ACPL And Cheat Detection
By Akhilesh Dongre, Jatin Madan and Sreya Vallabhaneni
Introduction
“Chess is not a game. Chess is a well-defined form of computation. You may not be able to work out the answers, but in theory, there must be a solution, a right procedure in any position”
— John von Neumann
Computers can give you a detailed and thorough evaluation of your play in online chess, using a variety of small metrics to indicate how efficiently you played your game. One of the important metrics is average centipawn loss, albeit it does require a little more investigation to understand it from the standpoint of a person.
What is ACPL?
The average centipawn loss (aCPL) is the amount by which a chess engine’s evaluation of the position changes after each of the player’s moves.
A centipawn is one-hundredth of a pawn, i.e., 100 centipawns = 1 pawn. And the average centipawn loss (aCPL) metric represents how much “value” you drop by making incorrect moves across a chess game.
The pieces typically have an integer value in pawns but using the centipawn makes it possible to analyze strategic aspects of the situation that are worth less than one pawn without the need for fractions.
Though we do have a direct correlation in the value of pieces and pawns, positional mistakes are harder for the human mind to measure using this metric for the game of chess. This is more in line with how computers analyze chess positions than our conventional understanding of the play.
Correlation between ACPL and ELO
At all levels of play, the average centipawn loss keeps decreasing as players get better.
Grandmasters and super-Grandmasters are the top human players in the world, hence an average centipawn loss of 20–10 is to be expected from them. The Norwegian GM won one of the twelve classical games in the 2018 World Chess Championship match between Magnus Carlsen and Fabiano Caruana, scoring an average of one point per game.
In a typical game, an international master player (between 2200 and 2500 ELO) may lose around 30 centipawns on average.
The average online player sees their ACPL stats between 50 and 100 frequently, with expert-level players (in the 1800–2200 range) going up to 50.
For novice players, this figure may even rise to 200 or more, indicating that they give away an average of more than two pieces’ worth of points for each move.
Logically, the graph demonstrates a negative correlation between rating and ACPL (ACPL declines as rating grows); when a player improves, they prefer to make movements that are closer to the ideal.
Can we use ACPL to predict cheating in chess?
Websites where you can make post-game analyses, and all mistakes, will be shown. If a player in every game had 0 inaccuracies, 0 errors, 0 blunders and games were longer than 30 moves — most probably this player cheated.
Conclusion
In human-versus-human games, a variety of several factors come into play, and just because a computer can predict the best move with absolute certainty, making some dubious decisions with the express intention of confusing your opponent and clouding the issue can be a wise move, particularly in low-time circumstances. (In truth, the players’ available time is a significant component in analyzing any human chess game but is not considered when calculating the average centipawn loss.)
It is a useful tool for identifying mistakes and serving as a starting point for your tactical and strategic development, but it is not the end-all-be-all of chess measurements.
It is wrong to cheat. In actual competitions, cheating is a serious offense. Cheating in casual internet chess matches is likewise wrong but harmless because there is not much at risk (other than pride). Some methods of cheating are simple to catch, while others are difficult.
References
- Chess Digits. Predicting rating from centipawn loss. Retrieved on 12/01/2022 from https://web.chessdigits.com/articles/predicting-rating-from-centipawn-loss
- Chess ACPL. https://dotesports.com/chess/news/what-does-average-centipawn-loss-mean-in-chess
- How to detect cheating in chess https://chess-bot.com/blog/how-to-detect-chess-cheat.html
For more on Hans Niemann:
For more on Stockfish:
Copyright 2022 Akhilesh Dongre, Jatin Madan, Sreya Vallabhaneni
Permission is hereby granted, free of charge, to any person obtaining a copy of this article and associated documentation files, to deal in the article without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the article, and to permit persons to whom the article is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the article.
THE ARTICLE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES, OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT, OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE ARTICLE OR THE USE OR OTHER DEALINGS IN THE ARTICLE.
