dkappe wrote: ↑Fri Sep 09, 2022 6:04 pm
CornfedForever wrote: ↑Fri Sep 09, 2022 3:25 pm
pocpit wrote: ↑Fri Sep 09, 2022 1:43 pm
PS: about false positives. The other day a guy with whom I play OTB told me about another guy being banned from lichess. Say player A (Elo 1800) and player B (elo 2400) know each other (they live in the same, small city). Player A knows that player B plays a certain Sicilian line, so player A prepares against this line and beats player B online. Player B reports player A, and player A is almost automatically banned: no questions, no explanations asked.
Well, that's not exactly the way it would work... one would never get banned in that situation - maybe not even looked into unless the people hosting the playing platform were total dolts - probably only if a cursory algorithm 'soft flagged' that person as a player to look into pending future results.
But yes, false positives (Firouzja for example) are a problem - but only to a degree as generally the disease is worse than the cure. Heck, I've had enough opponents be banned that I not longer play 'daily' chess anywhere and I gave up my (true) 'correspondence career' about 15 yrs ago). One in an online 'daily' tournament was....lets just call him a very highly rated Russian. He beat me in our game and I thought nothing of it (except that I really should learn the opening a bit better!), I noticed later that his account had been restored. Algorithms can only 'point towards a possible truth'.
Lichess isn’t very good at detecting computer cheats. I run two BOT’s (official, designated as such) on lichess. They are both small leela-style nets running at very low nodes. Against a human, they do very well, but an AB engine goes through them like a hot knife through butter.
I used to review the games they lost. Most often the pattern was that a weak player would lose a bunch of games, then break out stockfish to “teach that BOT a lesson.” (Frankly, the difference in play is quite obvious.) Then I’d report them. Invariably lichess would uphold my claim.
But life’s too short. I stopped reviewing the games and lichess didn’t pick them up on its own. As a consequence, the ratings of the bots fell by about 150.
I think your post gives a wrong impression. IMO the cheater detection at Lichess is superior to chess.com and chess.com doesn't care if you
report a cheater with a non paid account anyway.
The difference is that obviously Lichess OTH only checks Human vs. Human games and also only rated games (of course).
This might be suboptimal for chess bots, but the whole platform was founded for Humans!
Your experience with chess bots has no meaning in this regard and I have surely played 10000s of games in the last ten years there,
with ratings mostly between 2100-2300. Ofc I only play no inc games and the slowest I play is 3+0 because of all that cheating to minimize
the effect.
I guess I had pointed out to hundreds of cheaters over a decade there, but after my last return I noticed the cheater rate
seemed to have a major decrease. At some time 1/10 - 1/20 games (in a rating region not under 2100, except in daily/hourly tournaments)
were cheated.
Regrettably I even personally know some cheaters during the corona leagues (DSOL) and it was even for me surprising what people started cheating.
Even some 70+ aged ones. There is an article in a German chess online magazine, which did stats on a full such corona season over all leagues
until the top. And their good message was that nearly 90% of the games looked clean. But each and every league had several bans despite
the 'Horvig-Bot' scam bot posters et al, who spread internet cheating since several years and seem to allow more and more 'clever cheating'.
Anyhow the bad news then is that around 10% of all those games were most likely cheat games despite no money (IIRC) was involved!
https://perlenvombodensee.de/2021/04/11 ... elen-fair/
