Chess.com FAQs and Discussion on Cheating

That Hansel is so hot right now.

I totally agree with you.

You know, you're right, McCarthyism was perfect...

All marriages last, whether you realise it or not...

1. Ultimatums such as this "you're either with us or against us" have consistently been the slogans of autocrats, though at best they are horrible policies of well-wishers.

2. People are wrongly accused all the time. For instance, I accuse you of cheating. If your logic were correct, then by this virtue, you are a cheater, and you should admit this and accept your banning immediately. That's ridiculous, but on a case-by-case basis, it is your argument. Cheaters are probably here and in large numbers, but you shouldn't ban someone just for playing well. Please disagree with that and make it even more obvious to everyone just how ridiculous your position is.

3. What lies? The only lies I see are you pointing to strong likelihoods and opinions as settled facts, which they clearly aren't. And it is ridiculously obvious that the cheaters have no motivation for admitting their cheating, and really nothing you're going to say could change that, even if they read this thread, which is incredibly unlikely.

On the respect issue, I respect every human being to a certain extent intrinsically because of their value as a human being, their dignity, their autonomy. See the works of Immanuel Kant. Further than this I have an added layer of presumed respect for any random person which I meet, because most people are good people. This is not very high to begin with, but the more I'm around a person, the more it usually grows. However, it's also very possible for someone to lose my respect and earn my disrespect, which happens very often, including in your case, Richie. It is very rare for me to respect someone at a very high level, as this must be earned, generally over large periods of time, but it does undoubtedly happen.

So yes, I respect all people. I respect neo-nazis. I respect Hitler.

Respect does not mean agree. Respect does not mean like.

As for many definitions, that's true of almost any word, to the extent that if all definitions were always open, and you never used context, it would be nearly impossible to parse most language which is commonly used, and I'd certainly have no idea what you are saying.

A sensible post (finally) though one that I'll raise a couple of minor qualms. Your statemtent in the post above (1264) is true, but your previous accusations about lying painted a completely different picture, not conclusively, but through implication. One might call it a Clinton-esque deceit. I wouldn't, but then I don't hold as much against Clinton as most people do, so they'd probably be apalled that I thought you to be more deceitful in this case.

Furthermore, your contention that you don't appreciate it is meaningless. I and many others here don't appreciate the nature of many of your postings. So what? Why are you allowed this privelege while others are not?

However your points about proof are completely right and compelling. I actually hadn't thought about this in quite some time because I laughed at the idea that they had some kind of conclusive proof when I first read it, immediately dismissed it, and then basically forgot they even tried to make such a ridiculous claim. Of course it comes to human subjectivity in the absence of admissions (which you aren't going to get), and I am on the side of having very high standards in situations like this - I'm of the opinion that a false positive is much more harmful/injurious than a false negative.

Steve: I think your conclusions are wrong. I do not think a reliable analysis can be obtained using 30s automatic analysis of Fritz. I guess that, if things were so simple, online cheaters from this site and others would have been eradicated for a long time.

You also contradict yourself: you say that a human with a top 3 mathup over 90% is a cheater. But you show that a strong engine (Shredder) had a matchup of 85% in more than one game (15 or so). So, Shredder plays like a human, and any player here using Shredder for assistance is fine. Is that your conclusion? You fail to give a clear answer.

The idea of not counting forced moves is not mine. It belongs to a person who did some serious research on the cheating issue. I posted once the link, I guess you did not take the time to read it :)

http://www.cse.buffalo.edu/~regan/chess/fidelity/

It also looks to me that you only copy-paste analysis from Red Hot Pawn. Karl also was surprised that your conclusions are not clear at all, you indicate that some engines do play like humans. The bottom line I think : low quality analysis. I assume that a reliable analysis would take at least an hour per game. A serious one would take a whole night :)

Cheater_1 is an expert on the subject, so we should listen to him.

P.S. You sound a lot like him lol.

I'm enjoying the conversation between SteveC and costelus, who seem to be the only experts on the subject we're discussing.

Costelus, I don't think he's using automatic analysis since that usually doesn't show non-best matches. 

Automatic analysis is a terrible way to do this kind of testing so I hope that's not what's used.  The move hashing that occurs in retrograde analysis (i.e. automatic) greatly distorts the results and doesn't give an accurate picture of what someone would see if they were cheating in infinite analysis mode.  Strong human moves will often show up as matches even though the engine never would have seen the move on its own.  Any results derived from automatic analysis would be extremely unreliable and I sincerely hope no one is getting the boot based solely on this method of testing!

That said, I agree with you that 30s is far too brief a time.  It would be much better to base the eval on depth rather than duration.  In certain complex positions, 30s on good hardware is still not enough to correctly evaluate the position and in CC games 30s is laughable.  (Maybe that's why so few top players get banned?) 

If I recall correctly Kenneth Regan uses depth rather than duration when testing.

Ozzie: No way, I am not an expert. I just stumbled over Kenneth Reagan home page looking for his research papers. And I noticed his interest in the cheating issue. I simply read some of the information he put on his webpage. 

He clearly outlines the importance of hashing and `playing` with the engine, since otherwise it will never be able to see some good moves as matches. Here is his methodology, in short (using his words):

• (a) As before, start on (at least) the previous move to the desired range of moves, to fill up the hash table.
• (b) Single-line mode until getting a PV at depth 17 (or 18). Clip analysis then.
• (c) While at that depth, increase the # of lines computed fully to (I recommend) 10. Clip analysis at 18/18, which marks the end of the 17-ply round (or at 19/19 if you went to depth 18 in single-line mode). The calculation will re-start from the beginning, but it will have the benefit that evals of many lines found in single-line mode will already be stored in hash.
• (d) Step ahead to the next move while in 10-line mode. Wait until 11/11 shows. Clip that 11/11 if you can. (This may help fill some hash for later use by step (c), and might also reflect a reality of people getting a quick peek at options before going into single-line mode for optimum depth.) Then click the "-" button to reduce down to single line mode and goto (b).

RhP mods ... maybe mister cludi, who was a game mod there until he was banned for cheating :)

Well, I think that your analysis is a good trade-off between resources and results. It is likely to find all the cheaters in live chess and all the stupid cheaters in correspondence. But it will probably not detect strong ICCF cyborgs (who, on this site, are cheaters). 

The issues with forced moves is also raised by Reagan. I believe that he is right (by the way, he is IM). Again, I am not arguing with you, I am not the expert. But I am convinced that, without eleiminating forced moves, Kramnik should have been labeled as cheater in the toilet cable case.

The single game case is really not terribly relevant to the issue at hand anyway, since catching a cheater who has only played a single game on Chess.com is of less value than catching a cheater who is extremely active.  Building a bullet proof methodology so that you can handle the exception at the expense of a methodology that can efficiently deal with the rule just isn't pragmatic.  If the trade off is that a larger game history is required in order to arrive at a reasonable degree of certainty using the more efficient methodology then I think that this is a perfectly reasonable approach.