal of the American Statistical Association 22(158), pp. 209–212, doi:10.1080/01621459.1927.10502953. jstor.org/stable/2276774 (accessed 2 October 2026).

Data: public Lichess games from January to August 2026, 74.2 million games with an engine rating. A mistake costs at least 10 points of winning chances. Lichess analysed these games on request, so they are not a random sample of all games. Groups by the rating of the player who moves, ranges are 95 percent Wilson intervals [1].

Try it on your own games

Enter your Lichess name and see where your knowledge ends.

Try it on your own games