FM Borislav Ivanov Disqualified

The science of statistics is designed to draw conclusions about populations and samples, not individuals.

And the games by Borislav Ivanov form a sample, which is a valid object for statistical analysis.

True ... but there are questions that can't be adequately addressed by that sample, one of which is, "Does he cheat?" If Ivanov is a true rarity, you end up drawing the wrong conclusion.

EDIT: You could use the sample of his games forensically, but it would have to be a designed experiment. 'Making the engine-recommended move most of the time' is not compelling, but 'always duplicating the same error' is more convincing.

You could understand that a 2600 performs around 2400-2700

but it would be too difficult to believe that somebody can perform at 1900 and at 2700.

I beg to differ, his games form an adequate statistical sample. What are the chances he plays like a machine in a sharp position against a 2600 GM, and then like a patzer in a much simpler position? Probably not much higher than zero.

No, that question is perfectly normal. Taking into account all the chess players on Earth, it becomes ovious that one with a rating performance of 1900 cannot end up with a rating performance of 2300+ in 10 days. No matter how hard you train or practice, that just won't happen. The human brain is not a computer that can be programmed to play with Grandmaster strength in a flash.
Sure, geniuses do happen (Magnus Carlsen), but they still can't play with the accuracy that computers play. Check out some games from the recent Grandmaster Supertournament 13'. You will see that even grandmasters make mistakes at times. That doesn't happen with Borislav - his moves match with the engine's almost 100%. You won't see any blunders in his game, unless, of course, he decides to shut off his engine and get back to his original 1900 strength.
We know this due to statistical analysis. The rating points curve slowly, because a human player improves slowly. If there is a rocketing curve, it means that the human player didn't improve slowly, but, rather, changed dramatically (became Houdini or etc.). That is a cause for concern and suspicion.

Regardless of rate of improvement, humans simply cannot start to think like can engine (comb through many candidate moves at 18ply in half a minute). Even at the highest level, human chess is based on a number of logical parameters, and practical decisions (thus, human games don't look anything like those betweem engines).

Some people just can have off days, I know I do it! Just because he's good some of the time doesn't mean he is cheating at that time! If you can show me the device, then I'll be convinced. Otherwise, you could just be punishing a player for have some particularly good games!

Agreed. Statistical analysis used to flag an improbable pattern for investigation. But coincidence is expected in statistics. It's a science of probability, not absolutes ... that's why a competent statistician will not use statistical methods to draw conclusions about an individual.

This is not just a simple coincidence. Gaining 500 rating points in 10 days can not be a coincidence. Statistically, the probability is so much close to zero, that we can already make conclusions (i.e. he is using some forms of cheating).

Theory of statistics proves that he is cheating - this is how chess.com finds and bans the cheaters on this site. For FIDE, this is seemingly not enough. We need practical proof - we'll have to catch him with a cheating device. But both proofs are pretty much valid and can be treated as undeniable forms of evidence. Miracles don't happen, sorry.

*facepalm*

Going into centaur tournaments and scoring well here does not invalidate previous or future silicium-free CC performance.

If you cannot understand that some people are much better at CC than at rapid chess, you can accuse this guy of cheating too. Pretty lame cheater, who reached a drawish position against a 1932.

You seem to not understand the science. Treating it as proof of an individual case is incorrect. You can only make a probability statement, with the proviso that the probability is only valid if all modeling assumptions are correct.

I've been a professional statistician for nearly 30 years, and this particular aspect of statistics, misapplication to the individual, comes up frequently. It's a science of large numbers, not a science of one.

The probability is only considered if it is relevant. In the case of Borislav you can, with great confidence, state that he has cheated, because his rating has increased dramatically over a short period of time. The probability of such a phenomenon is too small to be taken into account. It produces a simple conclusion that he was cheating.

Statistics is just as valid as any other science. I don't care how many years you have worked with it. We are not basing our conclusions about an individual on nothing, we are basing them on a big number of games, that were played against other people. It is a completely valid decision to make conclusions about a player from his games:

A 1500 player has beaten five 1600 players, but lost to a 1700 player, Conclusion: he has improved and his playing strength is that of a 1600 player.

A 1900 player has beaten six grandmasters (2600+), but lost to a couple of 2100 players.
Conclusion: Obvious cheating.
 

Individual case is different from individual person.

Computer match rates on different games, as well as blood tension taken at different days, do constitute multiple samples although they are taken on the same person.

If Ivanov was consistently performing at 98% match, although incredible, you could argue that he is just very computer-like in thinking, and that the individual characteristics account for the high percentage - as old age could account for not-population-average blood tensions.

But when the match percentage has two peaks, one at the 1900-level-statistical-match mark and another one over everything any other player ever did, it's just impossible to be "off days". If he had off days, he would perform more often in the middle of his two ratings.

Now, of course that's a probability and not a certitude. But when it comes under 10^-20, it's hard to defend it comes "by chance", and it's reasonable to say it's "impossible" and act accordingly.

Not that it is illegal, it is simply impossible for a human being. If someone plays like an engine once or so, you can say that he has had a lucky day. But if he play like an engine for a few days, than starts playing like a patzer, then like an engine again... Patzers cannot become engines, and engines don't become patzers. Not that quickly.

Let's all vote Marvin the ultimate-expert-in-all-fields and find something else to talk about.

Its quite telling that the ones defending Ivanov are themselves weak players , no masters defend him that I am aware of. 

I'm not sure why they just didn't set him up... let him enter a tourny and half way through a game that looks suspicious (maybe have someone putting his moves into houdini) run a metal detector wand over him while watching that he doesn't turn off any devices. 

I am not defending Ivanov (chances are he was cheating) but this issue would be lot clearer if he was actually caught.