How Much Accuracy Means Cheating

How much accuracy means cheating? I saw players accusing others about using engines at 90+ so I wanted to ask.
At a rapid game

Depends on the position. If it's obvious piece captures and take advantage of huge blunders, then it doesn't prove cheating. Where it does however would be in a long very confusing game where finding the right move would be nearly impossible for any human player. I'd say in a 40+ move game if the accuracy is over 90% centipawn loss under 20 you can have enough reasonable suspicion to report the game for mods to review it.

the main problem with this is, that lichess analysis is goofy in a way

consider this my rapid game lichess.org/5ESx53FP/black where i had accuracy 93% ..and i am about 1200 fide elo.
The problem is, that chessbase 16 (i ran stockfish on it too) gives me only 49% engine correlation! You can also just turn on the engine on lichess...and you will see, that i didnt played best moves...and still had 93% accuracy!



As was pointed out on different thread, on lichess page lichess.org/page/accuracy explaining, what accuracy on this site actually means, you can understand, that it really can't be used for cheating detection. You can play as bad as you want and as long as you dont considerably lower your practical probability of winning, lichess is not considering your moves as poor.

It all depends on your opponent. If he/she plays badly enough, then every move is winning and then every normal/average move of yours is considered accurate, or lets say "accurate enough to still win". It doesnt have to be accurate in the sense of "best move" is.

I usually report when the following three things happen at the same time.

1) I feel strange during the game, like if it feels like the opponent is playing very fast yet making very profound moves, without being much higher rated than me

2) they get somewhere below 2 0 0

3) they get acl below 15

And in most cases the reports do indeed result in a ban, so the method is effective. However, I jave to say that it also often happens that players whom I did not suspect turn put to he cheaters as well.

High accuracy does not necessarily mean cheating. Even low rated players get high accuracy scores now and then.

Unlike some others, I never report anyone for cheating (I do report people for other things, though). I do not think I am expert enough to tell a difference between a false positive and an actual cheater, and when choosing between falsely accusing someone and letting an occasional cheater go unpunished, I choose the latter.

Cheating is, mostly, not determined on the basis of accuracy.

Accuracy and ACPL can be used to detect cheaters, but only the most flagrant cheating and requires enough games to do.

A better way to detect cheating is to find some way to model the difficulty of a move to find. The greater the tightrope you have to walk to play a move the less likely a human would be able to play it. If. a player plays many deeply profound moves in a game chances are they are a cheater.

I imported two tournament games of mine, 94% and 95% is feasible. :D

@Sarg0n thats otb tho