Suggested Cheat Detection Metric

I've been playing a lot of 'classical' on here the past couple of weeks, and sometimes you just know you're playing against a computer cheater based on their move timing, yet they're also not making the absolute best move 100% of the time, presumably in an attempt to avoid automatic cheater detection.

My anecdotal experience with these players is that they are extremely reticent to allow the evaluation to flip from positive to negative territory on themselves. If they are playing white and are at, say, +.7, then they may be willing to make moves that put them down in the +.3, or +.4 range. They are almost never willing to go from +.1 to -.1.

I would aggregate data on these moments specifically, where the evaluation actually -flips-. Players with a significantly low propensity to transition from a small winning edge to small losing edge should draw the eye of cheat detection.

Or maybe this is already one of the metrics used?

I'm just making a suggestion for possible improvement to the -automated- cheat detection system.


Better than that?

Modern AI resemble self-learning „expert systems“.

Read this:

„Instead these systems are based more on probability. Big websites have millions of games to analyse - including games with confirmed GMs to learn from. A fraud detection algorithm can consider unlimited questions like:

If black has a 3 pawn lead but lots of unguarded pieces and down a knight, how likely are they to aggressively recapture instead of defend?
If black has been playing aggressively, how likely are they to play a very difficult to calculate defensive move?
When black has 15 available moves, 10 of which are pretty good, how long will it take for black to play a move?
Black has played 20 straight GM level moves and has a strong material lead. What are the odds that black will now play an amateur blunder? A "look I'm not cheating" move?
Is black suddenly playing like a GM whenever there's one really important move?
Does black suddenly play very well mid game if they are slightly behind? How likely is a human to do that?
Is black playing hard to calculate moves just as fast as easy to calculate recaptures?“ ...

I don't understand what you're saying. Are you saying something like . . . 'cheat detection isn't just looking at the acp, it's about finer decision-making indicators'?

"Players with a significantly low propensity to transition from a small winning edge to small losing edge should draw the eye of cheat detection." Significantly, as in statistically significant. This is prima facie "based on probability." It is the suggestion of a single metric to be added, not to replace the whole current system with my one suggestion.

#5 Machine learning derives this and trillions of other metrics, and uses those with the greatest predictive accuracy.

GAZILLIONS. @lazerblood69 No, bro in all fairness why are you making out of the blue suggestions without ever having seen Irwin used? To my knowledge it works fine.

I thought that out of the blue suggestions was pretty much the point of this forum. I imagine that it does work 'fine' but I am just suggesting that I think this is a good metric to explore to potentially make it work -better-. Is this somehow controversial? Is Lichess' build completely unsupervised?

This topic has been archived and can no longer be replied to.