intelligent cheating?

All I'm saying is there is chicken and egg problem. To train an AI to detect cheating you need to feed it tons of clean data. you can't generate that data with the AI and even if you could it would cause overfitting. Effectively you have to bootstrap it with data generated from games reviewed with human intuition which won't provide clean data and thus let clever cheaters win.

You might be able to partition things with k-means algorithm, i wonder how effective that is

I am learning (1100ish) and so I am studying theoretical lines and general concepts of different openings and strategies. Most of the GMs out there have videos and tricks on how to play the lines best (youtube is a great resource). I learn a lot of solid moves and I make mistakes along the way either through lack of memory or just get baited into some other line. The issue I have playing people rated my level is that they don't know the theoretical ways to play against my opening so they play random stuff that I can't make sense of what they are doing (many times they start destroying me). Now and then I play someone who plays higher rated and follows the lines the way the GMs tell you they are going to be played. I then get to use my GM trick I learned and beat a higher rated player. I may make a few mistakes, but generally I have learned a gambit that gives me enough of an advantage to sneak out a win now and then. I don't cheat and maybe they aren't cheating either.

Of course it exists, it's extremely hard to spot it let alone prove it. I would venture to say it's almost impossible to have a definite proof if cheaters conducts his deed in a very careful and intelligent way.

I'm not going to promulgate the ways one could convey such misdemeanour because that'd only play into the hands of cheaters, but I definitely want to bring home to you that such things do happen, probably more often than you think, especially at higher echelons.

Sure it is hard to spot that‘s why there’s a algorithm trained with millions of games. It sees patterns where humans don’t. It is beyond such deceiving techniques mentioned in posting #2. A move-by-move evaluation might be to ineffective though; an AI which creates a general „impression“ of the given data is the better approach.

I recommend to contact the admins or the official stuff. Recently I saw data from this hobby cheating-hunter groups here. They produce nice tables, alas worthless if they produce false-positives. I think they have some work to do if they cannot differentiate between good play and cheating. As mentioned above just comparing single moves with Stockfish won’t work (time, complexity of the position, alternative moves, hidden patterns and so on)

No bad idea either to make the RL name mandatory in official championships.

FredtheCrusher

Hi

I agree with you. Many cheaters are clever.

Many don't know and report, so cheaters are not identified.

also common in slower games is people playing the opening from a physical opening book, that sort of cheating can never be caught... also one cud play some 2 pawn gambit( analysis will already see that as 2 mistakes) then u develop n attack with computer moves and when u +8 in eval make sum suboptimal like teh 7th best moves in position for a few moves so eval goes down to +3 or so, comp thinks u blundered in analysis, then u just winh teh game from there, voila, easy cheating and never get caught

#18, is as easy as eating cherry-pie. You mean with some simple home-made tricks you can cheat long-term? People, people!
For example, this example given above has a steady trend, no turning points. This is one of many patterns.

PS: countless users have been banned because they regarded themselves „intelligent cheaters“.

#18 I know a way you can never get caught... get good at chess and play moves without cheating. It even works in real-life tournaments.