Tuesday 5 May 2020

There is (some) truth in numbers

Just as last month everyone was some sort of epidemiologist, this month, everyone (in the chess world) has become an expert at online cheating. At the moment I am seeing a tasty mix of  "why don't they ban obvious cheaters"/"why did they unfairly ban my friend" posts, which I am sure will continue for the next few months.
While I have my own strong opinions the topic, I'd rather share an amusing observation from one of the online tournaments I help organise.
One of the obvious metrics for suspicion (apart from the always popular "he was rated 400 points below me"), is how close a players move match the top engine choice. This is closely followed by "average centi-pawn loss" per move. And if a lowly rated player scores too high, then "there is something afoot at the Circle K". In a first round game (where top seeds meet lower rated players), both players scored on 90% on computer first moves. The winner (or was actually the higher rated opponent) made no errors at all, while the loser only made 1 mistake, and a single blunder. So while you may think that it would be difficult for both players to play an almost perfect 32 move game of chess, it is less surprising when you know that the one really bad move was to drop the queen on move 14! 
In fact one of the things I try and do after each tournament is look through game (including upsets), to see if I can find something for my tournament stream. And so far, while I have found a number of well played games, I have found no evidence that an engine was involved in any of them.

No comments: