Would be Tal be flaged as cheater on Lichess?

Many people here claims and even report "perfect game" players as a cheater.
Regarding that, simple question is: "Would be Tal be marked as a cheater on Lichess?". Based particlulary on this game and assuming no body know about him:



1 Inaccuracies
0 Mistakes
0 Blunders
8 Average centipawn loss

It is failrly easy to play such "perfect" game against beginners and players who blunders - all "right" moves are simply forced up to the checkmate or resign. However, I have impression that many people here doesn't consider that as truth.

There are millions of other points of metadata. It’s not about ACPL alone.

@Sarg0n : Probably the first post here that I really don’t understand.
What is it about this metadata (I know what metadata is)?
And what is ACPL?
Thanks for any hint and advice.
Best regards

Magnus bullet game:
2 Inaccuracies
0 Mistakes
0 Blunders
7 Average centipawn loss

@Makropoulos #3

Sarg0n obviously do not understand what term "Metadata" actually means, so do not bother yourself to understand this comment - just ignore it.

ACPL is explained. It is actually commonly in use ACL. But I did not refered on ACL, it is anyway useless info (as I stated here many times). It is just about:

- Inaccuracies
- Mistakes
- Blunders

Inaccuracies compared to "perfect" Stockfish move actually can be taken with reserve, but mistakes and especially blunders certainly not.

Main question is: when and on what info Lichess "perfect anti-fraud algorithms" decide that player "is a cheater"? On 0/0/0, ?/0/0...

@AngelaRoss : Thanks Angy. Yet another abbreviation I missed.

Some links for the uninitiated:

en.m.wikipedia.org/wiki/Metadata
en.m.wikipedia.org/wiki/Sidecar_file
en.m.wikipedia.org/wiki/Chess_piece_relative_value

I couldn’t resist

Lichess is open source; including the cheat detection code. This has been said many times in the forum.

From github.com/clarkerubber/irwin

"Similar to cheatnet, it works on a similar concept of analysing the available PVs of a game to determine the odds of cheating occurring."

"main.py covers accessing the lichess API (modules/Api.py) via Env to get player data; pulling records from mongodb, analysing games using stockfish, assessing those games using tensorflow and then posting the final assessments."

It doesn't matter what criteria people use in reporting someone because you don't get marked as a cheater because someone reports you as such. They can report you based on any criteria they want; valid or not. Excessive invalid reporting is probably punished because it would cause much work by those involved in handling the reports.

Nothing else is going to be said about how cheat detection works for obvious reasons.

I had one long game with 8 ACPL in 3 min blitz.

Average ACPL matters.


My average ACPL in blitz is 50, Magnus may be 25, other GM may be 35.