Borislav Ivanov and Rapid Chess

[Kenneth Regan]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Mr. Regan (or is it Dr. Regan?) [Dr. Regan]
Do you consider it at all possible that Ivanov, by means of removing himself from any chess activities against human opposition for months at a time, and by during those same months spending considerable man-hours playing chess games against Houdini 3 and Rybka engines, could possibly have learned to play at the level of those engines

I measure the level of those engines and the level of Mr. Ivanov's games in the 3100s (3150 for Veliko Tarnovo, counting all 6 played games), with high confidence above 2950. So the straight answer is No. Hence ditto your two following questions. Overall I repeat what I'm quoted as saying earlier in this thread, and note that the investigation has moved on to some more "forensic" matters having to do with individual games and moves and observer reports.

[D]o you think that the only way to then prove any guilt on Ivanov's part would be for others of similar rating level (~2300 ELO) to undertake his methods for the same amount of time and then attempt to replicate his Houdini-level play against human opposition? I personally would have no idea how many such individuals would have to take such steps, nor what the statistical results would have to be... if N such rated players took the steps, and only 1 of them achieved Ivanov's results, how low would N have to be to prove Ivanov's innocence on cheating? It's a fascinating question!

My January report gives that answer. To find another person able to replicate his results, you would expect to need N = over 1 million such individuals. That is actually the primary meaning of the z-score "odds"; the interpretation as odds against a "null hypothesis" of no-cheating needs further caveats as given in my report. Hence I doubt you would find enough takers for an actual such test. You could, however, look to the entire actual past history of chess as your guide. My report also addresses this.

(Of course, I also considered the possibility of simply having Ivanov himself play against Houdini and Rybka in controlled conditions...

Mr. Ivanov was offered the opportunity to demonstrate his skill, albeit under RF surveillance, which I am told has not been done at any of his tournaments, but did not accept. He did agree to take a polygraph test and be examined this past Wednesday at the federation HQ in Sofia's central sports palace, but he showed up at a tournament in Varna instead, perhaps not realizing that a court had refused to hear his appeal to lift a 4-month suspension. The federation regards the matter as closed until the end of the suspension.

[Louis Morin]

Re : Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Mr. Ivanov showed up at a tournament in Varna instead, perhaps not realizing that a court had refused to hear his appeal to lift a 4-month suspension. The federation regards the matter as closed until the end of the suspension.

Most people here seem to consider that there this whole Ivanov affair is merely a cheating problem with a strong computer.

I do not agree at all. This is a cheating problem with hi-tech devices enabling information (i.e. chess moves) to be sent back and forth with no possibility of detection.

Take this example. The World Open will be held on July 3-8, with a first prize of $13,000 in most sections. So imagine a team of two players who want to win the U1600 section. Player A is rated 1530, shows up for all games and sends all opponent's moves to player B. Player B is a strong expert or a master. Upon reception of each move, he analyses the position, then sends his suggestion to Player A. Of course, chances of Player A winning his section are very real. And you know the best one? Should anybody in the World Open have suspicions and ask for all moves to be computer-checked, nothing will be found. Why? Because no computer was used for cheating! Which means that several people a lot smarter than Mr. Ivanov may have been cheating in tournaments with complete impunity for a very long time.

So I'm afraid that Mr. Ivanov's suspension will solve nothing at all, quite the contrary. And the same can be said for computer-checking of moves, because cheaters can use any computer other than the strongest ones, or even no computer at all, and they can cheat for any given number of moves rather than for all moves. Unless one finds exactly how the cheaters cheat with hi-tech devices, this cheating problem will go on and on, and eventually spread in local tournaments as well. Sooner or later, one will have to take the trouble of finding out the modus operandi of this scheme.

[John Coleman]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

As I wrote before (up thread), I want prefer to see some sort of evidence in the form of a commumication device, or an accomplice. Or both.

[Gary Ruben]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Would you mind posting 3 of the games in pgn format? I'd like to go over them and see what conclusion I get. Also the time control for the games.

[Roger Patterson]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

...

The one unprecedented development I know is robots capable of optically recognizing a chess position on a generic tournament set: one by Alexandra Kosteniuk's father (http://www.chessblog.com/2013/05/one...ch-sergey.html) and one I lost to in Toronto last summer (http://spectrum.ieee.org/tech-talk/r...e-of-the-turk- which links to this BBS). The latter required me to center the pieces, but my CS departmental colleagues tell me that a more-robust system capable of viewing a position and outputting a FEN position code is well within the capacity of an advanced amateur hobbyist.

.....

that would be a pretty cheap and simple replacement for organizers for DGT boards.

[Erik Malmsten]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Would you mind posting 3 of the games in pgn format? I'd like to go over them and see what conclusion I get. Also the time control for the games.

Bulgarian FM Valeri "Tiger" Lilov [and his manager Alex Karaivanov] discusses the games on a video:

2012 Ivanov games
http://www.youtube.com/watch?v=Jr0J8SPENjM

He also looks at tournament results and in a few days or weeks Ivanov performance raised rating by 400 points, with crushing checkmate attacks against 2600 opponents.

2013 games
http://www.youtube.com/watch?v=qhfCU...ture=endscreen

Also a strategy of setting the engine to exchange pieces for a draw with 2200 players.

He recommends organizers buy signal jammers.

[Gary Ruben]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Most of what the speaker said on that 2013 youtube presentation sounded like opinion. I noticed where he mentions losing the first game against even a weak opponent. Assuming it's a swiss format event that would give weaker pairings or good players having a bad event for a few rounds.

I'd like to see the actual games which from which the analysis of cheating came. I want to compare them to my books and computer programs as well as databases. Also to look at the play of the opponents in those games.

Swapping off pieces to make draws or even taking 4 move draws is not against the rules.

I looked at one tournament on his FIDE card and only one in 2013 stands out. By I don't know how strong his opponents really play. We've seen Canadian players play in Europe and take apart the opposition. Under rated players against over rated opponents. It's kind of a grey area.

Signal jammers are an idea if they don't interfere with medical devices like implanted pacemakers and defibrillators. I heard they were working on wireless ones but don't know if they are in use.

If signal jammers were put in use and top players started performing badly what would that tell us? Or if top players wouldn't play in such events.

I wonder if it's an advantage to play someone who thinks you're cheating and somehow using a computer.

[Kenneth Regan]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

I'd like to see the actual games which from which the analysis of cheating came. I want to compare them to my books and computer programs as well as databases. Also to look at the play of the opponents in those games.

Attached are all the games from the 2012 Zadar Open and last month's Old Capital open, plus the one reconstructed accused game from the April Kustendil Rapid. Notice I've given you whole sets and my work is first based on whole sets. I read accusations against subsets of games from other events, but the selection criteria for those games have not been made clear to me, and I have not run full tests on them.

---Ken R.

[Roger Patterson]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

.....

He recommends organizers buy signal jammers.....

If you don't mind going to jail http://www.ic.gc.ca/eic/site/smt-gst...g/sf10048.html

or perhaps causing the death of someone because a 911 call couldn't go through.

[Gary Ruben]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Thanks for posting the games. I'll have a look at them over the next week. On a fast look at the rapid game with Georgiev (I assume the GM), the event doesn't show as being rated by FIDE. Also he's close to 60. Maybe if it wasn't rated he didn't take it very seriously.

The game itself, a Trompowsky, goes into a line I wouldn't expect. 3. Bxf6 gxf6 is not what I'd play. Black weakened his king side and it's often more difficult for him to develop. 3. ... exf6 is more what I'd expect.

It's important to look at the play of the opponent because often a loss comes from a bad plan in the opening.

[Felix Dumont]

Re : Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

3... gxf6 is perfectly playable. Black only mistake is to play against a computer. He obviously had no chance. White keep playing lots of extremely weird moves that no human would ever play (in fact, he's the one to get out of the opening book).

[Paul Bonham]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Most people here seem to consider that there this whole Ivanov affair is merely a cheating problem with a strong computer.

I do not agree at all. This is a cheating problem with hi-tech devices enabling information (i.e. chess moves) to be sent back and forth with no possibility of detection.

Take this example. The World Open will be held on July 3-8, with a first prize of $13,000 in most sections. So imagine a team of two players who want to win the U1600 section. Player A is rated 1530, shows up for all games and sends all opponent's moves to player B. Player B is a strong expert or a master. Upon reception of each move, he analyses the position, then sends his suggestion to Player A. Of course, chances of Player A winning his section are very real. And you know the best one? Should anybody in the World Open have suspicions and ask for all moves to be computer-checked, nothing will be found. Why? Because no computer was used for cheating! Which means that several people a lot smarter than Mr. Ivanov may have been cheating in tournaments with complete impunity for a very long time.

So I'm afraid that Mr. Ivanov's suspension will solve nothing at all, quite the contrary. And the same can be said for computer-checking of moves, because cheaters can use any computer other than the strongest ones, or even no computer at all, and they can cheat for any given number of moves rather than for all moves. Unless one finds exactly how the cheaters cheat with hi-tech devices, this cheating problem will go on and on, and eventually spread in local tournaments as well. Sooner or later, one will have to take the trouble of finding out the modus operandi of this scheme.

A point well raised, Louis. What the advent of computer engines has done, then, is to expose this entire question of wireless transmission of moves. FIDE must eventually, and soon, change the course of chess history. The pure skill aspect of chess, once its proud hallmark, is now its undoing. Luck and randomness MUST be introduced to some degree, and if that degree is to be minimized yet have the desired effect, Option Chess is one of many potential and practical solutions. Kevin Pacey's Throne Chess is another, but leaves open the possibility of middlegame cheating where Thronemate is not a factor (see Kevin's blogs if you don't know what Throne Chess is).

The scheme you describe is good for a one-off or maybe a few events, but eventually succumbs to the Law of Diminishing Returns. In fact, it starts with Diminishing Returns because a prize now has to be split betwee 2 people, and Player B is presumably giving up any chance of winning his / her own section prize to help Player A. The likelihood of this happening has to be low, but not zero. But as Player A moves up in the rankings, Player B becomes less useful and needs replacing with a better Player B, who may demand a higher return. So I think we can safely say Carlsen isn't at the top of the rankings via this scheme.

Regarding "cheaters can use any computer other than the strongest ones... and they can cheat for any given number of moves rather than for all moves". This in fact is (further) evidence that Ivanov may not be cheating at all, but instead has used an intensive learning process to emulate Houdini over the board. Why would be cheat on all moves, knowing this would more deeply incriminate him? It's very possibly because he's not cheating, and this may also be why he continues to play fearless of being caught. If he wants to keep playing, he will have to succumb to test conditions (signal jammers, Faraday cage, whatever), and my interest in this whole affair is to see whether he might confound everyone by still playing at Houdini strength. Only time will tell.

[Gary Ruben]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

Felix, I admire that you have made up your mind. In fact, everything is playable depending against whom you wish to play it.

It remains the theory I have is that in my book, a paper book, says black weakens his king side and it's often difficult for him to develop.

So we have the opening, the opinion of my book on the Trompowsky, and the game itself in which black finds white attacking the weak kingside (like the book suggests) while nursing a passed center pawn. I still have to check how he got to the end position.

[Garland Best]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

It's simpler to obtain RF detectors anyhow, sweep the players for "bugs". http://www.spyshops.ca/Bug&CameraDetectors.html

I'm surprised this hasn't been tried.

[Paul Bonham]

Re: Re : Re: Re : Re: Re : Re: Re : Re: Re : Borislav Ivanov and Rapid Chess

I measure the level of those engines and the level of Mr. Ivanov's games in the 3100s (3150 for Veliko Tarnovo, counting all 6 played games), with high confidence above 2950. So the straight answer is No. Hence ditto your two following questions. Overall I repeat what I'm quoted as saying earlier in this thread, and note that the investigation has moved on to some more "forensic" matters having to do with individual games and moves and observer reports.

...To find another person able to replicate his results, you would expect to need N = over 1 million such individuals. That is actually the primary meaning of the z-score "odds"; the interpretation as odds against a "null hypothesis" of no-cheating needs further caveats as given in my report. Hence I doubt you would find enough takers for an actual such test. You could, however, look to the entire actual past history of chess as your guide. My report also addresses this.

Thank you, Dr. Regan, for responding. Many people on this board have me on their "ignore list" because I tent to ask questions and raise points that make them feel uncomfortable. When I mentioned the hypothetical defence lawyer, I thought it might prevent you from any kind of response, but it needed mentioning, because (unfortunately) it might all come to that. Stranger things have happened.

I respect greatly the work you have done and continue doing. Back in 2006, I started thinking about a move-based chess rating system. I even wrote up a few things about it. Much later I came across your Intrinsic Rating system (hope I haven't botched the name), and I smiled to myself and thought, of course, someone well advanced in the CS field would already have thought of this.

When I read through your pdf document on the Ivanov case, I was encouraged by how you were apparently trying to cover every possible bias and not let it color the results. I do believe you are an objective and fair researcher.

But in your response it seems (to my perhaps uneducated mind) that the particular bias I was introducing is being dismissed. For example, you mention measuring the level of the engines and of Ivanov's games with high confidence, and then just say "the straight answer is No" to my first question. But I'm raising the objection that your measurements aren't really what's important. I'm certainly not disputing the calculations. I'm agreeing that Ivanov's Intrinsic Ratings for the games in question may be on par with Houdini's and Rybka's.

What I'm saying is that there is this possibility sitting out there that Ivanov is a new type of human chess player, one who possibly learned from intensive training against Houdini / Rybka how to behave exactly like them. Ivanov claims to have done this intensive training. Let's assume for agrument's sake that this claim can be proven correct, i.e. he did the training just as he claims.

How can any of your measurements have any bearing on Ivanov's innocence / guilt on the cheating charges if the only thing your measurements confirm is that yes, he played at Houdini level? The hypothetical defence lawyer will agree with you: "Yes, Ivanov was playing at that level".

"But," the lawyer will continue, "He learned to play at that level, by means of unprecedented and intensive training against Houdini."

But you say your straight answer is that there is no possibility of this. What I'm left wondering is, how can any statistical method argue against that if there are no statistics on it? No one but Ivanov has performed his specific intensive training (another assumption perhaps, but a fairly safe one unless you know of specific people who have replicated his steps).

Later, you respond that something like 1 million players would have to replicate Ivanov's steps to produce 1 more Ivanov. Given that no one else has (presumably) actually undertaken Ivanov's steps, I can only assume you are extrapolating this from, as you mention, the entire past history of chess.

But the entire past history of chess is mostly devoid of computer engine technology. No one until the mid-2000's has had access to engines of Houdini's capabilities. Of those, has any single serious chess player besides Ivanov taken it upon himself / herself to drop everything, stop playing chess against humans, start playing chess intensively and daily against only Houdini / Rybka, continue this for months on end (how many months I don't know, but let's assume at least 3) with the goal of learning to play like Houdini?

You see, this is the crux of what the defence lawyer may use against your evidence. Your methods are fine for saying "Player so-and-so played at an Intrinisic Rating level of X over N number of games." But, the lawyer will argue, your methods cannot say anything as to whether Player so-and-so LEARNED to play at that Intrinsic Rating level via some new means that has become available (in this case, intensive training against Houdini / Rybka). And so, given a fair and objective judge, I believe such a lawyer can have your evidence dismissed, if Ivanov were to claim to have learned to play at that level by those means. If no evidence of electronic cheating comes forth, the onus would be on the prosecution to prove Ivanov could not have learned to play like Houdini.

I certainly am willing to be corrected and educated if there is a flaw you can point to in my argument. By the way, if you'd prefer not to hash this out publicly on this forum, please do PM me and we can discuss it by email and I will keep it private.