Cheating in chess

[JustinHorton]

Can I also specifically reiterate with regard to this

That is this one. It takes a bit of patience but is worth a careful reading.

It's interesting that they make the point that a defence is to show that the engine matching was down to pre game preparation. That has echoes of the chess.com accusations against ejh where he was able to demonstrate that the supposed engine assistance was taken from material on his bookshelf. Being "daily" chess, akin to correspondence, this was within the rules and to an extent the point of that mode of play.

that while this may be the case, it may also be that what I did was irrelevant since to this day I do not know what the evidence was on which I was initially accused. This is something people need to think seriously about.

[JustinHorton]

Do you follow 12.26?

No, but I've had it explained to me elsewhere on this site so I can always look it up if I can only remember where....

There are some parts of that judgement which bothered me a little, which is not at all to say that I think they came to the wrong conclusion. Just to go through the relevant paragraphs:

12.4.One of the best-known applications is credit scoring, which is used throughout financial services. Scoring models process a customer's credit history, loan application, customer data, etc., in order to rank-order individuals by their likelihood of making future credit payments on time.

This doesn't tell you at all how well "predictive analytics" perform in this field, does it? But I am not under the impression that financial services have a stellar record in this area.

12.29.The interpretation of odds against fair-play hence involves a so-called "transposed conditional"---and it should be understood that they are only "face-value odds" which may be lessened or strengthened by other factors. Keeping this in mind helps avoid well-documented "fallacies" of statistical interpretation which have figured in some historical judgements now regarded as incorrect.

I think I would have appreciated some information as to which fallacies those were, and how the above would help us avoid them.

These might be quibbles, and I only mention them in passing, but what did bother me more seriously was this:

13.7.Apart from the slightly anomalous result produced by Stockfish, the outcome of the various tests is a z-score of about 4,50 (translating into odds of 300 000:1), an IPRNB in excess of 3 000 rating and values of 73 and 87 on the ROI.

"Slightly anomalous"? Here's the table from the preceding paragraph 13.6:



It's way out of line with the other three engines, and I would want to hear some stronger reason why this might be so than appears in the report, where indeed nothing further is heard of it.

[NickFaulks]

Do you follow 12.26?

No, but I've had it explained to me elsewhere on this site so I can always look it up if I can only remember where....

So somebody thinks it's right?

[JustinHorton]

Far as I recall, and if I ever track it down, I'll draw your attention to it

[Matthew Turner]

Do you follow 12.26?

No, but I've had it explained to me elsewhere on this site so I can always look it up if I can only remember where....

There are some parts of that judgement which bothered me a little, which is not at all to say that I think they came to the wrong conclusion. Just to go through the relevant paragraphs:

12.4.One of the best-known applications is credit scoring, which is used throughout financial services. Scoring models process a customer's credit history, loan application, customer data, etc., in order to rank-order individuals by their likelihood of making future credit payments on time.

This doesn't tell you at all how well "predictive analytics" perform in this field, does it? But I am not under the impression that financial services have a stellar record in this area.

12.29.The interpretation of odds against fair-play hence involves a so-called "transposed conditional"---and it should be understood that they are only "face-value odds" which may be lessened or strengthened by other factors. Keeping this in mind helps avoid well-documented "fallacies" of statistical interpretation which have figured in some historical judgements now regarded as incorrect.

I think I would have appreciated some information as to which fallacies those were, and how the above would help us avoid them.

These might be quibbles, and I only mention them in passing, but what did bother me more seriously was this:

13.7.Apart from the slightly anomalous result produced by Stockfish, the outcome of the various tests is a z-score of about 4,50 (translating into odds of 300 000:1), an IPRNB in excess of 3 000 rating and values of 73 and 87 on the ROI.

"Slightly anomalous"? Here's the table from the preceding paragraph 13.6:



It's way out of line with the other three engines, and I would want to hear some stronger reason why this might be so than appears in the report, where indeed nothing further is heard of it.

Justin,
You make some excellent points here. Ultimately, I think what it shows is that you cannot just say "oh a Z score of 3 means this", you actually have have to look at the evidence a bit more deeply and holistically.
I agree that it is worrying when one engine produces a Z score of 3 and another 4.5. I am afraid the arbiter, or the appeals committee, or whoever it is, has to look through the evidence and try to understand what is going on and it is bloody difficult!
Here is a potential explanation (I have no evidence this is actually the case, but it is the sort of thing you might consider). A player might try to be a clever cheat, so they select computer moves until they get a really good position and then start throwing in a few bad moves or even outright blunders. We try to cut these out by stopping evaluating the moves once a player has a certain advantage, in this case +3. It is noticeable that there are more moves being evaluated by Stockfish, so maybe this took in a number of deliberate blunders that were excluded by the other machines. That would obviously lead to a lower Z score.

[Kevin Thurlow]

"I'm a little disappointed to see a player unceremoniously booted off the England junior team at the end of the Glorney Gilbert Online International back leading a team at the Junior 4NCL this evening. I'd have thought there would have been some form of sanction or deterrent, not just a shiny new Lichess account and away you go!"

Going back to this, I'm assuming this is one of the favoured juniors, rather than just an ordinary junior? Or are 4NCL saying he was innocent?

[Mick Norris]

The general approach seems to be being banned in 1 competition doesn't bar you from another

So, get banned in the Glorney but still ok for County Championship and 4NCL

Get banned in 4NCL still ok for County Championship etc

Intriguing

[Kevin Thurlow]

And in Mind Sports Olympiad, someone accused of cheating at "Othello" got a 10 year ban.

https://msoworld.com/2020/08/fair-play- ... ification/

Interesting to see different approaches to the same problem.

[Mick Norris]

A ban from MSO only so far, though, yet to hear from the Federations

[David Sedgwick]

And in Mind Sports Olympiad, someone accused of cheating at "Othello" got a 10 year ban.

https://msoworld.com/2020/08/fair-play- ... ification/

Interesting to see different approaches to the same problem.

Congratulations to Etan Ilfeld and the MSO. Full investigation; fair opportunity for the accused person to respond; decisive and severe action.

[Matt Bridgeman]

Yes, somehow more satisfying than a player being banned in the 4NCL Online Congress and then turning out 2 weeks later to win as top board in the Counties Online. Or a player being banned in the Glorney Cup Online to turn up 3 weeks later to win as a top board in the J4NL.

[Roger de Coverly]

Full investigation; fair opportunity for the accused person to respond; decisive and severe action.

They had it easy, as it appears that the accused person wasn't a good player. Many of the chess cases are where in effect a player of reasonable strength attempts to follow ICCF rules when playing Blitz or Rapid online. Interviewing them about their games wouldn't help unless their success could be attributed to superior pre-game preparation which could exonerate them..

[Matt Bridgeman]

The UK Chess Challenge Terra Final is on Lichess on the 5th and 6th September, with a couple of thousand pounds in prize money on the table. I can’t think of an online event with such high prize money. I think the event just relies on Lichess for anti-cheating measures. It could be interesting if the Mega finals and Giga finals are anything to go by;
https://www.delanceyukschoolschesschall ... ules-2020/

[Wadih Khoury]

The UK Chess Challenge Terra Final is on Lichess on the 5th and 6th September, with a couple of thousand pounds in prize money on the table. I can’t think of an online event with such high prize money. I think the event just relies on Lichess for anti-cheating measures. It could be interesting if the Mega finals and Giga finals are anything to go by;
https://www.delanceyukschoolschesschall ... ules-2020/

The final four will be with webcam, but not the qualifying Swiss.

[John Swain]

The UK Chess Challenge Terra Final is on Lichess on the 5th and 6th September, with a couple of thousand pounds in prize money on the table. I can’t think of an online event with such high prize money. I think the event just relies on Lichess for anti-cheating measures. It could be interesting if the Mega finals and Giga finals are anything to go by;
https://www.delanceyukschoolschesschall ... ules-2020/

The final four will be with webcam, but not the qualifying Swiss.

I understood from one of the official Twitch broadcasts a few weeks ago that Megafinal and Gigafinal games were also subjected to analysis by Ken Regan's anti-cheating software.

The decision to use a webcam at the Terafinal for the final four in each section as an additional anti-cheating precaution sounds eminently sensible, although it only reduces the possibility of cheating, rather than eliminating it altogether.

It is interesting that very few of the alleged cheats in this year's online UKCC have indicated that they were cheating, despite the inducement that the slate would be wiped clean for 2021 if they did so: https://www.delanceyukschoolschesschall ... -mistakes/