Some thoughts on anti-cheating systems

[Roger de Coverly]

It would be impossible for me using Mathews machine to access my copy of Rybka
which is installed on my computer.

I think they figured that out years ago. They spy on your computer to make sure you aren't switching to Rybka. Actually looking at your own Rybka is not even necessary if they supply real time computer analysis whilst the game is in progress. Again I'd expect they've got that one partly covered.

[Mick Norris]

I know it’s nothing to do with the topic, but was the opening post the longest ever on the EC forum?

I hope so, and let's hope it keeps the record too

It was obviously well though out though, which can't be said for all opening posts

[Matthew Turner]

Genuine question - which bit? I was under the impression at least one person had got banned based purely on 2(?),4(?) 4ncl online games.

There were players banned after playing just a handful of rounds in the 4NCL league. As neither lichess nor the 4NCL will publish much of the reasons for the bans, it's open season for speculation. It could have been something as straightforward as a spectator with the same IP address following the game. Given that both lichess and chess.com appear to supply the very obvious security risk of live analysis whilst the game is in progress, that could be an explanation.

Martin,
The second bit

Roger,
A player will only be flagged if they play abnormally well. Someone accessing analysis from the same address alone would not trigger a ban in itself

[Roger Lancaster]

I don’t think they can be using 5 sigma right now, else no one in the 4NCL on line would be that near getting banned (on moves alone.).

Well, like a number of other forumites, I have just come off a 2 hour conference call, and that is simply not true.

Having been on the same conference call I can confirm that, unless you believe [and, to be clear, I don't] that someone is fabricating the statistical data about 4NCL performances, then Matt is correct. If my memory serves me right, the peak performance was somewhere around 5.6 sigma,

[NickFaulks]

Because in the case of Ken Regan’s software it Is publicly available information.

Can you provide a link?

[Roger de Coverly]

A player will only be flagged if they play abnormally well.

The offence of playing well whilst not in possession of a high enough rating?

[Matthew Turner]

Roger,
The way that the FIDE rating system works is that each one of z equates to 200 points. So if you you have a z score of 4 you are playing at 800 points above your rating (adjustments are made for players with very low rating). There are zero players who have managed this in the 4NCL. There are a significant number of players who are managing this in the online 4NCL. How would you explain that?

[Roger de Coverly]

How would you explain that?

Let's clarify. You are saying that if someone plays well above their nominal rating they are cheating?

That's placing a lot of faith in the reliability of ratings. Which ones are being used? Someone playing on lichess might have FIDE standard, FIDE rapid, FIDE Blitz, ECF standard, ECF rapid, lichess Rapid, lichess Blitz, lichess Bullet. They may all be inconsistent.

[Matthew Turner]

Same players same ratings used. One event zero players with z score of 4 (as all the historic and mathematical evidence would predict) the other event a significant number of players with a z score of 4.

[David Sedgwick]

I don’t think they can be using 5 sigma right now, else no one in the 4NCL on line would be that near getting banned (on moves alone.).

Well, like a number of other forumites, I have just come off a 2 hour conference call, and that is simply not true.

Having been on the same conference call I can confirm that, unless you believe [and, to be clear, I don't] that someone is fabricating the statistical data about 4NCL performances, then Matt is correct. If my memory serves me right, the peak performance was somewhere around 5.6 sigma,

If my memory serves me right, the sigmas at that level were recorded by players who have now been banned. No-one still playing has recorded anything above 4 sigma.

As Matt says immediately up the thread, that is itself highly suspicious.

[Roger de Coverly]

Same players same ratings used. One event zero players with z score of 4 (as all the historic and mathematical evidence would predict) the other event a significant number of players with a z score of 4.

Are you saying that the 4NCL online league was calibrated against the 4NCL over the board league?

[Matthew Turner]

Roger,
The tests in both leagues use exactly the same parameters.

[Roger de Coverly]

The tests in both leagues use exactly the same parameters.

Are you saying then that the performance of a player in the online league was compared against performance over the board and discrepancies noted? Was this a factor in the lichess decision to ban players?

[Matthew Turner]

If my memory serves me right, the sigmas at that level were recorded by players who have now been banned. No-one still playing has recorded anything above 4 sigma.

I don’t believe that was explicitly said and I’m not sure that it is true

[Matthew Turner]

Roger,
The Ken Regan tests are calibrated (as you put it) over millions of historic games to assure the accuracy. The test carried out on the 4NCL and the online 4NCL use exactly the same parameters. I don’t think I can put it any clearer than that. In the 4NCL there are no players with a z score of 4 (as historic data and maths would predict) In the online 4NCL there are a significant number. It really is your turn to answer a question, how do you explain that?