I run two neural network bots that are relatively weak on purpose. They do fairly well against humans, but stockfish goes through them like a hot knife through butter.
Now it was pretty easy to tell who the cheaters were, and I reported them dutifully. I suggested to the lichess folks that they might want to check the bot’s losses as a matter of course. Their response: “nah, we got this. Thanks but no thanks.” Since that time I’ve stopped reporting cheaters. Lichess cheat detection rarely flags cheaters and the bot’s blitz rating has gone down 80 elo.
So, maybe rethink how good your cheat detection is?
I don't understand. Did you cheat to catch cheaters? Doesn't make a lot of sense.
Or did you create an official lichess bot? In that case, what makes you so sure they cheated? Also, why would someone cheat in an unrated game against a bot? Seems pretty odd.
They are official bot accounts. How to tell a cheater? If a 1300 player wins and corresponds 100% with stockfish’s move choice over a 60 move game, it’s suspicious. When lichess concurs after reporting the game, then it’s good enough for me.
Cheating is prohibited regardless of whether the game is rated or not, and those who cheat in unrated games appear to also cheat in rated games (the loss of elo should have been a clue that not all games are unrated). Have you had any experiences with cheating in unrated games?
<Comment deleted by user>
I highly doubt that Lichess responded saying "thanks but no thanks". Catching and marking cheaters often takes a long time and a lot of data, usually it won't be done because of one game.
Disclaimer, I am not a moderator and have no inside information about Lichess' cheat detection.
@Alzheimerhead I don't know if 30% of registered users cheat, but if they do then most of them sure don't last long.
I think the problem here may be that people might be more likely to cheat against a bot since "it's only a bot" or maybe "it's also a computer after all" not realizing that this is still cheating and a bannable offense. So probably they don't cheat against other humans, just against that bot in one or two games. Which then presumably makes it hard to ban them since one or two games as noted above might not be sufficient evidence. (maybe this could be reduced by having some kind of box when challenging a bot that says "Using engine assistance against a BOT still is cheating and you will can be banned for it!" that you have to click, to make sure people know)
Though I'd also note that just because someone beat one of the nets might not be 100% cheating. The nets being rather small means that they have some blind spots so you can just get lucky. Though of course if someone with a way lower rating beats it, especially in multiple different ways, then that is suspicious.
A guy I introduced to Lichess got his account flagged for cheating, even though he didn't cheat. At least that's what he says and I don't see any reason to doubt him, not like he was playing with no errors or anything, and I doubt he's the type to install cheating software and such. I've also analyzed all games I played with him and it's not like he played flawless with no mistakes and blunders. Anyway point is, the cheat detection isn't always accurate and sometimes flags non-cheaters. Aside from this being a serious allegation to get accused of cheating like that (basically destroys your reputation online as far as chess is concerned, and perhaps other sports as well; your reputation as an honest player is basically tarnished after that), what bothers me is that he couldn't get his account reinstated as a normal account so that he could play rated games and so on. It's one thing if he actually cheated, but not much indicates that he did.