Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / What is the correspondence rating that rybka can get?
1 2 Previous Next  
- - By Uri Blass (*****) [il] Date 2007-06-26 05:03
I believe based on my experience that rybka can get ICCF correspondence rating above 2500(generally it is not only rybka and in the last years simply using the best software on fast hardware could be enough for that purpose).

Thibault de Vassal  claims that he is sure that rybka cannot get rating of 2400 at FICGS even in case of using fast hardware.

I looked at FICGS rating list and I clearly doubt it.
http://www.ficgs.com/rating_list.html

What is your opinion?
Parent - - By Kapaun (****) [de] Date 2007-06-26 05:21
Based on a certain person whom I suspect of being a pure engine player with monster hardware I would guess that 2400 is possible, while 2500 would be really, really hard. As a centaur, however, 2500 should not be too hard.
Parent - - By Uri Blass (*****) [il] Date 2007-06-26 05:35 Edited 2007-06-26 05:37
Thanks for the information.

I want to be sure that you mean not to ICCF rating but to rating of
http://www.ficgs.com/rating_list.html

Note that I only played correspondence chess in ICCF  so I do not know if achieving high rating in ficgs is harder than achieving high iccf rating in ICCF.

considering the fact that the leader (I do not consider the top 2 players that did not play a single game) lost 3 games out of 24 I think that the level is not very high(for comparison the leader of ICCF lost in the last years only one game out of many games that he played .

The leader is   Balabaev Farit with rating of 2580 and I did not count the top 2 players above 2600 who did not play a single game.

Uri
Parent - - By walden (**) [nz] Date 2007-06-26 06:30
Balabaev Farit is rated like 2200 at ICCF and 2580 at FICGS so you work that out!!!
Parent - By Uri Blass (*****) [il] Date 2007-06-26 06:47
Thanks for the information and it seems that it is easy to get high FICGS rating.

It seems that Balabaev Farit is not that weak and maybe he lost many games on time
I looked at his last tournaments and he did not perform bad and he should have ICCF rating of 2400-2500.

1 AUT 10115 GM Teichmeister, Dr. Sven 2590  Austria    loss
2 RUS 141241 SM Turkov, Vladimir Sergeevich 2496  Russia draw
3 EST 930112 IM Siigur, Jüri 2488  Estonia  draw
4 GER 80939 GM Hertel, Peter 2666  Germany  draw
5 IRL 260147 SM Nightingale, Darrell 2493  Ireland  draw 
7 NOR 360341 SM Haugen, Arild 2536  Norway draw
8 NED 370374 GM Maliangkay, Rudolf J. 2578  draw
9 POR 390156 IM Reis, Luís Simões dos 2479  Portugal  draw
10 SLO 480033 SM Pucelj, Joze 2493  Slovenia  draw
11 POL 420112 SM Marcinkiewicz, Jan 2447  Poland draw
12 LTU 920051 SM Novikovas, Vitalijus 2530  Lithuania draw
13 LAT 910062 SM Saksis, Juris 2453  Latvia   draw

Uri
Parent - - By Kapaun (****) [de] Date 2007-06-26 07:06
I didn't talk about FICGS, but not about ICCF, either (I was talking about the German BdF-Server RemoteSchach). Anyway your question, how the ELO here translates to the ELO there, is hard to give an answer.
I tried some comparisons but the result was unclear. While some people have a lower ELO on ICCF (up to 200) some of the really active and known persons with titles (for instance Uwe Mehlhorn or Peter Schuster) seem to rather have higher ratings on ICCF (up to 200 as well).
I guess it's up to you to draw your own conclusions. :)
Parent - - By nuff (**) [us] Date 2007-06-26 09:11
Correspondence chess should be renamed Postman Chess as its an embarrassment to the game. Where is the skill involved in taking moves from your engine and putting them in the ICCF or whatever just like a postman? Show me the raw results of the players are they GMs in OTB chess or only when they are postmen in ICCF?
Parent - - By Zruty (*****) [ru] Date 2007-06-26 09:30
nuff, have you actually met any correspondence chess player?

I have, and he described to me some of his playing strategies, and the course of his analysis in one particular game and I was amazed!

Correspondence chess still is the fencing duel of one mind+engine versus another mind+engine and no doubt. Although I'm afraid we are deviating from the topic.
Parent - By nuff (**) [us] Date 2007-06-26 09:47
Can you show me where to browse a few of these CC games? How many of these CC players have won the freestyle as it takes more than monster hardware to win the freestyle? Some of those freestylers are only 1500 so I doubt their "contribution". I'm lucky I learned chess by myself and can go toe to toe with strong players so I have no need of computer assisted chess. I cannot run the engine tournaments as that in my view is a waste of time I would have spent on ICC/Playchess.
Parent - By Kapaun (****) [de] Date 2007-06-26 10:11
Guys, why do you feed the troll?
Parent - - By Yoav Dothan (**) [il] Date 2007-06-26 09:44
Dear Nuff,

The best answer to your statement would be - for you - to play in a single CC tournament of high level.

Try to put the moves from your engine in the server , and you will be surprised to see that you will be crushed by most of the other players.

Then you can think deeply - why?
Parent - - By Uri Blass (*****) [il] Date 2007-06-26 11:31
I think that it is dependent on the time per move that you use and the definition of high level.

My guess is that using my brain give me only something like 100 elo relative to using an engine with no help for the same time.

My iccf rating is above 2600 so I guess that an engine with no help can get rating of 2500.
My hardware is not the best(I used only single processor machine and only rybka 32 bit) and I can imagine that somebody who is using hardware that is 10 times faster without using his brain may even get rating of 2600.

Uri
Parent - - By FICGS (**) [fr] Date 2007-06-26 12:18
Hello again Uri :) .. Hello to all,

As I said in the same discussion at TCCMB (The Correspondence Chess Message Board) that's not so simple, there are many things to take in consideration in this discussion :

http://ancients.correspondencechess.com/index.php?topic=109.msg826;boardseen#new

First, a few ratings (particularly top ratings) in example don't mean anything, most are peaks so you need to look at the average.. rating changes are quite different at ICCF & FICGS.. For sure top rated players - in any organization - will lose some points as time flies nowadays, because many 2200-2400 are getting stronger (as Rybka's getting stronger)... GM Balabaev had a rating peak because of the FICGS WCH cycle, his future rating will decrease in a few days to 2494 already.

Anyway if you look at en engine running during several days, you understand better what one call horizon... monster hardware x 10 couldn't change things much... and it becomes obvious that's quite impossible for any engine to reach a top correspondence chess rating (at FICGS at least).

It would be interesting to make the test anyway :)

Best, Thibault
Parent - - By Uri Blass (*****) [il] Date 2007-06-26 13:31
My experience is that engines can change their mind after many hours

I agree that the advantage of being 10 times faster is smaller at correspondence time control but I think that it is still significant and my opinion is that when being 10 times faster may give 150-200 elo improvement at tournament time control it still can give 50-100 elo improvement at correspondence time control.

Uri
Parent - - By FICGS (**) [fr] Date 2007-06-26 16:00
It depends on the opponents... In an engine vs. engine match at CC time controls, it would most probably be true... Against centaurs, it has to be proved cause the key positions are almost impossible to be solved by engines...
Parent - By Uri Blass (*****) [il] Date 2007-06-26 16:12
This may be correct when there are forced line but I often get quiet positions when there are no forced lines.

Uri
Parent - - By Kapaun (****) [de] Date 2007-06-26 13:59
It would be interesting to make the test anyway


How is FICG's position to engine use anyway? I couldn't find something about it on the website...
Parent - By Marc Lacrosse (**) [be] Date 2007-06-26 15:56
Completely allowed without restriction of any kind.

... That's precisely why I play there.

Marc

"11. General rules

11. 1. Netiquette

Computer assistance is authorized, as any other kind of help."
Parent - - By Hetman (*****) Date 2007-06-28 16:50
Hi,
lets test Rybka rating. I am playing some games against SIM's. I can delegate one of them to Rybka only.
I do have 32bit version single proccessor. What time per move or analysis depth you suggest ?
Regards
Hetman
Parent - - By Uri Blass (*****) [il] Date 2007-06-28 16:57
I suggest 10 hours per move
Parent - By Hetman (*****) Date 2007-06-28 19:04
Could you give any guarantie ?
Parent - - By MightyMouse (**) Date 2007-06-26 11:52
Based on my long experience and estimations, an ICCF rating of 2400-2450 is probably possible using a computer alone, along with the best chess software, but no human involvement. Raise that to 2450-2500 if using the latest and fastest hardware, updated at least every 2 years. Beyond that, it depends on the chess IQ of the operator who can surely see some occasional improvements on the computer's strategy and move selection. Rule of thumb: add 25-50 Elo points for the average-strength human involvement, 50-100 Elo points if the human is master strength, and 100+ Elo if the human is GM strength.
Parent - - By Uri Blass (*****) [il] Date 2007-06-26 12:04
Based on your theory only players with GM strength could get ICCF rating above 2600

Based on your theory an engine cannot get more than 2500 when even master strength is enough only for 50-100 elo improvement.

It is clearly not correct and there are many players with ICCF rating above 2600 who are not GM's OTB.

Uri
Parent - By FICGS (**) [fr] Date 2007-06-26 12:23
Achieving a top correspondence chess rating is not only a matter of chess skills. It is also a lot of work, an enormous amount of time...
Parent - By Kapaun (****) [de] Date 2007-06-26 13:51
There may be many reasons for that. So there are people who are playing only CC, for instance. Or there maybe even some who are quite weak OTB, but still can add a substantial amount of ELO to their engine, when playing as a centaur. Some may believe that this is not possible, but it certainly is...
Parent - - By FICGS (**) [fr] Date 2007-06-26 12:20
Don't forget an essential point : Any player using Rybka could predict most moves from Rybka !  .. That's an enormous advantage...
Parent - - By Yoav Dothan (**) [il] Date 2007-06-26 13:41
I agree that CC is a lot of  work. The best example could be in the endgame, where weak players relying on their computer make bad mistakes and may lose a drawish ending.  I was given an endgame that the computer analysed as draw , and found very quickly a sure way to win - that the computer did not "see".  The second example is in the openings - if you play against a good program without an internal book - you see at once how badly it plays. I win most games VS. chessmaster when I play against him without the book. 

To sum the CC issue - everyone that thinks that it is an easy task - just mailing the moves - better to try it.

To Uri's question : A high level CC tournament is of category 10 and above. I played a few category 13 tournaments , and it is very - very difficult.
Parent - - By Uri Blass (*****) [il] Date 2007-06-26 16:04 Edited 2007-06-26 16:07
I agree that correspondence game is an hard work but I disagree that computer with no help will lose most games in category 10.
Edit:Of course I assume using opening book that is not too big when you may stop using opening when the position is not very popular
in GM games in the last year.

I play in tournament in category 11
http://www.iccf-webchess.com/EventCrossTable.aspx?id=7670

I expect to win the spanish player to get 8.5/12 but
I think that engine with no help that is used for many hours per move
could get there 6.5/12
(I say engine with no help and not rybka with no help
because at the beginning of the tournament rybka was not available)

It is not something that I can check and it is only a guess.

report of my games including guessing how well computer with no help
could do.

1)Alfonsas Kupsys computer with no help could probably lose
that game.
I drew
2)Dr. Milan Mraz
I won thanks to a clerical error and computer with no help
could probably draw that game(my opponent saw that I did not
choose rybka's move that was a clear draw earlier in the game so
he expected some logical move in order to try to win
and not rybka's drawing move later and it was
one of the reasons for his error in reading my move
when I chose rybka's move inspite of believing it is a drawing move
because I decided that I probably have nothing better
and have no time to do serious analysis).

3)Ing. Marek Kolcák
I lost because of falling into a trap in the opening when I did not
trust theory statistics and same result for computer with no help
when the line is not played often.
4)Siegfried Kluve
draw and I guess same result for computer with no help

5)Zigurds Dauga
draw with black when everything is known theory
and same result for computer with no help.

6)Teijo Oikamo
I won and I guess same result for computer with no help.
Based on my memory
I do not remember playing moves that chess programs did not
suggest after long analysis.

7)Áskell Örn Kárason
draw and I guess the same result for computer with no help

8)Alberto González Freixas
I expect to win it and when Alberto could play better in case of
choosing computer moves the computer does not know how to punish
the not correct moves of Alberto in the endgame so I guess
draw in case of computer with no help.

9)Heinz Polsterer
similiar case to gonzalez except that I have no strong feeling
if computer alone could draw or win.

10)Keith Kitson
similiar case and I won when I am not sure if computer alone
could draw or win.

11)Olivier Bouverot
winning thanks to a clerical error and I assume that in this case
computer could also win it(my opponent lost 7 out of 12 games
by default when against me the loss was not by default but by
resigning after the clerical error)

Having one opponent who lose games on time or quit
is part of correspondence chess so I cannot delete this game.

12)Rune Bergquist
I drew.
I am not sure about the result of using computer with no help
and it may lead to 3 possible results.

I chose a bad opening with white.
It seems that my opponent used Fritz but at the time that
I discovered it I could not find a way to get more than a draw
against Fritz.

Uri
Parent - - By Hetman (*****) Date 2007-06-26 19:45
Hi,
I do not know how many CC games have you played on that level.
If it has been one tournament only it could be some probability  factor involved, the luckiness ;-).
I think that in the early endgame and in the late middlegame phase computers are relatively weaker.
You have to be not bad centaur to get SIM. I am not sure of course, I  palyed cc many years ago, it was other chess.
Regards
Hetman
Parent - - By Uri Blass (*****) [il] Date 2007-06-26 19:59
I played in 3 tournaments
I got in all of them a performance that is above 2600 based on my calculation.

I have the following results in tournament for rating:
1)6.5/8 in the israeli final(at that time I used engines like Deep Fritz6 and chessmaster6000)
and average of the opponents was above 2400.

2)8.5/10 in the OLYMPIAD XV - Preliminaries 
6th board of israel(8/10 was needed for the SM norm)
http://tables.iccf.com/world/olymp/ol-xv-s3B6.html

3)7.5/11(I expect 8.5/12) in the VI EU-MSM, Final.
http://www.iccf-webchess.com/EventCrossTable.aspx?id=7670
Parent - - By Hetman (*****) Date 2007-06-26 20:06
.... The opponents were not using programms ? ;-)
in my shy opinion the game between programms shall finished with the draw.
Rgds
Hetman
Parent - - By Uri Blass (*****) [il] Date 2007-06-26 20:10
most of the opponents were using programs but I played in less tournaments so I could use more computer time.

The game were not games between programs because I also used my brain and I guess the same was correct for most of my opponents.

Uri
Parent - - By Hetman (*****) Date 2007-06-26 20:14
so it is not prove for programm getting SM ;-)
Rgds
Hetman
Parent - - By Uri Blass (*****) [il] Date 2007-06-27 03:06
It is not proved and it is simply my opinion.
The fact that most of the opponents use their brain does not mean that they also used computer analysis for hours and I know that there were clearly cases when they chose inferior moves to the move suggested by the program.

Even if the opponent  is not an engine it does not mean that he never played computer moves and I remember cases when I saw computer moves that were suggested after 20 minutes of search but not after some hours of search.
Parent - By Lee Ma Hong (**) [ph] Date 2007-06-27 03:21
I remember cases when I saw computer moves that were suggested after 20 minutes of search but not after some hours of search.

yes, I've had the same experience since the time when Fritz was the de facto engine choice and dual core processors were still on the manufacturers' drawing boards. it was easy to beat opponents by searching longer and deeper, and combine this analysis with brain power.
Parent - - By Kapaun (****) [de] Date 2007-06-26 13:41 Edited 2007-06-26 13:44
Actually, I'm not so sure about this prediction thingy, because evaluations don't necessarily seem to be the same, especially not if multi processor engines are used (which is almost mandatory nowadays). Furthermore it seems to depend very much on how you use the engine: Do you just let her running with 1 thread? For how long? Or do you use it in analyze mode? If so: How many threads do you chose? Did you clear your hash before? All this can greatly change the output, even if you don't interfere personally doing an interactive analysis. In my opinion it is almost impossible to predict the move even if you know the engine which is used - at least as long as there are several good moves at the top. If it's only one move standing out, then it's pretty clear usually. But then any other engine is likely to show this move, too...
Parent - - By nuff (**) [us] Date 2007-06-26 16:04
I'm still unconvinced about CC skills as evidenced in their non-participating in the Freestyle en-masse. Surely with a cool $16,000 they could do what they do in years in a shorter time? What is the benefit of CC is if you are 1500 to start with and now 1200? Who was say 2000 before CC(the postmen type) and 2500 in OTB chess because of CC?
Parent - - By walden (**) [nz] Date 2007-06-26 20:45
so Uri
you are saying you just use computer to play all your games?
Parent - By Uri Blass (*****) [il] Date 2007-06-27 03:21
No
I did not say it.

Uri
Parent - - By Lee Ma Hong (**) [ph] Date 2007-06-27 00:52
nuff,
there is a very obvious difference between correspondence chess and freestyle events - the rate of play. many (or should I say majority of) CC players go into correspondence chess because they prefer playing at longer time controls. they can study positions for hours on end in a relaxed atmosphere, which is not true in freestyle events nor in OTB play.
Parent - - By Yoav Dothan (**) [il] Date 2007-06-27 03:33
The time for thinking is one factor - the other is the number of games. I play usually 40-50 games at the same time , so I usually study one or two openings deeply. It helps me also for OTB chess (which nowadays I almost do not play).  So - Uri's ideas of using computer for many hours is not practical if you play more than one tournament. In ICCF site it is very easy to see how many games every one is playing, and every half a year there is the new rating published with the number of games finished. If you finish an average of 30-40 games a year it is clear that you could not play an average of 1600 moves and spending many hours on each , unless you have the facilities of a computer labaratory...

So - the whole idea is combining the brain, and the engine , and mainly check your idea with the engine. Many times I thought for a long time only to find an immediate refutation by the engine. So what? I can think again - I have time. Sometimes , of course I make mistakes.

I really wish that everyone that thinks that he will win many games with the help of his engine to try to play in a  tournament.

I played in the European TC VI, Final board 2 a Category 13 GM=7 tournament and made +5 =7. I do not believe that any engine could reach such score.

Even a Category 10 GM=9½ like Altshuler Memorial where I played with +8 =6 that Uri thinks is easier could not be won by an engine with such a result. IF everyone would use Rybka  - most games would end with a draw.

I could give more examples but I think that the idea is clear.
Parent - - By Uri Blass (*****) [il] Date 2007-06-27 07:45
I admit that your results are more impressive than my results considering the number of games that you play.

Opening choice was relatively the weaker point in my games in the last tournament and
I think that I could expect only 50% or maybe slightly more than it if you adjudicate the positions based on the positions when there are no more games in the database.

Maybe choosing 1.d4 instead of 1.e4 in previous tournament was a practical mistake.

The main problem was with white when I failed to get an advantage out of book in many cases when in one case I even got an inferior position.

I expect to get 4.5/6 with white thanks to better playing in the middle game and the endgame but
I feel that I could get at least 1/2 point more with white with better opening choice because I did not win against one of the weakest players with white.

I also believe that I could get additional 1/2 point with black with better opening choice so I guess that with your opening
I could get one additional point.

Uri
Parent - - By walden (**) [nz] Date 2007-06-27 08:31
Interesting I have rating of about 2500 and finish about 60-70 each year but I must say i can spend over 10 hours each and every day on moves as i have lots of free time maybe on weekend i could spend up to 16 hours  each day so not so hard to finish lots of games!!
Parent - By Uri Blass (*****) [il] Date 2007-06-27 11:51 Edited 2007-06-27 11:53
getting more than 2600 is clearly harder than getting 2500
I got more than 2600 and Yoav's performance in many tournaments is more than 2600(his iccf rating is 2589).

I suspect that by doubling the time that you use per game you cannot get more than 50 elo improvement.

Uri
Parent - - By FICGS (**) [fr] Date 2007-06-28 02:24
Freestyle is knowledge & brute force... Correspondence chess is analyze... another game...
Parent - - By Alkelele (***) Date 2007-06-28 03:10
I can assure you that there's a lot of crucial centaur analysis going on in Freestyle games. Actually, the human analysis element is probably more decisive in Freestyle games than in Corr. games, since you have to make a lot of decisions based on incomplete information, that is, you have to use some kind of experience/intuition. Also, the choice about which lines to investigate (first) is more important in Freestyle than in Corr. chess, since the clock is always ticking...

Did you try participating in any of the Freestyle events?

When Freestyle was discussed one of the first times in the FICGS forum, there was a consensus that a pure Rybka would be stronger than a human, because of the short time control. The evidence so far suggests that this is completely wrong -- centaurs have proved to be much stronger than pure engines. I suspect you underestimate the human factor in analysis during Freestyle games. Brute force is simply not enough...

If you like, you can take a look at my last game from the final. I was black vs. a pure engine that I believe had stronger hardware than me. I severely outplayed the pure engine from a slightly inferior position (but didn't achieve enough advantage to win). The critical moves Nd7-f8-g6 and 19...Bf8 would not have been played without some human intervention.

[Event "6th Freestyle Tournament, Final, 9th ro"]
[Site "playchess.com #101760"]
[Date "2007.06.24"]
[Round "1.1"]
[White "PvP, Rybka 2.3.2a mp"]
[Black "Flyingfatman"]
[Result "1/2-1/2"]
[ECO "C67"]
[WhiteElo "2435"]
[BlackElo "2695"]
[PlyCount "104"]
[EventDate "2007.06.24"]
[EventType "match (rapid)"]

1. e4 e5 2. Nf3 Nc6 3. Bb5 Nf6 4. O-O Nxe4 5. d4 Be7 6. d5 Nd6 7. Nc3 Nxb5 8.
Nxb5 a6 9. Nc3 Nb8 10. Nxe5 d6 11. Nd3 O-O 12. Re1 Re8 13. Qf3 Nd7 14. Bd2 Nf8
15. Re3 Ng6 16. Rae1 Bd7 17. Ne2 c5 18. dxc6 bxc6 19. Ng3 Bf8 20. Rxe8 Bxe8 21.
Bb4 a5 22. Bc3 f6 23. Ne4 Ne5 24. Nxe5 fxe5 25. Qg4 Qe7 26. Nd2 Bf7 27. Nc4 a4
28. Rd1 h5 29. Qe2 Qe6 30. b3 axb3 31. cxb3 h4 32. h3 e4 33. Ne3 Qg6 34. Qc2
Be6 35. Kf1 Bc8 36. b4 Ba6+ 37. Kg1 d5 38. Bd4 Bd3 39. Qb3 Qf7 40. a3 Bd6 41.
Qc3 Bb5 42. Ng4 Bc7 43. Rc1 Bf4 44. Ra1 Qc7 45. Be3 Bd6 46. Bc5 Bxc5 47. Qxc5
Qd7 48. Qe3 Qf5 49. Rc1 Qg6 50. Ra1 Qf5 51. Rc1 Qg6 52. Ra1 Qf5 1/2-1/2
Parent - - By Banned for Life (Gold) Date 2007-06-28 05:59
I think your arguments are certainly true for your team (FS), but you may be the exception that proves the rule. Most of the top teams have extreme hardware. I believe there is a reason for this.

Regards,
Alan
Parent - - By Kapaun (****) [de] Date 2007-06-28 10:28
Yes, there certainly is. If everybody uses the same software, and if everybody (at least in the finals) is a top chess player, then hardware is the way to make a difference...
Parent - By Hetman (*****) Date 2007-06-28 16:53
the same in formula one, the car and the driver :-).
Good driver in the good car is giving chances for win.Bad car and good driver or bad driver and good car no chance to win
Rgds
Hetman
Parent - - By FICGS (**) [fr] Date 2007-07-14 00:08
Analysis, I mean hours of analysis..... A few variants during a few minutes, even with Rybka is very different of a correspondence chess analysis.

I did not participate in Playchess freestyle events, but I played some centaur games, including freestyle "lightning" matches at FICGS... Ok, let's say "small analysis" :)

About the discussion in FICGS forum, we probably thought about a very fast time control... (actually I'm quite surprised, as I agree that a centaur is obviously stronger than a pure engine at any time control)
Up Topic Rybka Support & Discussion / Rybka Discussion / What is the correspondence rating that rybka can get?
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill