Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

[Rene Preotu]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

Guilt by statistics strikes me as laughable.

Recently my son was added to the computer list (C) on ICC because he won 7 games (3 minutes time control) in a row. His ICC rating was over 2400 and his opponents were 2159, 2429, 2277, 2164, 2258, 2270 and 2226. ICC sent me an email (see below) but didn't give me any details about the "other factors". I managed to get the PGN files for the 7 games and found out that 76.7% of his moves matched with Houdini 3 moves. I'm sure he didn't use "computer assistance" during these 3 minutes games but the statistics are against him. We went with option 1 because he needs ICC for online training. For blitz he's now using the new FIDE Arena and Playchess.com.

You have been added to the computer list on the Internet Chess Club, and a (C) will appear by your handle. Our computer-detection group has analyzed a number of your games and other factors, and determined that you used computer assistance in some of your ICC games. This is a violation of ICC rules, unless you have a (C) beside your name. This decision cannot be appealed, but you have 3 choices on how to continue from here:

1) You can continue on ICC with your current handle and the (C).

2) You can have the (C) removed if you promise to never use computer-assistance on the ICC and place the following note in your
first finger note for at least 60 days. While playing on ICC, I used a computer chess program to assist my play. This is against ICC rules. I apologize for this and promise not to use assistance of any kind in the future. Message FREEBIRD as soon as you put the note in your finger, so we can start the 60 days. If you remove the note in less than 60 days, we will put the (C) back on your account.
NOTE: If you are caught cheating again on this account or any other account, appropriate actions will be taken at the sole discretion of ICC management. This might include, but is not limited to, a permanent (C) or even banishment from the club with no refund.

3) You can quit the ICC and get a pro-rata refund of your remaining membership time. Read "help refund" for details. But you cannot
register another account on the ICC. If you do, the new account will be banned. Asking for a refund means you are leaving the ICC
permanently.

[Kerry Liles]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

Recently my son was added to the computer list (C) on ICC because he won 7 games (3 minutes time control) in a row. His ICC rating was over 2400 and his opponents were 2159, 2429, 2277, 2164, 2258, 2270 and 2226. ICC sent me an email (see below) but didn't give me any details about the "other factors". I managed to get the PGN files for the 7 games and found out that 76.7% of his moves matched with Houdini 3 moves. I'm sure he didn't use "computer assistance" during these 3 minutes games but the statistics are against him. We went with option 1 because he needs ICC for online training. For blitz he's now using the new FIDE Arena and Playchess.com.

You have been added to the computer list on the Internet Chess Club, and a (C) will appear by your handle. Our computer-detection group has analyzed a number of your games and other factors, and determined that you used computer assistance in some of your ICC games. This is a violation of ICC rules, unless you have a (C) beside your name. This decision cannot be appealed, but you have 3 choices on how to continue from here:

1) You can continue on ICC with your current handle and the (C).

2) You can have the (C) removed if you promise to never use computer-assistance on the ICC and place the following note in your
first finger note for at least 60 days. While playing on ICC, I used a computer chess program to assist my play. This is against ICC rules. I apologize for this and promise not to use assistance of any kind in the future. Message FREEBIRD as soon as you put the note in your finger, so we can start the 60 days. If you remove the note in less than 60 days, we will put the (C) back on your account.
NOTE: If you are caught cheating again on this account or any other account, appropriate actions will be taken at the sole discretion of ICC management. This might include, but is not limited to, a permanent (C) or even banishment from the club with no refund.

3) You can quit the ICC and get a pro-rata refund of your remaining membership time. Read "help refund" for details. But you cannot
register another account on the ICC. If you do, the new account will be banned. Asking for a refund means you are leaving the ICC
permanently.

I'm sorry to hear he needs ICC for training :) - on receiving that I would have picked #3 and had some choice language for those
fools (almost my last choice of word for them) ...
How exactly does one use computer assistance in 3 minute games?
It seems hard to have houdini or any other engine cough up anything on demand let alone on a lot of moves.

What would be interesting would be to have a computer analysis of Botvinnik's games - most of them (all?) were before there were any
computers or chess-playing computers. I wonder how many of HIS moves would be flagged as 'computer like'?

[Bob Gillanders]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

Recently my son was added to the computer list (C) on ICC because he won 7 games (3 minutes time control) in a row.

Wow! I guess we should now add a (C) beside his name on the preregistered list for the Mississauga Open!
Just kidding of course.

This is clearly a case of "guilty by statistics" gone mad!

[Peter McKillop]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

I'm sorry to hear this, Rene. It must be very frustrating for your son to be falsely labelled and punished without so much as a chance to defend himself and without any reasonable way to fight the decision.

[Ken Craft]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

I'd have responded with 4) take appropriate legal action.

[Kerry Liles]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

I'd have responded with 4) take appropriate legal action.

Unfortunately, I think that taking action against the ICC is probably as easy as taking action against FIDE and IlluisionOf...
I doubt you would get very far without deep pockets and a lot of free time. Even if you "win", what exactly might you achieve?
I think Bob Gillanders hit the nail on the head with his comment: "guilty by statistics" gone mad!

[Peter McKillop]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

I'd have responded with 4) take appropriate legal action.

I'd have wanted to do the same thing, Ken. Practically speaking though, you'd be looking at trying to prove...what?...that ICC is wrong in taking the position that winning 7 blitz games in a row with an x% Houdini correlation is proof positive of cheating? Would you need an expert to testify? The legal fees would be through the roof, I bet. Provided they don't go too far out on a limb, ICC must think they're bullet proof on these types of decisions for the simple reason that the cost of fighting their decision would be way out of proportion to the benefit.

[Mathieu Cloutier]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

The probability of Paul Bonham being reasonable in a discussion is so low that I don't even bother anymore...

[Paul Bonham]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

The probability of Paul Bonham being reasonable in a discussion is so low that I don't even bother anymore...

I see you're still upset that I spanked your ass over your misuse of statistics. Don't get mad -- get educated.

[Paul Bonham]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

EVERYTHING in the scientific/engineering domain is based on statistics and probability. Do we know 100% the speed of light, no. But do we know it to a high degree of probability ? Yes.

Yes, even our very existence and everything around us is simply probability. And when it comes to deciding any court case, it is always possible even in the face of indisputable evidence, such as being caught in the act, there may be extenuating circumstances by which the accused could be found not guilty, but we may ignore or not even know of such circumstances and go ahead and pronounce guilt. So I agree with your basic premise that nothing is ever 100% certain.

But we should still strive for very stringent requirements on probability before we make decisions on guilt or innocence. The essence of our judicial system is that one is innocent until proven guilty "beyond a reasonable doubt", which is a vague term. We should make sure it doesn't become more and more vague. Nigel would appear to prefer that we do away entirely with that phrase and replace it with "on the balance of probabilities", which if you think of the word "balance", could be construed in the extreme case as 51% probability. If it is 51% probable that you are guilty, that tips the balance and we can pronounce you guilty. This is where Nigel appears to want us to go.

Yet the very DNA testing he uses as an "example" is subject to very much higher probabilities, according to
http://en.wikipedia.org/wiki/DNA_profiling

"When using RFLP, the theoretical risk of a coincidental match is 1 in 100 billion (100,000,000,000), although the practical risk is actually 1 in 1000 because monozygotic twins are 0.2% of the human population."

Very clever indeed of Nigel to use an example of at worst 99.99% certainty to push an agenda that pushes that certainty down to (in the extreme case) 51%. If Nigel disagrees with the 51% number (again, suggested by use of the term "balance"), I would press him to provide us a probability number that he would deem sufficient to pronounce guilt or innocence, and then to explain to us how he arrived at that number.

If you read Rene Preotu's post from today, you'll see that his son has been convicted of computer use on ICC, with no chance of appeal, on the basis of only 76.6% of his moves matching any of Houdini's top 3 choices during his games. How much lower can that number go to pronounce guilt with no chance of appeal?

Not only that, but they give his son 3 options, EACH OF WHICH IS BASICALLY AN ADMISSION OF GUILT. There is no option provided for fighting the decision or argueing innocence. So it seems the chess establishment, in its rush to prevent cheating, is taking steps that in the end will likely choke off future growth and perhaps even kill the game. Rene's son may some day realize how he was abused in this case and may simply quit organized chess, and many others like him.

And I've already given examples of where this kind of agenda leads: the Salem Witch Hunt, Germany from 1933 to 1945 (the Gestapo), and late 1950's America (McCarthyism). How low do you think the probabilities were in each of those eras for guilt to be pronounced?

"All I can say is, be careful what you wish for. I'm sure we can find some probabilities that apply to YOU"

What on earth are you talking about. Either you do have something statistically true to a high probability (that doesn't break the laws of probability) or you don't, it's not a question of being able to "find" something on a whim. If you are able to find every chess player guilty of cheating through your "analysis", it is your analysis and premises that are flawed. Not the laws of probability.

See my response above to see "what on earth I am talking about". It is indeed a question of being able to find something on a whim, as long as you can show N% probability where N is a number that creeps lower and lower towards the 51% ideal that Nigel appears to desire. So I advised him to be careful what he wishes for, the "secret police" may serve and protect... and may also arrive at Nigel's front door to convict, with no chance for appeal.

And by the way, I love your conclusion that "it is your analysis and premises that are flawed". Exactly. The computer engine analysis, and the premises that go with it, are definitely flawed -- that is my very point. When 76.6% of a Top 3 match is enough to convict, and we have innocent people being convicted even now as we speak, "something is fishy in Denmark".

[Peter McKillop]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

Laugh away.

As I noted in the the Borislav Ivanov "cheat thread", probabilistic evidence is very common. This sort of language is even used when, strictly speaking, no exact percentages are given. "On the balance of probabilities", said one Canadian civil servant, "so-and-so is not a competent parent." And, on the basis of their expertise (and the expertise of others in agreement), a child is taken away from her mother. Happens all the time in this country and, with minor language changes, in dozens of others. And it happens with all sorts of cases and circumstances.

But you're welcome to retain your opinion. I might caution you, however, that DNA evidence is ALSO given in terms of probabilities, so I guess you must be in favour of jettisoning that sort of evidence as well?

Good luck with that. I'm sure David Milgaard, Thomas Sophonow and others who spent decades in jail for wrongful convictions in Canada will be happy to go ... back to jail. After all, they were freed based on DNA evidence. To hell with probabilistic nonsense, eh?

It was my impression that Kerry was referring to 'guilt by statistics' in the matter under discussion, cheating in chess. How like you, Nigel, to twist his words and make it appear that Kerry wants the use of probability-based evidence stopped everywhere.

[Dan Alistair]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

And by the way, I love your conclusion that "it is your analysis and premises that are flawed". Exactly. The computer engine analysis, and the premises that go with it, are definitely flawed -- that is my very point. When 76.6% of a Top 3 match is enough to convict, and we have innocent people being convicted even now as we speak, "something is fishy in Denmark".

I do not know enough about the statisical basis of chess cheat catching software to comment. But I think others have pointed out flaws specific to your analysis, for example the one about sample size which is huge.

Sorry but based on some of the things you have said i have to trust their statements rather than yours :).

I would think ICC used some reliable modeling from some really smart statisticians before they accuse people. It is idiotic to compare it to the Salem witch hunts or Macarthism. Those were not based on raw computer data as that would be available to a sitelike ICC.

[Kovalyov Anton]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

Chess is not a single move game, nor it is a chance game. Every strong computer move is a result of calculating millions of complicated positions, that prove why this move is good. If someone plays every single move, no matter how complicated or simple it is, at a constant rate of 5-10 sec/move, while everything seems to work out great, then something is certainly wrong, but there also other quite convincing proofs against Ivanov.

I'm not a number person nor I'm good in statistics, but as far as I know the number of different positions possible after just the first 10 moves equals a 30 digit number. There is no such thing as patterns in computer games. Given today's technology if you try to study chess by checking out and remembering the best possible moves in every possible position, it will take you several billions of years...

If you are so interested in chess, go out and play some games with real people, try some blitz tournament, maybe buy some good chess books, and if you like it, take some lessons. After this, I assure you that you will be laughing at yourself while remembering your chess related arguments in this site. From this moment I will abstain myself from explaining further how chess actually works and how chess players actually think during a game, to someone who evidently has no clue.

[Paul Bonham]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

I do not know enough about the statisical basis of chess cheat catching software to comment. But I think others have pointed out flaws specific to your analysis, for example the one about sample size which is huge.

Sorry but based on some of the things you have said i have to trust their statements rather than yours :).

I would think ICC used some reliable modeling from some really smart statisticians before they accuse people. It is idiotic to compare it to the Salem witch hunts or Macarthism. Those were not based on raw computer data as that would be available to a sitelike ICC.

I am not using those events as a direct comparison to right now. I'm saying that once we start on the path of lowering the probabilities needed to convict, we are on a slippery slope towards returning to those kind of eras. So we need to guard against lowering those probabilities imo, and so far at least most people in our civilized society agree with this, otherwise our judicial system would already be lowering their standards.

Also, I have not agreed that my one-game analysis convicts Kovalyov of cheating. But -- and this is a very key but -- by just mentioning that he matched Stockfish to such a degree, I have implanted in everyone's mind the nagging question... "could he have cheated?". That's the whole point here: we are using these measures to implant a SUSPICION of guilt. And someone (presumably an "expert in statistics") is going to decide that if 4 games or 6 games or 8 games or 12 games is enough to turn that suspicion into a conviction. Apparently some such person or persons at ICC decided 7 games was enough to convict Rene's son, with no appeal possible. Not only that, but we don't know HOW they arrived at their numbers. What engine(s) did they use? What search depth did they apply? These were 3 minute games; as someone pointed out, how could someone possibly manage computer engine cheating in such short time controls?

This is all "secret police" stuff. Very dangerous and leading to a slippery slope.

[Peter McKillop]

Re: Chess Cheating on Superficial Evidence: Time to Put Up or Shut Up

...
I would think ICC used some reliable modeling from some really smart statisticians before they accuse people. It is idiotic to compare it to the Salem witch hunts or Macarthism. Those were not based on raw computer data as that would be available to a sitelike ICC.

But Dan, their modelling would be based on some assumptions, wouldn't it? I mean, how would ICC do this? Would they put a call out to all cheaters: attention cheaters....we're doing some statistical investigations so we can set up cheating detection software....please identify yourselves so that we can study your results vis-a-vis our non-cheating control groups? I'd really be interested in finding out how they do this. Also, Rene Preotu mentioned elsewhere in this thread that his son came to ICC's attention because he won seven blitz games in a row against competition which, with one exception, was significantly weaker. I wonder how many ICCers there are who haven't won 7 blitz games in a row at least once in their ICC careers?

As for the Salem witch hunts and McCarthyism, you may disagree with Paul's choice of analogies but, if you do, you should be able to come up something more intelligent than 'idiotic', shouldn't you? Personally, I think our beloved and thoroughly impartial moderator, Mr. Nigel Hanrahan, should apply the Xs of disapproval to your word 'idiotic.'