XV. World Correspondence Chess Championship Qualification

For convincing argument, you need to use concrete numbers here (number of games played, number of moves per game, engine match ratios).

In a high-level game even consulting an engine just once would be essential. Just once. You have examples of high-level GMs asking at the press conference: "did I have a win? I should have made this move, right?". Such form of engine-assistance is impossible to detect offline.

Let's compare it to athletics. I don't know exact rules there, but I assume if your testosterone level is above the 99th percentile of normal healthy population (usual cut-off in clinical chemistry), you shouldn't be allowed to compete. It is irrelevant whether you cheated taking pills & shots or were on special bull balls diet or you are just unlucky 1%.

Control samples for cheat detection come not from normal population but from elite CC GMs (pre-computer ICCF WCC finals). 99.9% is nothing special, ordinary cheaters here (e.g. Yelena Dembo, Ouachita, Achmatova) were way above 99.999%. And they thought they couldn't get caught.

First example: A game has 35 non-book moves. Human would make 29 moves computers like (top3 83%) and 6 moves computers don't like that much (out of top3). If a player makes additional 4 moves computers like, we get top3 33/35. If we have player's sample of 10 such games in a tournament, top3 becomes 330/350 (94.3%) with z-score +7.34 and probability somewhere around 10^(-13).

OK, second, more realistic example: In chess, moves tend to come in logical series and most of cheaters suffer with cognitive bias. Let's assume that completely checked honest player decides out of blue sky to cheat in such prestigious competition. He/she thinks that 3-4 times during each game is sufficient enough to get an advantage and avoid positivity, and he/she maintains enough self-control to avoid consulting engines more (btw, this is not usual, cheating is addictive). However in serious competitions, there are no cheap tactical shots, so if he/she wants to gain advantage from it, he/she must follow the computer's complex line for additional 3-5 (or more) moves. Good CC game has 20+ nonbook moves and let's assume he/she makes 300 non-book moves in 10 completed games. I must point out this is terribly small sample for classic top3 method as presented by Steve Collyer. So in these games he/she consulted an engine 30 times and followed the recommended line on average for 4 moves each. That is 120 pure computer moves, 180 other moves. Let's assume that from these 120 pure computer moves detection engine catches 95% in top3 and thus misses 6 moves. Average elite human top3 match is 83%, that is only 31 moves computers dislike from the remaining 180 moves sample. (300 - 31 - 6)/300 = 87.7%, z-score +2.20 (due to only one tail P = 1.38% or 1 of 73) => definitely enough to raise red flag and have deeper look into his/her games, compare this result with his/her previous games, make blunder graph, compare top1 vs top3, run detailed test with refined sample etc. Actually alone top3 z-score between 2.0 and 3.0 isn't sufficient for lifetime ban, but it might be sufficient to refuse awarding medal/title/prize...

You make one of two assumptions:

1. either the player is really weak and needs to use an engine almost always in order to win

OR

2. both players are computers, so both of them must play accurately.

But, if you have one strong player, in a single game, who will look at the output of an engine in order to make a tactical shot or not, you will never be able to find this out.

Indeed, here's the crux of the difference between motarlan's comment "Nobody is able to detect engine usage. 4-5 times in a game" and your followup "Actually 4-5 more engine moves per game during tournament are sufficient enough". Come to think of it, the word "more" already shows that you're talking about different things.

With the expected T3 match of 83% (make it even 80%), using engine 5 times in a game will provide only one more matching move (in the other 4 cases, a well-considered human move would fit in the top 3 computer choices).

Similarly, to register 4 additional T3 matches, the player would have to consult the engine 20 times during a game.

So both of you are correct.

See example II.

The bottom line is that such careful cheater chooses not only this one actual computer move, but the whole set of subsequent computer-like moves appears in the game. He does this a few times during a game in several games and then he is surprised he got caught.

Another major issue is cognitive bias called "hindsight bias", observed very often with centaur correspondence players. They would swear they play solely own moves and ideas and quite reluctantly they admit "little help" from the computer. In fact they adopt computer's move as their own after they saw it on the screen. This plagues cheaters too. Cheaters must also care to avoid getting caught. So these cheaters consult engines a lot thinking if they choose only moves they would choose without engines anyway, they will not get caught. Wrong! Hindsight bias sets in and bends their judgment so they end up playing almost exclusively computer moves. Top3 analysis gives z-scores in the "idiotic cheating" range (above +5). Then they feel offended when we call them "idiot cheaters" and come to conclusion cheating detection has been erroneous, because they "almost didn't cheat".

The bottom problem IMO is that you suppose cheaters are stupid, or become so because they become addicted to the computer.

If only one decent player manages to really restrain his computer use, working hard like a human, and only after checking the computer, and this not for all moves, I don't see how he can get caugth on say 100 games, let alone 10. And yet, he will steal the top prizes.

Btw, this is interesting. I have searched your archive and found this game:

http://www.chess.com/echess/game?id=40400192

It is obvious your opponent defended against dangerous Marshall Attack only with his own guesswork while you used book or database. It wasn't too hard to find few decisive final moves. This definitely does not constitute a proof of engine usage.

There is another issue I want to point out. What makes me wonder how a player of your caliber can miss checkmate in 1 at move 23? You may say you were drunk or it was mouse slip or you wanted to play cat-mouse game and toy with your opponent... blah-blah-blah. But it can be also a sign of primitive dirty attempt to deceive cheat detection. This "blunder" doesn't ruin anything, the position is easily won anyway.

Let's imagine such thing happened in an official Anti-ICCF tournament. I am t3-analyzing the player and trying to decide whether he deserves his medal. The result come borderline with z-score around or slightly under +2, so I replay his games and now see this. I assure you I would cancel his results by default without hesitation and recommend at least 3 years ban from official Anti-ICCF competitions.

Don't feel bad, but I am getting kinda curious about your rated games here...

Precisely, in centaur, computers are allowed and thus you are bad if you just sit and follow the computer.

But in 'honest' CC, sitting and letting the computer play instead of you usually wins. "So why bother ?" is the reasoning of most cheaters, which leads them to be detected, or at least 'detectable'.

A centaur player that goes into CC for cheating would not be easy to spot if he knows how to restrain himself.

@ Firebrand. Did you give up playing rated c.c games here?

You haven't played one for years.

I think you're being a bit draconian in your approach, the fact firebrandx didn't see the mate and had his own plan which was different kinda supports that he isn't using an engine to play I feel.

I wasn't suggesting anything wrong at all in your game PB illustrated Firebrand.

Was making the observation that whilst you choose frequently to comment in the forums here on chess.com you cannot currently be considered an active playing-member. These days you rarely ever play any chess at all here.

" There is another issue I want to point out. What makes me wonder how a player of your caliber can miss checkmate in 1 at move 23? [...] I assure you I would cancel his results by default without hesitation and recommend at least 3 years ban from official Anti-ICCF competitions."

You are sick. Really.

You aren't in position to judge my reasoning. Especially when you cherrypick.

I meant such sign would be - but only in case of borderline positivity - the additional ground for disqualification. It can be:

a) attempt to deceive cheat detection by throwing blunders

b) unsportsmanlike behaviour (toying with opponent)

c) sharing the game with another, much weaker player

d) oversight (very unlikely with WCCC candidate player)

Would you also disqualify players who are too tall?

Austin Lockwood apparently agreed with you.

An account with the name "AustinLockwood".

Sockpuppetting is not that hard.

I think it might have been Austin Lockwood, noticed by bananaman. He has no reason to stay and play here as he runs his own CC playserver, aside from ICCF. I have no reason to answer his disappeared post. He knows my email. And apparently he has more important things to do than discuss cheat detection.

My assumption is following: 90 cheaters out of 100 are plain stupid, either thinking cheating is undetectable or underestimating detection or suffering with cognitive bias. They get caught. From the remaining 10, 9 have no own chess skill and experience whatsoever to emulate excellent human play with computer assistance. They may choose between getting caught and/or losing to much better human players. The remaining one must work hard, have sufficient knowledge and keep own statistics with good self-control. Still it doesn't guarantee him full success in both. Also keep in mind all participants are checked before, people who were caught cheating somewhere in the past or were found positive in testing, aren't allowed to participate. Scrupulous honest players will not start cheating out of blue sky.

This thread was about one tournament on this site, it is not to be used to discuss cheating, accusations of cheating, those previously banned for cheating or anything else cheat related otherwise the thread will be locked.

@FirebrandX -- did the email came from ICCF domain? Did you check the headers for spoofing/open relay traces? There are websites that can do email header analysis for you, e.g., http://www.spamcop.net.

Just curious. Claiming very loose association with ICCF while serving 2nd term on the Services Committee would make me pay extra attention there.

Of course, it may be all legitimate.