Chess.com FAQs and Discussion on Cheating

SteveCollyer,

Have you considered volunteering your time to this website? =)

Thank you -- I must have missed it.

Now, as for ozzie's request....

While I agree that it shouldn't materially affect the outcome of the analysis, one could tweak a working cheat detection algorithm to also attempt to detect cyborg DB users.

Clear something up for me.

 
Some people seems to infer that simply because moves played by strong masters do not agree with the 1st or even 2nd or 3rd choice moves of powerful programs that the masters are therefore, if not weak, at least weaker than the programs. 

I personally don't read that into the equation. How I see it is that programs  simply play differently than humans.  Of course, they also play much better than most people, but GMs play roughly equivalently. Anyway, the bottom line is that I see the move choice difference, not so much a matter of strength, but of vision.  A human views things in plans and ideas while a computer views things exclusively in lines and variations.

Because of this difference in vision, I would think that computer users should be able to be identified with relative accuracy.

Totally agreed, batgirl.

I think it was a facetious extension of my request of you.  Really appreciate the quick data gathering though -- it provides some valuable perspective that I think is a much better basis for comparison to the turn based chess here than OTB master games.

Steve: What is the depth you use for analysis? I am very surprised of the amount of time used to analyze so many games. And what exactly is the methodology you used? Did you let the engine perform an automatic analysis?

Your conclusion is quite clear: a strong CC player can obtain matchup percentages close to an engine. I have serious doubts about this.

I also think that forced moves or situations where there is a clearly winning move and other not so great moves should not be counted as matches. You are in check, you have two legal answers, if you choose one you will be mated in 3, if you choose the other, you play on. This is not a matchup!

I would be very interested to see a statistic for this game:

http://www.chessgames.com/perl/chessgame?gid=1430952

I'm afraid it is not clear at all and the data you provided is inconclusive. Your data:

{ Shredder - Intel Core 2, 8 x 3.16Ghz (Games: 9) } 
{ Top 1 Match: 271/467 ( 58.0% ) 
{ Top 2 Match: 344/467 ( 73.7% ) 
{ Top 3 Match: 374/467 ( 80.1% ) 
{ Top 4 Match: 399/467 ( 85.4% ) 

So, Shredder plays worse than a strong human? Or, in other words, if you use Shredder to make your moves, your play will look like a human. 

Could you also post a complete analysis, for any game? Not just some percentages, but the complete analysis, move by move. In this way one can see if your results can be duplicated.

Cyborg databases: huge collections of engine games. The engines are allowed to use opening books and tablebases.

Let's say we take a very sharp opening like Marshall attack. After let's say move 17 (all known theory until now) I put my engine to make a tree of the best 3 answers and the possible continuations for the next 5-6 moves. I store all the results in a database, which I am allowed to use while playing. In this way I am sure that, if my opponent choses to play Marshall, he will be cooked.

The cyborg database will not give me the possibility of winning the whole game only by looking up the moves. It will give me an edge over my opponent.

Last, I should add that those databases are not that uncommon for a guy with plenty of time. Professional GM's use such methods when studying opening lines. Besides, there are hundreds of centaurs games completed every day on ICC, playchess or ICCF.

Kepler: I noticed that most of the top players here (some of them banned in the meantime) deviate very early in the opening, sometimes even earlier than the normal level of OTB play. I've always said that they, being such strong players, don't want to follow the lines developed by patzers like Karpov or Kasparov.

You should also note that the goal is not to decide if a game is played by a human or by an engine alone. The goal is to discerne between a game played by a human and a game played by a human assisted by an engine. Thus, I think that out of the 3 criteria you ennounce, only the second one is valid (lack of tactical mistakes).

Yes Kepler, we agree: a smart cheater is very hard if not impossible to be caught. That's why I guess ICCF allowed engine usage. A smart cheater will drive his engine, finding good lines that an engine alone will never be able to find (thus, they would appear as non-matches in any analysis).

However, on a private website like this, things are simpler. The owners can ban anyone without having to provide huge piles of evidence. If X is banned, he is basically told: "either you cheat or you are too strong to waste your time here instead playing with the top GM's in the world". 

Fortunately for us, a "smart" cheater must have a certain OTB strength (I would say at least 2000 ELO), from which he can effectively drive his engine. A patzer cannot help his engine too much. This greatly reduces the number of "smart" cheaters. 

Lastly, I want to say that 100% detection is impossible. ICC with its Dos Hermanos tournament also acknowledged this, when - in the final stages- they sent arbiters to the houses of the players to make sure they do not use outside assistance.

I know at least one GM who quit playing cyber chess due to his frustration over cheating ( on ICC ). This GM was also a candidate for the WC in the mid 80s so I doubt it was just paranoia on his part. ICC has caught even strong GMs cheating in their money events , like Dos Hermanas, some of the GMs caught are famous. Why would such GMs cheat ?! Well, because even they know that they cannot beat those people who are using the strongest programs against them. I think the problem is only growing worse and I have been playing net chess since 1996. The people who dont want/need to cheat are eventually faced with either fighting fire with fire, quitting cyber chess, or just accepting losing to someone who doesnt even know what co-ordinate squares are and cant answer simple questions about the position yet play like a GM.

 Yes, I think you are right. Among the top 100 players here only 5 are titled players ! How is that possible ? There are many titled players here on chess.com who do not play and I think the reason may be that they do not want to face the strong programs that they KNOW they will face.

 Yes it is and is true in other, more serious areas as well. Obesity is a problem in people younger and younger these days and part of the problem is technology. When I was a kid I can vividly recall that our parents had to "make" us come inside because we were always playing outside. Today it seems kids must be made to go outside in many cases ! Ofcourse in the 60s kids didnt have nintendos, computers, their own TV , audio "books"  etc. Technology may indeed be both a blessing and a curse. Perhaps its how we use it that decides which it is ?