English Chess Forum

Chris Rice wrote:Round 6 White: Ivanov (2303) - Black: Georgiev, Kiril (2657)

1. d4 Nf6 2. Bg5 d5 3. Bxf6 gxf6 4. Nf3 c5 5. e3 Nc6 6. c4 dxc4 7. Bxc4 cxd4 8.Nxd4 Nxd4 9. exd4 Bg7 10. Nc3 O-O 11. O-O f5 12. d5 Qd6 13. Re1 Bd7 14. Qe2 Rfe8 15. Bb5 Rac8 16. Rad1 a6 17. Bxd7 Qxd7 18. Rd3 Rc4 19. h3 b5 20. a3 Qd6 21. Qh5 Qg6 22. Qd1 Qd6 23. Rg3 Kh8 24. Qh5 Rf8 25. Qxf5 Rf4 26. Qg5 Qf6 27. Qh5 Rf5 28. Qg4 Rf4 29.Qd7 Qd4 30. Rge3 Qd2 31. Ne4 Qxb2 32. g3 Rxe4 33. Rxe4 Bf6 34. Rxe7 Bxe7 35. Qxe7 Kg8 36. d6 Qxa3 37. Qg5+ Kh8 38. d7 1-0

So did Black play poorly, did White play well? Some commentary is really needed. I can accept that some players can play much better at faster time controls than they do at other time controls. Also, how do you compare human blitz or rapidplay play to computer play? Is that the right approach to take here or not? And are all the games available? A reminder that the time control used was 10 mins + 5 secs increment.

Christopher Kreuzer wrote:

Chris Rice wrote:Round 6 White: Ivanov (2303) - Black: Georgiev, Kiril (2657)

So did Black play poorly, did White play well? Some commentary is really needed. I can accept that some players can play much better at faster time controls than they do at other time controls. Also, how do you compare human blitz or rapidplay play to computer play? Is that the right approach to take here or not? And are all the games available? A reminder that the time control used was 10 mins + 5 secs increment.

Not sure I'm really qualified but it looked like an error free game from White for sure. It would be interesting and perhaps gives us a clue as to whether he is cheating or not to see Ivanov playing to see if there are any hesitations or changes of mind when he moves the pieces. One thing I did notice from the game was move 23 where he played Rg3 when he could have repeated the position with 23 Qh5. How many players of his "strength" would decide to go for broke against a 2657?

Chris Rice wrote:

Christopher Kreuzer wrote:

Chris Rice wrote:Round 6 White: Ivanov (2303) - Black: Georgiev, Kiril (2657)

So did Black play poorly, did White play well? Some commentary is really needed. I can accept that some players can play much better at faster time controls than they do at other time controls. Also, how do you compare human blitz or rapidplay play to computer play? Is that the right approach to take here or not? And are all the games available? A reminder that the time control used was 10 mins + 5 secs increment.

Not sure I'm really qualified but it looked like an error free game from White for sure. It would be interesting and perhaps gives us a clue as to whether he is cheating or not to see Ivanov playing to see if there are any hesitations or changes of mind when he moves the pieces. One thing I did notice from the game was move 23 where he played Rg3 when he could have repeated the position with 23 Qh5. How many players of his "strength" would decide to go for broke against a 2657?

In a rapidplay game? Probably quite a few. My impression of the game (though I'm not qualified to comment either) is limited to wondering whether White won that f5 pawn, or whether Black gave it up for counterplay, and whether White defended accurately (clearly he did) or whether Black missed some better lines. The impression given is that White stayed in control, and after winning that pawn won with some ease, while Back flailed, collapsed and lost without much resistance. Without knowing, I would have assigned the ratings the other way round...

I've read online that White played 73% 1st choice Houdini-3 moves, and took no more than 10 secs over each move.

Michael Bennett wrote:I've read online that White played 73% 1st choice Houdini-3 moves, and took no more than 10 secs over each move.

73% isn't that high really. If you were just relaying engine moves, you would get 100% provided the comparison engine was identical to the one in use. I suppose you could have a hypothetical customised interface that selected randomly from a list of top choices. At ten seconds a move, you've got the issue, that if he is using an engine during the game, as to how the moves are communicated to the engine in that period of time.

The alternative hypothesis remains that "he got good", probably by intensive training with computer engines and it would appear a highly developed ability to second guess them. In the Trompovsky game, the position reached with that pawn on f5 is reasonably generic, so the idea that a player would investigate with a computer engine before the game, the best methods of play, is plausible.

Roger de Coverly wrote:

Michael Bennett wrote:I've read online that White played 73% 1st choice Houdini-3 moves, and took no more than 10 secs over each move.

73% isn't that high really. If you were just relaying engine moves, you would get 100% provided the comparison engine was identical to the one in use. I suppose you could have a hypothetical customised interface that selected randomly from a list of top choices. At ten seconds a move, you've got the issue, that if he is using an engine during the game, as to how the moves are communicated to the engine in that period of time.

The alternative hypothesis remains that "he got good", probably by intensive training with computer engines and it would appear a highly developed ability to second guess them. In the Trompovsky game, the position reached with that pawn on f5 is reasonably generic, so the idea that a player would investigate with a computer engine before the game, the best methods of play, is plausible.

73% 1st-choice moves is still a slight increase from Zadar, the event that kicked it all off (see page 7 of this thread.) The key was that almost 95% there were top-4 choices.

Also, perhaps in the tactically sharp positions Ivanov seems to get into, Houdini at 10 seconds per move may not suggest the same moves as it would after a longer time? I'm not sure if the investigation reported by Michael takes this into account or not.

To do anything with engine comparisons you have to track it quite carefully over time etc. Plenty of games needed. There are a fair number by now perhaps.

It would then help with plausability rather a lot if you could dream an even semi plausible method for cheating that would have a chance of working at this sort of time limit..... Stuff like external operators would I think pretty clearly take too long.

MartinCarpenter wrote:T
It would then help with plausability rather a lot if you could dream an even semi plausible method for cheating that would have a chance of working at this sort of time limit.

Magic glasses with some form of head up display would help with the move notification. How the engine finds out the board position is the issue. If they are playing with DGT wireless boards, it might be possible to hack into the moves being broadcast. But are these boards really that common in the events he plays in?

Roger de Coverly wrote:

MartinCarpenter wrote:T
It would then help with plausability rather a lot if you could dream an even semi plausible method for cheating that would have a chance of working at this sort of time limit.

Magic glasses with some form of head up display would help with the move notification. How the engine finds out the board position is the issue. If they are playing with DGT wireless boards, it might be possible to hack into the moves being broadcast. But are these boards really that common in the events he plays in?

I don't think they used digital boards at Plovdiv and I haven't seen him wearing glasses either in any of the photos. There is one scenario I can think of that's reasonable though it would need outside help. If a camera had a good shot of the board (perhaps one of his supposed paramours or bodyguards observing the game with a hidden camera) then someone externally could be feeding him the moves through a simple earpiece. To do it at 10 seconds a move should be possible.

I wonder if they could get him to publicly annotate his games afterwards, then we might get more of an idea of how good he really is.

The following analysis was posted in the comment section on http://chess-db.com/public/game.jsp?id= ... 7.21940736

Anonymous wrote:Settings were 12-22 ply depth/45 seconds, hash 256mb
Engine Houdini 103a
Core2Duo 2.00 GH
{ Ivanov, Borislav (Games: 1) }
{ Top 1 Match: 25/34 ( 73.5% )
{ Top 2 Match: 33/34 ( 97.1% )
{ Top 3 Match: 34/34 ( 100.0% )
{ Top 4 Match: 34/34 ( 100.0% )

{ Georgiev, Kiril (Games: 1) }
{ Top 1 Match: 17/33 ( 51.5% )
{ Top 2 Match: 26/33 ( 78.8% )
{ Top 3 Match: 27/33 ( 81.8% )
{ Top 4 Match: 27/33 ( 81.8% )

AS good as that game seems, it seems to me to be a rather similar style to Capablanca - Simple yet effective. A style anyone could strive for and possibly attain. Playing error-free in complex positions is more disturbing than this game IMO - blitz or not.

Chris Rice wrote: There is one scenario I can think of that's reasonable though it would need outside help. If a camera had a good shot of the board (perhaps one of his supposed paramours or bodyguards observing the game with a hidden camera) then someone externally could be feeding him the moves through a simple earpiece. To do it at 10 seconds a move should be possible.

If a camera was concealed about his person or in his clothes, that's one less assistant needed. Resources of chess events don't generally run to having a nurse or doctor, so an examination for hidden earpieces isn't likely to be possible. If the method relies on a third party supporter keying moves, perhaps trying to unmask them is an approach as well as identification of the signalling transmissions.

I'm assuming he doesn't wear headwear whilst playing. That would be an obvious place to conceal a camera and earpiece.

Eoin Devane wrote:The following analysis was posted in the comment section on http://chess-db.com/public/game.jsp?id= ... 7.21940736
{ Top 3 Match: 34/34 ( 100.0% )
{ Top 4 Match: 34/34 ( 100.0% )

That 100% starts to look suspicious. Assume a hidden camera, hidden earpiece and an assistant with an engine and microphone. The assistant has to respond within 10 seconds. Depending on the interface and settings, engines will give a first suggestion almost immediately. Whilst searching to a greater depth may refine the first suggestion, unless there's a horizon effect, unlikely in a relatively simple position, the first suggestion is likely to remain valid as a top four.

Drop the assistant I think and transmit things to/from some remote system automatically. A human assistant inevitably needs a bit of time to en/decode the signals etc and I doubt if there's really time for that here.

Although automatically interpreting camera shots of chess boards isn't a remotely trivial task.

MartinCarpenter wrote: interpreting camera shots of chess boards isn't a remotely trivial task.

Good point. Perhaps that might explain why he lost in Round 4

http://chess-results.com/tnr95086.aspx? ... 821&snr=10