Analysis of Lichess' cheating detection with Machine Learning (ML) - a mis-use of ML & doesn't work

@bufferunderrun said in #90:
>
Well me for example, I never cheated and I also don't have multiple accounts, neither did I spread misinformation neither did I harass any player, yet my account got banned, I appealed against such decision and I wasn't given any hearing, then I created a new account, and all of a sudden it gets shut down, I don't know what I did to deserve that either, so now I'm facing an IP ban, and I never for once cheated, so this is very annoying and depressing, if you accused me of a crime, shouldn't there be atleast tangible evidence to prove I committed the crime, so lichess can just come out randomly and ban whoever they want, whosoever they may be, whatsoever they may have done, this is most likely going to make me quit playing chess, so this is lichess definition of free and fair

We do not discuss individual cases in public or with third parties. Please see lichess.org/page/appeal

The problem is that you don't even discuss individual cases at all. Not even with the person concerned. You label real people without ever providing evidence or even proof. This is a key point of criticism from the author of this thread.

Because it doesn't work for Lichess (or any other site), for reasons stated many times.
Implicit hints at better cheating, and unproductive debates with cheaters who just can't accept they were caught.

@Deadban said in #85:
> Your "reliable source" is someone who was shadowbanned by Lichess for spreading misinformation. LOL, I'm done dude.

except I think they were not particularly wrong on this subject?

@Cedur216 said in #94:
> Because it doesn't work for Lichess (or any other site), for reasons stated many times.
> Implicit hints at better cheating, and unproductive debates with cheaters who just can't accept they were caught.
The usual blah blah

> except I think they were not particularly wrong on this subject?
When it comes to correspondence chess, I know I'm right.

#96 EOF

Indeed you are correct to a degree when it comes to correspondence. When you throw out move times and add in opening book, it becomes a lot more difficult to catch cheaters. Truth is, people can play incredibly strongly in correspondence as the limitations of timed chess do not apply, and without move times, you’ve thrown out an entire super useful method of detection.

Here's an update from lichess:
After reviewing the situation, we have re-opened @RealEmms, as it was closed in error. There is no IP ban. I hate lichess currently, only God knows how many errors have been made

Is there anything else relevant or is that it @IrwinCaladinResearch ?