Lichess must fight cheaters

i) Lichess can automate the procedure

It already does. It marks up people as cheaters in the majority of cases.

ii) compare average inaccuracies, mistakes, blunders with recent games

An inaccurate way of doing it. For example, this game: http://en.lichess.org/pp7fqeeW I have 0 0 0 with 9 centipawn loss, but we can all agree that no cheating was involved there. However, because I predominantly play bullet, that game is probably by far an outlier.

A better way would be to detect consistent patterns of low centipawn average, or 000, which is already done.

iii) Analyse time per move

Already done.

iv) Analyse if errors begin after a certain point

Not an efficient method of detecting cheaters - the player may just have memorised the first 5, 10, 15 moves in a particular line.

v) analyse errors before a particular time

Not an efficient method of detecting cheaters; maybe he's awful at openings, but great at tactics or endgames.

---

Just keep reporting cheating accounts. Moderators can see a lot more about a player than you can. I'd agree that consistent games of 000 (or 100) and low centipawn loss is suspicious, so PM me the account names you suspect, and I'll check it out - THIS IS ONLY FOR THIS THREAD, NOT IN GENERAL.

Also, how can you help lichess - if you think a player is cheating, report them, and then analyse 10 - 15 of their games. If they are cheating, the system will automatically pick it up and auto-ban them.

All that stuff is already implemented. It's been there for years actually, and it continually gets improvements.

~70% of reports are generated automatically.

~40% of all moderator actions (marking engines/boosters/trolls) is done automatically. But that proportion is probably greater if you focus on cheaters alone.

#9 I tend to spread a few numbers in the forum from time to time, last 2 days there was 53 accounts marked as engine; 28 of them automatically by the system. During the same period there was about 2300 new accounts created.

@Cynosure
Average centipawn loss or nbr of inaccuracies, mistakes, blunders can mathematically help if the average is comparing not a single game (as per your example) but the average of let's say 5-10 games. If your example was repeated more than 5 times in a row or within 10 games, it could raise an alarm.

The same could be done by tracing early and end game performance, comparing not single games but a set of last played games. An awful player at openings could not suddenly play as grandmaster in 5 consecutive games for example.

All this is food for thought for making Lichess an even better chess site, with less cheaters bugging it. The increased anti-cheat effort can advertise the site for inviting even more players to register here.

DarthNick, one has to do some kind of move-matching with computers. You can't base cheating detection on such vague feelings. Also the comps are very strong and never play a really bad move, so they don't play awful openings and then "like a Grandmaster.

I agree detecting cheaters is vital, but I think the work it would require to implement those ideas successfully and without a myriad of false positives (ie, inconsistent players, people experimenting with new openings) would create far more of a headache than other methods of detecting cheaters.

Another cheater busted and I am going to PM me you info Cynosure. I think it is worth the effort to try more to prosecute these losers, or else the site will become as chess.com...

You should use the report button and not contact us by private message. Our inbox is not a report list.
Exceptions are welcome, but the report button is what people should use and what we pay attention to.

Well I have done this multiple times already for the same and other users + on post #11 Cynosure said he could receive a PM.
It is tiring for all of us this procedure. Simplest thing for myself is to stop playing here but I like the site and want to help it.