Cheating in chess

[Matthew Turner]

John,
There is no cheat detection system (to my knowledge) which takes account of the opponent's rating.
Lichess and Chess.com would not normally know your OTB rating so that wouldn't be a factor (except they would check out the OTB rating before banning a titled player).
In short ratings are really not important in Lichess and chess.com systems.
If your game has a near 100% correlation with Stockfish then obviously at some point you are going to get flagged up, but you'll have a few games to wait unless they have other evidence against you.

[Jonathan Bryant]

I'm just glad that lichess was open minded about considering all of these and not just the algorithms...

Except they weren't. At first anyway.

[MartinCarpenter]

Does anyone know whether lichess and chess.com use their own rating systems to determine whether someone has played above their grade, or even if they take rating into consideration at all?

They do yes. They try to evaluate the expected strength of your moves given the grade.

Then assign the relative chances of you have played that well by chance vs having used external computer help.

Nearly all of my opponents at longer time controls have been graded between 180 and 210 ECF but their online ratings have been 1500-1800, which means there will be many seriously under-rated players when these ratings stop being provisional. Does this increase the risk of being called out for cheating?

For those people who are hugely underrated? Yes, I think it must.

Especially if they've actually played a bunch of games against each other in closed groups like club rapid plays etc, so there's a decent amount of 'evidence' backing the grades.

If someone then starts playing more generally and gets a huge grading uplift? I dunno. It might be theoretically possible to notice this and compensate for it but this isn't a situation that the chess sites will have faced before, and I'd not be amazingly confident.

[Matthew Turner]

Does anyone know whether lichess and chess.com use their own rating systems to determine whether someone has played above their grade, or even if they take rating into consideration at all?

They do yes. They try to evaluate the expected strength of your moves given the grade.

Then assign the relative chances of you have played that well by chance vs having used external computer help.

You can test out whether this is true. Simply set up an account and play ten games awfully, then play normally and see if you get flagged.

[JustinHorton]

John,
There is no cheat detection system (to my knowledge) which takes account of the opponent's rating.
Lichess and Chess.com would not normally know your OTB rating so that wouldn't be a factor (except they would check out the OTB rating before banning a titled player).
In short ratings are really not important in Lichess and chess.com systems.
If your game has a near 100% correlation with Stockfish then obviously at some point you are going to get flagged up, but you'll have a few games to wait unless they have other evidence against you.

When I was appealing against being barred by chess.com they specifically suggested I tell them my rating (not that the appeal waa anything other than a joke).

I do not see how they could possibly make judgements without, at some level, taking a view of your playing strength.

[Wadih Khoury]

Taking a view of playing strength is a double edged sword.
It may work for established adults, but is a nightmare for juniors.

Even adjusting for the fact they are juniors does not necessarily work. How much do you adjust by? 50? 100? 400?

I see routinely the best U10 U12s beat OtB players ranked 500+ higher than they are. And I'd doubt that algorithms adjust that much, otherwise they'd have a more false negatives for the average juniors that need less of an adjustment.

[Matthew Turner]

With the rather major caveat that we don't know how Lichess and chess.com systems work! I think you have to realise that there is an awful lot of data available to these providers, so it is a bit misleading to think about ratings as static data points. If you think about playing strength as defined by characteristics of play with observable trends and evolution I think you get a better feel for how it (maybe) all works.
Corollaries of this are
1. Rating are not so important
2. A 1200 junior beating an 1800 can become much more explainable and expected.

[Roger de Coverly]

If you think about playing strength as defined by characteristics of play with observable trends and evolution I think you get a better feel for how it (maybe) all works.

At least some of the accounts set up for 4NCL and ECF events through lichess and chess.com have been new especially for the events. Being a new account doesn't seem to prevent a ban after a handful of games.

[Matthew Turner]

Roger,
Let me give a silly example,
You have a player who plays 60 moves in 2 games and 80% match stockfish' first choice
Would you flag them?
What if four moves match the first choice and then one doesn't and this is repeated throughout the 60 moves would you flag them now?
If can see that each move that matches stockfish takes between 20 and 25 seconds, other moves average three and a half minutes. Would you flag them now?
If you can see that the player switches screen on each stockfish move, but not or the others
Would you flag them now?

I am just trying to show that you can very quickly accumulate a lot of information on which to a judgment.

[Roger de Coverly]

If you can see that the player switches screen on each stockfish move, but not or the others

I'm aware that ICC's specialist software of twenty years ago was allegedly capable of spying on its user. But can something running on a server using a browser observe what else is going on at the user's computer.?

[Colin Purdon]

I'm aware that ICC's specialist software of twenty years ago was allegedly capable of spying on its user. But can something running on a server using a browser observe what else is going on at the user's computer.?

No, or at least not normally.

Browsers are allowed to send information like what OS you are using, browser version, screen resolution and similar but not information about what else is running on your device. However, if you have an old browser that supports ActiveX (usually that would be IE) and weak security settings then it would be possible to acquire information about other running processes. With any reasonably modern browser (including IE or Edge) with default settings this is not possible.

If a cheater is using a server-based engine on the same server on which he or she is playing the game, then that (unsurprisingly) would be detectable.

[Jacob Ward]

If you can see that the player switches screen on each stockfish move, but not or the others

I'm aware that ICC's specialist software of twenty years ago was allegedly capable of spying on its user. But can something running on a server using a browser observe what else is going on at the user's computer.?

No - but it can tell (if the browser tells it, which most browsers will by default) whether the chess.com browser window currently has 'focus' (i.e. is the window the user is currently active in). So it can potentially spot switching from window to window - although as with everything, a more cunning cheater could disguise it.

[JustinHorton]

Roger,
Let me give a silly example,
You have a player who plays 60 moves in 2 games and 80% match stockfish' first choice
Would you flag them?

You'd absolutely need to see the moves before even thinking about it, wouldn't you?

[Matthew Turner]

Roger,
Let me give a silly example,
You have a player who plays 60 moves in 2 games and 80% match stockfish' first choice
Would you flag them?

You'd absolutely need to see the moves before even thinking about it, wouldn't you?

Justin,
You are meant to carry on reading my post, not stop at that point!
I think given the metrics that I outlined most people would conclude that the player was cheating. It is interesting to note
1. It is very few games
2. I haven't used rating at all (so the same conclusion would be reached with Magnus Carlsen or a beginner)
3. As you rightly point out I haven't looked at the moves at all

I hope it is food for thought.

[Matthew Turner]

Roger,
Let me give a silly example,
You have a player who plays 60 moves in 2 games and 80% match stockfish' first choice
Would you flag them?

You'd absolutely need to see the moves before even thinking about it, wouldn't you?

Justin,
There might be another useful way of answering your question/thinking about it
1. If you were a person at Lichess responsible for issuing a flagging, not necessarily
2. If you were a person responsible for hearing an appeal, then yes I would have thought absolutely.