Cheaters are killing lichess : a systemic problem

Watching lichess quite regularly and each time, almost every day, there is a cheating incident be it a strong player, often a titled player, loosing rating points to a newcomer or a tournament being dominated by an unknown player crushing everyone. Usually, these new players will be recognized a few hours later as cheaters and expelled from the rated games but it doesn't stop the flow : another game, another tournament, another cheater.

For the regular users, it's devastating : they come to this site to have fun playing chess, to have some friendly competition and maybe to contribute to the building of a small Internet community. Instead, the very game they came for is rigged by anonymous persons : they loose and loose again against machines. Playing in a tournament means, usually, that you have to wait a few hours after the tournament end in order to know the real winner.

The problem here is not the cheaters themselves but how cheating is systemic on lichess.org and how it became part of the user experience. For me, it's amazing that so many people could have some pleasure doing such a thing, lose their time doing it and maybe their money (since some cheating software are not free) but the question is not the immorality of the cheaters or their shady reasons for cheating and spoiling other people enjoyment. We have to confront the fact that cheating is systemic on lichess.org and hurting the game and the users.

(One explanation to the massive amount of cheaters should be addressed : it is possible that this constant flow of cheaters is organized and concerted)

The website is free : there is no premium, the most amazing features are free -- and that's good, but there is consequence to this gratuity : it is very easy to cheat. Almost nobody is going to pay for this kind of stupid and vain activity but if it takes just of few minutes and no money to have the pleasure of beating an IM, to win a tournament and being able to boast (and to have the short admiration of some naive followers), it's no wonder that every day, from the billions of people using Internet will emerge someone, and usually more than one, willing to cheat.

The community must do something to reduce this problem. We need a systemic answer to this systemic problem. We need new rules to prevent cheating, to prevent the spoiling of the user experience. Reducing the access to the tournaments could be a first step. Asking for a few casual games before being allowed to play rated ones could be another one. I don't know but one thing is clear : if nothing is done, the regulars users and the good players will leave lichess for another website who will have the courage to address this problem.

The problem is..., how to detect one someone is cheating and how to automate it. If we answer these questions satisfactorily, we will be able to fight against cheating.

A new comer is not necessarily a bad player but maybe just someone that has just discovered lichess!

Experience/reputation system is the solution.

You can register a new account in 30 seconds, but to build up experience and reputation takes a long time, a lot of activity on the site, and last but not least good behaviour to earn upvotes and not to earn downvotes.

In such a system a cheater will find that they only see seeks of provisional players or players with low reputation. They will soon find out that the whole thing is not that great fun anymore, because they can never get games against experienced and respected 2200+ players.

I know its frustrating, been there ! Personally though, I have encountered only 3 or so cheaters in the 1.500 or so games that I've played here. That makes 1 cheater for every 499 honest spars. The problem with automating cheater recognition is algorithmically someone should answer ethical questions. As in, should a driverless car choose to kill one pedestrian instead of %80 probability of killing two ? Therefore, i think the current human investigation before terminal decision, is necessary. Yes, it slows down detection, but compared to being called out a cheater due to a false-positive is much worse than 1 cheater in 500 games. Another issue is, we shouldn't rush in to call one a cheater. I am no great player but I have scored some 3-5 centi-pawn loss games here and there, some successively. So stuff is possible you know. We should employ some self restraint for the gray areas I think. In conclusion, I try not to allow that 1 in 500 to ruin my day for me. I try to minimize that frustration to a few seconds. And every now and then If its extremely apparent I go ahead and report, without deterministic language. After all I wasnt sitting next to them to know for sure.

Alphilinte #1

I've written about the cheater problem several times in the forum here.
For me the main problem is seeing cheaters in tourneys.
Sometimes even 2 or 3 cheaters in 1 tourney.
In regular games I could "block" those users if I wanted to, but in tourneys I cannot really avoid getting paired to cheaters (unless I notice them, and then I can choose to withdraw for a moment or for longer, or when getting paired, refuse to make a move).
The thing is also that it can take quite some tourney to do the compu analysis, to compare games analysis, and move times, look at the losses and draws, and ratings and ratings graphs of those cheater users, and then do the report.
When that happens during a tourney, the tourney is often over for myself, when I want to make a report. It takes a bit of time !

One solution that I see is that we collectively *c a r e*.
Meaning that if anyone suspected to have played against a cheater, the least thing you can do is to run a compu analysis (Stockfish) for your own very game.

That makes reporting in general much easier for anyone, and takes less time for those who care more about getting cheaters out.

For the rest, I would like to suggest that Lichess does automatic checking of the first 10 games of a user that joins a tourney for the very first time.

And since the daily and weekly and such tourneys seem to attract more cheaters, Lichess could give that more attention as well.

And cheat reports done in tourneys could perhaps go first in the reports lists for the moderators ?
And if one suspected player had 5 reports, this could go higher in the reports to do list ?

@GoMentalGoBloody #5

That you don't see many cheaters, doesn't mean that cheaters are not so much here.

In the last few weeks in the classical top10 at least 4 (!!) cheaters were caught thanks to the efforts of those who care to check their games for cheating.

You should have to earn the right to play in tourneys. Why is every fool allowed to play in tourneys? By fool of course I don't mean low rated players. But those who see this as the golden opportunity get several forced games in a row which they otherwise don't get and then misuse this opportunity.

Proposed solution:

1. Do not allow players with provisional rating to participate in tournaments.

This is also useful for another reason: players with low 1500 rating are paired with low-rating players. So there is another door for cheating: get provisional rating just to be paired with weak players and easily win tournament.

2. May be additionally restrict tournaments (because in tourneys one cannot really avoid getting paired to cheaters). Say, player must have 1 week from getting non-provisional rating.

I just want to add;
Before reporting someone, I intensively analyse their games (not only the one we've played)
If a 1700 with min 60 centipawn loss games suddenly strikes out a 10+ game winning streak with 3 centipawn loss in every game, ok I say something is fishy. But apart from such obvious cases, I mean "doubt" is like an enveloping shadow. I don't believe it is correct to rush such decisions, let alone automate.

What we do have is data. Lichess can calculate "expected move quality" for rating classes (for all moves of players with statistically significant games that are stuck between a rating range).

If the likelihood of a player's playing genuinely (calculated from the discrepancy of quality of moves for his rating range if it is stabilized ) falls below a certain % (analysed over all of their moves in all of their games), we may alert people by changing the color of the username to others. Then say lichess can allow entrance of certain shades of green to the tourneys.

This could solve the problem of existing cheaters. For new cheaters, we can change the account setup process to include some tactics and strategy puzzles with time controls, to establish an expected rating of the new user. (Yet still letting them start at the default entry rating) Then we can include new accounts into supervision as well.

However, such an implentation should at first calculate all such data and probabilities and check them against confirmed cheaters. If this shows great promise, all nice and good, if not all I've said should be scratched.

:P Im writing this as if I have an army of PhD students at my disposal. Until that day or somehow the idea being embraced by many, I advise restraint on cheater-calling and not letting it ruin your mood for more than a few secs.