Fritz10 in correspondence chess question

[Marc MP]

...

If rybka3 plays perfect chess in correspondence chess then the best that they can do is to draw the game and finding a line that rybka does not play perfect is not an easy job because you have no time to predict all rybka's moves and rybka may blunder at fast time control(1 hour per move) but not play the moves that you expect her to play at 24 hours per move.

Uri

Why would Rybka 3 play perfect by correspondence? Think about it, Rybka 4 is going to be at most equal to or clearly weaker than Rybka 3 in this case. That doesn't make any sense, does it?

[Tony Thomas]

...

If rybka3 plays perfect chess in correspondence chess then the best that they can do is to draw the game and finding a line that rybka does not play perfect is not an easy job because you have no time to predict all rybka's moves and rybka may blunder at fast time control(1 hour per move) but not play the moves that you expect her to play at 24 hours per move.

Uri

Why would Rybka 3 play perfect by correspondence? Think about it, Rybka 4 is going to be at most equal to or clearly weaker than Rybka 3 in this case. That doesn't make any sense, does it?

I thought that with enough time and proper analysis, the elo advantage diminished..

[Marc MP]

...cannot use more than some hours of computer time per move.

Uri

This is the oldest and the worst possibility how to play correspondence games. Only wasting time (at the end for both sides). Horizont effect still exist.

Hi Jiri,

It getting hard to explain simple things to some members in this forum. Not enough chess knowledge and too frequent use of a computer to analyse positions leads them toward thinking that Rybka 3's play will be unimprovable on recent hardware. I don't think either that the "horizon effect" will dissapears with Rybka 3. I've not seen any piece of evidence that it will. I personnally don't expect much improvement in this regard - i.e. in almost any closed position involving slow build-up. But for open/semi-open position, I expect a good improvement.

PS: Did you have other successes after your third place in the 5th PAL/CSS freestyle ?

[Marc MP]

...cannot use more than some hours of computer time per move.

Uri

This is the oldest and the worst possibility how to play correspondence games. Only wasting time (at the end for both sides). Horizont effect still exist.

Hi Jiri,

It getting hard to explain simple things to some members in this forum. Not enough chess knowledge and too frequent use of a computer to analyse positions leads them toward thinking that Rybka 3's play will be unimprovable on recent hardware. I don't think either that the "horizon effect" will dissapears with Rybka 3. I've not seen any piece of evidence that it will. I personnally don't expect much improvement in this regard - i.e. in almost any closed position involving slow build-up. But for open/semi-open position, I expect a good improvement.

PS: Did you have other successes after your third place in the 5th PAL/CSS freestyle ?

I want to add that I consider that Uri has sufficient (by far, I read on this forum recently that Uri's OTB elo is close to 2100) chess knowledge for what I mentionned above. What I find puzzuling is that he considers the possibility that Rybka 3 might play perfect in correspondence chess.

From what I read from the Rybka's forum, is that Rybka 3 on a Skulltrail (that is about 2.5 times stronger than a Q6600) reaches depth 27-28 on average after 5 days. To get the horizon effect removed, you need to reach depth =100 (and over for intelligent planning) to avoid 50 moves draws.

That is not going to change with Rybka 3. Human Grand Masters still have very many years to live before their knowledge about these types of positions being overwhelmed by computer analysis.

[Uri Blass]

...cannot use more than some hours of computer time per move.

Uri

This is the oldest and the worst possibility how to play correspondence games. Only wasting time (at the end for both sides). Horizont effect still exist.

Hi Jiri,

It getting hard to explain simple things to some members in this forum. Not enough chess knowledge and too frequent use of a computer to analyse positions leads them toward thinking that Rybka 3's play will be unimprovable on recent hardware. I don't think either that the "horizon effect" will dissapears with Rybka 3. I've not seen any piece of evidence that it will. I personnally don't expect much improvement in this regard - i.e. in almost any closed position involving slow build-up. But for open/semi-open position, I expect a good improvement.

PS: Did you have other successes after your third place in the 5th PAL/CSS freestyle ?

I want to add that I consider that Uri has sufficient (by far, I read on this forum recently that Uri's OTB elo is close to 2100) chess knowledge for what I mentionned above. What I find puzzuling is that he considers the possibility that Rybka 3 might play perfect in correspondence chess.

From what I read from the Rybka's forum, is that Rybka 3 on a Skulltrail (that is about 2.5 times stronger than a Q6600) reaches depth 27-28 on average after 5 days. To get the horizon effect removed, you need to reach depth =100 (and over for intelligent planning) to avoid 50 moves draws.

That is not going to change with Rybka 3. Human Grand Masters still have very many years to live before their knowledge about these types of positions being overwhelmed by computer analysis.

note that I do not think that rybka3 is going to play perfect in every position in chess but I think that rybka3 may play perfect in big part of the openings if it plays against perfect play.

There is a difference.
If chess is a draw then you only need to keep the draw to play perfect chess against the perfect player.

Playing perfect in equal position is often easier job then playing perfect in
endgame that is better for one side when it is not easy to win but possible to win.

I believe that in many positions rybka may play perfect simply because there is nothing to find in them.

tactical positions when 28 plies are enough to find a forced draw for both sides may be one example but closed positions may be another example and it may be possible that rybka can find the best moves thanks to better evaluation together with deep search.

It is simply impossible to win in many closed positions and rybka at 28 plies may be careful enough not to allow the opponent to win regardless of the color that rybka plays.

Uri

[Marc MP]

...

If rybka3 plays perfect chess in correspondence chess then the best that they can do is to draw the game and finding a line that rybka does not play perfect is not an easy job because you have no time to predict all rybka's moves and rybka may blunder at fast time control(1 hour per move) but not play the moves that you expect her to play at 24 hours per move.

Uri

Why would Rybka 3 play perfect by correspondence? Think about it, Rybka 4 is going to be at most equal to or clearly weaker than Rybka 3 in this case. That doesn't make any sense, does it?

note that I do not think that rybka3 is going to play perfect in every position in chess but I think that rybka3 may play perfect in big part of the openings if it plays against perfect play.

There is a difference.

Then you should say it. It is completely different then. And we wouldn't argue. You are usually very strict for that Uri. That is why I said so.

[Uri Blass]

...

If rybka3 plays perfect chess in correspondence chess then the best that they can do is to draw the game and finding a line that rybka does not play perfect is not an easy job because you have no time to predict all rybka's moves and rybka may blunder at fast time control(1 hour per move) but not play the moves that you expect her to play at 24 hours per move.

Uri

Why would Rybka 3 play perfect by correspondence? Think about it, Rybka 4 is going to be at most equal to or clearly weaker than Rybka 3 in this case. That doesn't make any sense, does it?

note that I do not think that rybka3 is going to play perfect in every position in chess but I think that rybka3 may play perfect in big part of the openings if it plays against perfect play.

There is a difference.

Then you should say it. It is completely different then. And we wouldn't argue. You are usually very strict for that Uri. That is why I said so.

I did not say that rybka3 plays perfect chess but that it may not be easy to find line that it does not play perfect chess in correspondence games.

Here are again my original words

"If rybka3 plays perfect chess in correspondence chess then the best that they can do is to draw the game and finding a line that rybka does not play perfect is not an easy job"

I did not say impossible and my idea was that they may not have enough time to find it because we talk about correspondence games that the opponents have the same time and not correspondence games when the human has 2 weeks per move against one day of rybka when rybka does not ponder.

Uri

[terminator]

Fritz10 on decent hardware does not want to play those moves after a few minutes more so 5 days allowed, so the 100% engine dump claim is wrong. It would be interesting to hear from those guys with Deep Fritz 10 on an Octa or 12 core system.

[Uri Blass]

Fritz10 on decent hardware does not want to play those moves after a few minutes more so 5 days allowed, so the 100% engine dump claim is wrong. It would be interesting to hear from those guys with Deep Fritz 10 on an Octa or 12 core system.

You talked about g6 and Rxg7 but the claim of Peter Coleman
was that Fritz10 play g6.
The typical correspondence player does not have fast hardware and based on my experience there were also cases when I could win games because my computer simply outsearched the opponent that probably used it for less time so I do not ask to use many days for analysis and few hours on single machine per position or maybe even only few minutes may be enough.

Note that he said Fritz10 and not Deep Fritz10 and I would like to see analysis by Fritz10 here.

Nobody posted analysis by Fritz10 that suggest that Fritz10 does not play
g6 and Rxg7.

Peter did not respond about the Rxg7 part because this move is not suspected to be some move by an human.

[terminator]

With 3 minutes per move on a weak computer Fritz10 comes up with for move 19 :-

Code: Select all

Del Vecchio,A - Schön,W, WC26CT03 ICCF 2006
rr4k1/2q1bppp/p1bpp3/2n3P1/2QBP2P/2N2P2/PPP5/1K1R1B1R w - - 0 1

Analysis by Fritz 10:

1. ²  (0.45): 19.a3 Rb7 20.Ka1 Rab8 21.b4 Nd7 22.f4 Nb6 23.Qb3 e5 24.fxe5 dxe5 25.Be3 a5 26.b5 Nd7 
2. =  (0.10): 19.Qe2 Qb7 20.b3 

(Enrico,  25.07.2008)

With 3 minutes per move on a weak computer Fritz10 comes up with for move 34 :-

Code: Select all

Del Vecchio,A - Schön,W, WC26CT03 ICCF 2006
1q6/5Ppp/4k3/2Q1p3/3pP2P/5P2/rrP5/2KR2R1 w - - 0 1

Analysis by Fritz 10:

1. =  (0.00): 34.f8N+ Qxf8 35.Qd5+ Kf6 36.Qc6+ Ke7 37.Qc5+ Kf7 38.Qd5+ Kf6 
2. =  (0.00): 34.Rxg7 Rxc2+ 35.Qxc2 Rxc2+ 36.Kxc2 Qc7+ 37.Kb1 Qb6+ 38.Ka2 Qa5+ 39.Kb3 Qc3+ 40.Ka4 Qc4+ 41.Ka3 Qc5+ 42.Kb3 

(Enrico,  25.07.2008)

It is likely he did not have a quad or even a dual core but you can see Fritz10 does not agree with those moves at 3 minutes per move. Fritz10 may be "weak" but ask Kramnik what happened. White's opening choices were not the best so in addition to getting decent hardware he may get a good book.