The Use Of Engines, Average Centipawn Loss And Online Cheating By Pureheart Loveday


    Being a mathematics enthusiast, the subject of centipawn loss is really of interest to me.

    Irrespective of the primary means of evaluating the strength of the two relative positions in a single game of chess viz. pawn structure, material, space and initiative, the most encompassing currency for evaluating any given position in modern times is what is referred to as the “average centipawn loss.”


    A centipawn is 1/100th of the material value of a single pawn; the unit of analytical currency to evaluate the relative strength between the move a player made and the best move available given the position they were given. Such that if a human player makes the best possible move for any position, they will lose 0 centipawns. The best move possible is simply the 1st choice move of the strongest engine like Stockfish 11 or 12. For example, if there is a mate in 10 somewhere and you took a route that leads to mate in 12, you lose some centipawn there. Similarly, if you play BxQ (taking a free Queen), and there was a QxP (sac) somewhere that can lead to mate but you didn’t play it, you equally lose some centipawn there.

    Nevertheless, average centipawn loss is the sum of centipawn loss divided by the total number of moves in the game vis-à-vis the arithmetical mean (average) of centipawn loss per move for any given game.


    The GMs of today are so accurate that their games could read around below 10 centipawn loss in classical time controls and 15 in rapid time controls (see: WC, Carlsen vs. Caruana, 2018), and even above 20 in blitz. It may interest you to note that WC matches between the years 1900 and 2000 were having average centipawn loss of 30 and even 40! Nonetheless, it is pertinent to note that “inaccuracies” and “mistakes” get labeled for moves that have 20-50 centipawn loss, and anything over 50 centipawn loss is a “blunder!”.

    So in a way, in classical time controls IMs and GMs are in the realms of +-20 average centipawn loss, whereas sGMs are between 5 and 15. But in blitz time controls, even Carlsen has never attained less 15! Even against a weak opponent!

    In retrospect, because of the prevalence and proliferation of the chess engines, chess players all over have become more accurate; becoming like the engines they create. However, even as human chess evolve in an ultra-modern feat, the engines equally evolve; we grow and they grow even faster! Consequently, the development of centipawn, a currency we can use to compare the relationship between our move and that of the strongest chess engines.


    Every player with his strength is relative to the average centipawn loss expected of them. You are rated 2200 for example, it is expected that you are to have an average centipawn loss of nothing less than 20. If you create a new account, the system imagines that you are 2200 and judge you at such, but as you grow over time, the system expects a lower centipawn from you.

    In other words, if you have an Elo of 2100 and your centipawn is 15 you will be flagged, without your knowledge and when the account is scrutinized, you will be blocked. However, if your Elo is 2900 and your centipawn is showing that same 15, you may be flagged but not blocked; when the system check and found out it is an, IM of FM and has found out the rating grew over time, it is expected to have low centipawn. A similar thing happened to IM Yarebore (rated 2900+ on lichess), who created a German account and started playing iron moves and then the account was blocked.

    So yes, you are rated 2150 and you play a whole game with an average centipawn loss 10, the account will be blocked so that you explain yourself. How you are able to do what WCs have never done in a 3 minutes game.

    I hope this post is expository enough.

    Gens Una Sumus


    1. I have just had a centipawn loss of 13 on lichess. Blitz, my rating is only 1737.
      I didn’t cheat. So was I just lucky ?


    Please enter your comment!
    Please enter your name here