Borislav Ivanov: a Lilov's add-on

[Hood]

ad b) the example with result of analysis: Capablanca using Stockfish! It is the case which makes that methodology false.

That is indeed a major point in this matter. Of course the inquisitors will ignore it.

Since the accusers are chessplayers themselves, and not weak ones, someone should really analyse their games and then confront them with the many, many engine agreements which will be found (except they are patzers which they are not). Then ask them for explanation.

No, you need to find "hard facts" like e.g. (1) an accomplice who transmitts computer moves to a hidden (2) radio device in B.I.'s ear. You need to find both, then you have something. Otherwise, you have nothing.

Isn't it funny that only patzers, as you call them, can suspect and confront the supposed "cheating grandmasters"? Of course, no strong, titled players are accusing grandmasters of cheating these days, but they are accusing B. Ivanov, because that's where the real incriminating correlation is found. Unfortunately, according to you this is "nothing", because you decided to pretend it is not evidence, just because you felt like it.

Think about it -- if titled players are cheating too, going by the same type of evidence that made Ivanov suspect, why isn't there an uproar about their cheating? After all, when GM Feller was cheating he wasn't allowed to get away with it. So it seems that fellow GMs don't like it when other GMs or strong players cheat. What's stopping them now? According to your theory, these supposed "cheating strong players" are just as guilty as Ivanov, but they're all getting away with it, suddenly (!!), while only poor Ivanov is picked on arbitrarily (yeah, right).

In the Feller case there were witnesses of sms's concerning his games
and persons involved in transmitting moves.

[Terry McCracken]

Don, I love your avatar....Let's Play Rex Chess! Better yet, rename Komodo to T-Rex Chess! or T-Rex ver.

[Don]

Because the methodology used by Lilov is false as the logic proves.
Only one thesis we can get from analysis is that there is a coincidence
between moves of BI and some program.
When we would analyse the games of other players we would find similar
coincidences.

What logic is false? He played many games identical to Houdini so he was with very high probability using Houdini.

[Don]

Don, I love your avatar....Let's Play Rex Chess! Better yet, rename Komodo to T-Rex Chess! or T-Rex ver.

Yes, we should go back to Rex!!

Don

[Don]

Don, I love your avatar....Let's Play Rex Chess! Better yet, rename Komodo to T-Rex Chess! or T-Rex ver.

The thing is very well done, it looks quite real.

[Mike S.]

I think there was a misunderstanding. I do not at all suppose that GMs are cheating. The idea is that you'll of course get many engine agreements with any (very) strong and precisely played game, including some of Fischer's or Capablanca's, even.

By applying that flawed method to themselves, the GM accusers would have trouble to explain their own "Houdini moves". Because so far that is their major argument.

It is also flawed for another reason: Even Houdini does not agree to 100% with Houdini. In very many positions, there is not just one particular Houdini move. Every computer chess fan knows on how many factors (time,speed,depth,hash,cores...) the move decision depends, and that especially mp results are sometimes not even reproducable under the very same circumstances.

If throughout an analysis of reasonable length the engine would e.g. change the pm twice, we'd have three different "Houdini moves", not just one. Now let's imagine a suspected player has chosen one of these. Wow. How meaningful is that?

[Terry McCracken]

I think there was a misunderstanding. I do not at all suppose that GMs are cheating. The idea is that you'll of course get many engine agreements with any (very) strong and precisely played game, including some of Fischer's or Capablanca's, even.

By applying that flawed method to themselves, the GM accusers would have trouble to explain their own "Houdini moves". Because so far that is their major argument.

It is also flawed for another reason: Even Houdini does not agree to 100% with Houdini. In very many positions, there is not just one particular Houdini move. Every computer chess fan knows on how many factors (time,speed,depth,hash,cores...) the move decision depends, and that especially mp results are sometimes not even reproducable under the very same circumstances.

If throughout an analysis of reasonable length the engine would e.g. change the pm twice, we'd have three different "Houdini moves", not just one. Now let's imagine a suspected player has chosen one of these. Wow. How meaningful is that?

And if Ivanov played and beat the top 10 in the world would you still be playing Devil's Advocate for him?

[Mike S.]

Ask IM Andrew Martin. (He had a look at at least two of the suspected games too; see YouTube.)

Anyway, I am not that much interested in the matter itself. I don't know if Ivanov is cheating but also, I cannot claim the opposite. I am no chess professional so I can remain cool about this anecdote.

I just think that this method with claiming to have found "100% Houdini moves" (absurd!) in a master game is flawed, and also rather meaningless on master level. We should get rid of that.

Analysing for evaluation difference (preferably over many games) is something much different. I did that recently, but for 10 games from different players only, as a kind of precision meter using a threshold of 25 centipawns:

http://www.talkchess.com/forum/viewtopic.php?t=48463

(Maybe a smaller threshold is considerable, sometimes.)

[Don]

I think there was a misunderstanding. I do not at all suppose that GMs are cheating. The idea is that you'll of course get many engine agreements with any (very) strong and precisely played game, including some of Fischer's or Capablanca's, even.

By applying that flawed method to themselves, the GM accusers would have trouble to explain their own "Houdini moves". Because so far that is their major argument.

I think the primary argument is his incredible and sudden increase in performance - and that is what caused the initial suspicion. It was NOT the move matching.

Every player plays "Houdini moves" so what has to be measured is the frequency of these over many games. There is no GM accuser who plays like Houdini. To say you play like Houdini doesn't mean you play a few moves that Houdini would, it means that you play all your moves like Houdini would.

Houdini doesn't always play exactly the same move in any position because chess program are not deterministic, especially in time control games and even more so with MP. But Houdini still plays like Houdini. To play like Houdini, you have to play a move that Houdini "could have" played and that is why if you do an analysis you need to look at any move that Houdini may have played - I think they were going by the top 2 moves that Houdini might have played.

I want to see for myself before making a judgment on the move matching but I don't think the empirical proof of his cheating requires this. The GM's know he is cheating and they are highly tuned judges when it comes to spotting anomalies like Ivanov as they have years of experience and they have seen it all. Some of us making these judgement have no clue about tournament chess and especially chess at high levels and we are just pretending we are experts giving our naive opinions.

It is also flawed for another reason: Even Houdini does not agree to 100% with Houdini. In very many positions, there is not just one particular Houdini move. Every computer chess fan knows on how many factors (time,speed,depth,hash,cores...) the move decision depends, and that especially mp results are sometimes not even reproducable under the very same circumstances.

If throughout an analysis of reasonable length the engine would e.g. change the pm twice, we'd have three different "Houdini moves", not just one. Now let's imagine a suspected player has chosen one of these. Wow. How meaningful is that?

As I said, I want to conduct my own study and relate it to many players - to see if I can construct a test. In the analysis they were doing it is said that he matched Houdini's top 2 choices every time. Presumably by looking at the top 2 you cover most of the cases of play that is not 100% deterministic but I want to see that for myself.

[Uri Blass]

I think that you miss the main evidence against ivanov and it is possible to find that he is quilty without a reasonable doubt even without looking at his games only based on his strange results(the only doubt based on his results is if he is quilty of cheating to get good results or quilty of losing games on purpose).

The main point is that ivanov is a weak player based on his worse tournaments in 2012-2013 and in some tournaments he has performance that is worse than 2000.

see
"35th International Open Bulgarian Chess Championship "Georgi Tringov Memorial" from february 2013

http://ratings.fide.com/individual_calc ... -03-01&t=0

He scored only 1 out of 5 against players with rating above 1942 when all of them have rating below 2100.

see
http://ratings.fide.com/individual_calc ... -07-01&t=1

He scored only 2 out of 4 against average belowe 2000.



On the other hand we have in the 19th International Tournament Zadar Open 2012 - Group A excellent performance of more than 2600 that include beating some GM's

http://ratings.fide.com/individual_calc ... -01-01&t=0

Other tournaments with strange results:

http://ratings.fide.com/individual_calc ... -12-01&t=0

2 draws against weak players rating of 2005 and 1875 and later 4 wins when 2 of them are against 2394 and 2411.

another strange result

http://ratings.fide.com/individual_calc ... -07-01&t=0

winning against the strong players and losing against the weak players.