I read this funny quotation ... you do not know what happen to blitz on Lichess.
"Why dont you blame yourself for your loss? You lose to real people which is 99% likely, cheaters are less than 1%. Cry it or learn and improve it, it is up to you."
I play many games and I found that the most cheaters is in blitz 5+3. I think there is 20%-10%. That is not 1% - 1% was two years ago maybe.
They play tons of wild/normal openings with EXTREME PRECISION and plays MIDDLE GAMES like masters.
Problem is that they have:
- low puzzle rating (about 1400 or less often)
- low classical skill (about 1500)
It is IMPOSSIBLE to play blitz with 1450 like master and have such low ELO in puzzles/classic games :):):)
I have 1900 ELO in puzzles and 1800 in classic and I can not win with CHEATERS ARMY and their TRASH OPENINGS :)
I can tell you if I play/train with engine and with which - but why to the HELL I have to train with ENGINES on lichess - no idea.
I like very much training with engines but not when I choose game with HUMAN :)
I think it is very easy detect cheaters in openings but it need to have some data to train neural network. If it will be available someone can train to implement.
"Presumably because this would require Lichess to have engine analysis running on every move of every game being played simultaneously. This is likely prohibitively costly to implement." that is bullshit - we can use post game analysis and openings books plus some "deep learning magic" - it is not magic if you understand how :)
Can we have access to games with marked cheaters to train model to detect cheaters?
For this is need PGN with times and moves and who is cheater?
Can you share this data? Cheaters can be marked as anonymous and cheater to make data anonymous.
@LaserGuy @drmrboss @bufferunderrun