Some thoughts on anti-cheating systems

[NickFaulks]

If by competitively you mean FIDE rated

No, I mean that they've not played locally-rated games either, that though they may have played casual games with a physical board and pieces, maybe at school or with a friend, they're otherwise pure internet players.

Many nations do not even have a local rating system, and we know nothing about games which are not FIDE rated. My guess is that every player currently capable of achieving an IM norm has played a great deal of meaningful chess with board and pieces, but I could be wrong. Also, what is true now may well not be true in the future.

[JustinHorton]

My guess is that every player currently capable of achieving an IM norm has played a great deal of meaningful chess with board and pieces, but I could be wrong. Also, what is true now may well not be true in the future.

Indeed not. Which just means that assumptions we're making now, and procedures we're developing which incorporate those assumptions, may not hold in the future.

I'm guessing that the test of playing in a monitored environment against a strong player still ought to hold for individuals, but it's not really available in a lot of situations.

[Matthew Turner]

Justin, but if you have a history behind it, then you can just use the online rating. I don’t see this causing a problem.

[MartinCarpenter]

Depends slightly what you mean by a history.

My York club has moved its internal tournaments online. Everyone concerned is mostly just playing in that so obvious ‘desert Island’ problems which won’t ever correct themselves. Many of them are horribly under graded right now.

This sort of thing is probably rather widespread at the moment.

[Matthew Turner]

Martin,
We don’t know how Lichess systems work, but taking the Regan analysis as a base.
I guess you are about 2100. So you’ll establish an online rating playing at around this level, but because you at in this bubble you stay at 1500. You now start playing better and your move matching puts you at a 2300 level and your account gets flagged
This shouldn’t happen, but in the real world it could and that is unfortunate.

What would happen next? I presume you would E-mail Lichess and say you are Martin Carpenter, you are a 2100 and you haven’t been cheating. They would review the evidence, see that it was all consistent with a 2100 and reinstate your account.

Is that really a failure of the system?

[JustinHorton]

Well, we would want to see if that was what actually happened.

[Roger de Coverly]

Is that really a failure of the system?

The failure is making accusations on the premise that someone isn't good enough to play the moves actually played. Rather than conclude that the rating is unreliable, jump to the conclusion with no other evidence that a player is cheating in some manner.

Maybe they have other evidence, but their approach lacks credibility unless it is disclosed.

[Mick Norris]

My guess is that every player currently capable of achieving an IM norm has played a great deal of meaningful chess with board and pieces, but I could be wrong. Also, what is true now may well not be true in the future.

Indeed not. Which just means that assumptions we're making now, and procedures we're developing which incorporate those assumptions, may not hold in the future.

I'm guessing that the test of playing in a monitored environment against a strong player still ought to hold for individuals, but it's not really available in a lot of situations.

Presumably you could have a Zoom meeting with said strong player, and explain enough to show your level of understanding?

I can't get away from the basic fact that for me now, playing chess is only ok when it is enjoyable (so I don't play in the evenings when I am tired after working, but I don't mind driving on the occasional weekend to play county or 4NCL or a very occasional rapidplay or busy persons)

[Matthew Turner]

Just to be clear, I think that the circumstances Martin describes are quite extreme, but even in these circumstances I don’t believe that he would be flagged because whilst his move matching performance rating would be out of line with his rating, his move matching performance wouldn’t jump, it would presumably still be within tolerance of the established performance. What I am saying is that even in this extreme example the situation would end up being rectified relatively quickly. I am basing this on what Theo Wait (Lichess board member) said About the appeals process - he is extremely candid with how much information he puts forward so I don’t think he would have said it if it wasn’t true.

[JustinHorton]

I can't get away from the basic fact that for me now, playing chess is only ok when it is enjoyable

Not sure I'd ever play at all by that criterion

[MartinCarpenter]

They are extreme from the perspective of this sort of anti cheating, yes. They’re also, I guess, conditions that roughly pertain for quite a few of the people who have just moved online which is why I mentioned it.

The other thing that might make it slightly more complex is that the 4NCL/County online etc are at much longer time limits than internal club events. Some people are obviously much stronger at one than the other.

Should all be something they could handle sensibly, hard to be truly certain given what we’ve heard of the general experience of appeals. Obviously it’d be trivial for the 4NCL/ECF etc to check known ratings.

[Matthew Turner]

What we heard from Lichess is that the computer throws out suspicions about a player and then this is reviewed by a human moderator. On appeal the evidence is reviewed by a moderator not involved in the original case.
The moderators are volunteers to do this because they like chess and want to support the community. I am sure if you had a pot load of money you could devise a better system, but as the accused you have a hearing and appeals can be and are successful.

[Roger de Coverly]

The moderators are volunteers to do this because they like chess and want to support the community. I am sure if you had a pot load of money you could devise a better system, but as the accused you have a hearing and appeals can be and are successful.

There was a team that withdrew from online 4NCL division 1 because one of its squad was banned. Presumably if there was an appeal process, they weren't satisfied by it.

[Joseph Conlon]

There's a lot here (and thanks to Roger for his thoughtful starting post even though I broadly disagree). Some comments:

1. For all the talk of secrecy, lichess is open-source. The anti-cheating software is at https://github.com/clarkerubber/irwin together with sample screenshots of what it produces.

I'm not a programmer and not competent to understand it, but if anyone can go and look and take the time to understand it if they want, its not that secretive.

2. Strong anonymous accounts appear all the time on e.g lichess without getting banned, including sitting at the top of the rankings. Right now three of the top ten blitz accounts are anonymous (i.e. without an associated title, and so lichess has no reason to know who the player is). Very likely these are training accounts of strong GMs, but how would lichess know this? So the problem of 'How to tell that this brand new account playing at 2700 strength is human?' is one that lichess etc are dealing with all the time, apparently successfully.

3. In principle you would expect lichess/chess.com to have much better cheat detection systems than Ken Regan as they have far more data to work with (not least move times). i.e. if someone is cheating in a games, the moves by themselves would give you a certain amount of statistical significance, but moves plus move times should give you more.

[NickFaulks]

1. For all the talk of secrecy, lichess is open-source. The anti-cheating software is at https://github.com/clarkerubber/irwin

Can someone who understands these things confirm that Lichess really do reveal their anti-cheating methods here? If so, why do they demand such a ferocious NDA?