Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / If it's overrated, then what's the real FIDE ELO of Rybka?
- - By SillyFunction (**) Date 2007-08-01 06:13
Some top GMs (like Magnus Carlsen as I far as I know) said that they used Rybka as an analysis tool. But they also said that Rybka is overrated. Then what do you think...

a) Given the number one human player on the planet and plus 300 to his ELO point for Rybka since Rybka can surely win and never lose a single game to him.

b) It's impossible to really know since there won't be any serious human tournaments for Rybka. From now on human ELO should not be compared to computer ELO any more.
Parent - - By Quapsel (****) Date 2007-08-01 07:12

> But they also said that Rybka is overrated.


What do they mean with that statement?

1.)
All Engine in the typical Engine-ELO-Lists are overrated!
This could be. I think, the relation between human-ELO and Engine-ELO is weak, ist known only very aproximately.
There are only too view serious Games between Top-Engines an Humans.
Every computerches-interested man would be happy to see more Games to calibrate the Engine-Elo-List with the human list.

2)
Specially Rybka is overrated against humans
This would mean one of the following points is correct:
- if other strong Engines play with a some Success against Humans, Rybka would not play
  with so much more Success, as it could be expected because of the Engine-ELO-Lists
- Rybka does not reach more Success than other Programs
- Rybka will have less Success than other Programs (annotated here only because of totality)


Here is a general open Question for me:
--------------------------------------------------------------
If Engine-A has for instance 70 ELO Points more than Engine-B, declared in good Engine-Elo-Lists,
will Engine-A then get on avarage that much more success than Enine-B against Humans, as it should be expected because those 70 Points ELO-Difference?

I could imagine, that generally the success against humans shows less differences than against Engines!

Quap
Parent - - By Vasik Rajlich (Silver) Date 2007-08-01 07:50
FWIW - all the evidence we have so far points to three things:

1) Elo is just Elo. There is no evidence to support the hypothesis for example that engine Elos "condense" when playing against humans.
2) CEGT ratings are quite well-synchronized with FIDE ratings.
3) Strong human players underestimate the strength of engines.

Vas
Parent - - By M ANSARI (*****) Date 2007-08-01 08:23
It is likely that the lower rated engines are actually much stronger than their ELO shows if they were to play humans, while the top rated engines might be a little inflated due to possible draws against much weaker humans.
Parent - By SillyFunction (**) Date 2007-08-02 00:30

> It is likely that the lower rated engines are actually much stronger than their ELO shows if they were to play humans


Agree absolutely.

> while the top rated engines might be a little inflated due to possible draws against much weaker humans.


What a pity, we have no way to prove. More and more, I feel that GMs have seen that it's pointless to prove how weak they are comparing to engines.
Parent - - By George Tsavdaris (****) Date 2007-08-01 10:32 Edited 2007-08-01 10:37

>1) Elo is just Elo. There is no evidence to support the hypothesis for example that engine Elos "condense" when playing against humans.


There is no evidence to support the opposite too. :)

I believe that some engine that is weaker in engine-engine games, can be stronger in engine-human games because of style.
Junior 10 1 CPU for example seems weaker than Fruit 2.2.1 in engine-engine games, but because of the more aggressive style of Junior, that should bring more headaches for humans than the more quiet Fruit, i expect Junior to perform better against humans than Fruit.

Also because ELO is ELO, that means a measurement of performance in a sample of players that have played between them, we can only make conclusions and predictions from that ELO, only about the performance between each player in THAT sample and NOT about the performance they would have against players outside from that sample of players.....

>2) CEGT ratings are quite well-synchronized with FIDE ratings.


How is that? How do you know that and how they've done it.....?

>3) Strong human players underestimate the strength of engines.


I believe that too.
Parent - - By SillyFunction (**) Date 2007-08-02 00:32 Edited 2007-08-02 00:44

>> 2) CEGT ratings are quite well-synchronized with FIDE ratings.
> How is that? How do you know that and how they've done it...?


It happens to be my question, too :-)
Parent - By Vasik Rajlich (Silver) Date 2007-08-03 11:31
If you look at the various man vs machine matches and engines participating in tournaments (this amounds to hundreds of games), you'll see that the engines invariably performed up to their CEGT ratings.

Vas
Parent - - By SillyFunction (**) Date 2007-08-02 00:25

> 3) Strong human players underestimate the strength of engines.


It's like, in the past, male players would always say that they were sick when they lose to female ones.
Parent - By bluemax (**) Date 2007-08-02 10:44

> It's like, in the past, male players would always say that they were sick when they lose to female ones.


Or that they were not watching the board!!!
Parent - - By lkaufman (*****) Date 2007-08-02 04:24
     I have to disagree with statement number 1. There is indeed evidence that engine Elos condense (by 20-25%) when playing against humans. The proof is the Swedish rating list, which has been around for two decades. Over the years, the list has been repeatedly lowered because the top rated programs never lived up to their ratings on that list. Now the top few programs (after Rybka) are probably rated approximately right vs. humans (at least based on slightly outdated versions), but the twenty year old machines at the bottom are way underrated, as compared to official ratings they got at the time. Of course rating levels may be a bit different now, but I'm talking about several hundred Elo here. The spread of the Swedish ratings needs to be contracted by 20-25% for the ratings at both extremes to be plausible.
     So in my opinion Rybka's rating on CCRL or CEGT is higher than it would be against humans, not because of anything specific to Rybka, but just due to the above. If we assume that the 2700 ratings are the ones that are likely to be about right (as there were several matches against 2700 level players that were closely contested a few years ago), then a 3100 rating on CCRL or CEGT probably means about 3000 or a bit more against humans. Based on the Ehlvest matches with reasonable estimates for the handicaps, this value of 3000 for Rybka on a quad against humans seems about right to me.
     So if Carlsen was talking about CCRL/CEGT ratings for Rybka, I would agree with his statement, but only for statistical reasons.
Parent - - By Vasik Rajlich (Silver) Date 2007-08-03 11:36
Ok, that's interesting. I'm not at all familiar with these older programs.

Are you saying that for example 'Chess Genius 1.4 SX1 OMAP 310 120 MHz' is stronger than 2150 FIDE?

http://ssdf.bosjo.net/nr048.htm

Vas
Parent - - By er (**) Date 2007-08-03 20:36
Some computer chess history:

In 1994, 13 year ago,
Gary Kasparov lost this rapid game (25 min /game, I think)
against Chess Genius:

[Event "London PCA"]
[Site "?"]
[Date "1994.??.??"]
[Round "1"]
[White "Kasparov, Ga"]
[Black "Comp Genius"]
[Result "0-1"]
[ECO "D23"]
[PlyCount "120"]
[EventDate "1994.??.??"]

1. c4 c6 2. d4 d5 3. Nf3 Nf6 4. Qc2 dxc4 5. Qxc4 Bf5 6. Nc3 Nbd7 7. g3 e6 8.
Bg2 Be7 9. O-O O-O 10. e3 Ne4 11. Qe2 Qb6 12. Rd1 Rad8 13. Ne1 Ndf6 14. Nxe4
Nxe4 15. f3 Nd6 16. a4 Qb3 17. e4 Bg6 18. Rd3 Qb4 19. b3 Nc8 20. Nc2 Qb6 21.
Bf4 c5 22. Be3 cxd4 23. Nxd4 Bc5 24. Rad1 e5 25. Nc2 Rxd3 26. Qxd3 Ne7 27. b4
Bxe3+ 28. Qxe3 Rd8 29. Rxd8+ Qxd8 30. Bf1 b6 31. Qc3 f6 32. Bc4+ Bf7 33. Ne3
Qd4 34. Bxf7+ Kxf7 35. Qb3+ Kf8 36. Kg2 Qd2+ 37. Kh3 Qe2 38. Ng2 h5 39. Qe3 Qc4
40. Qd2 Qe6+ 41. g4 hxg4+ 42. fxg4 Qc4 43. Qe1 Qb3+ 44. Ne3 Qd3 45. Kg3 Qxe4
46. Qd2 Qf4+ 47. Kg2 Qd4 48. Qxd4 exd4 49. Nc4 Nc6 50. b5 Ne5 51. Nd6 d3 52.
Kf2 Nxg4+ 53. Ke1 Nxh2 54. Kd2 Nf3+ 55. Kxd3 Ke7 56. Nf5+ Kf7 57. Ke4 Nd2+ 58.
Kd5 g5 59. Nd6+ Kg6 60. Kd4 Nb3+ 0-1

How many Elo for 'Chess Genius'?

egr
Parent - - By turbojuice1122 (Gold) Date 2007-08-04 01:54
Keep in mind that 10-15 years ago, people didn't know how to play against chess programs--they didn't know their strengths very well and they knew even less about their weaknesses and how to exploit them.  For example, I'd bet that any grandmaster rated above about 2700 could now beat the second Deep Blue (1997 version) in a six-game match without much trouble.  However, the correct methods of playing such programs weren't known then, and so their estimated elo in the 1990's was always far higher than what we'd estimate nowadays.  Kramnik, of all people, greatly illuminated the path of how to defeat the silicon monsters, beginning with, I think, the Frankfurt tournament in either 1999 or 2000.  Anyway, at the time, one would probably have estimated the rating of Genius on a 200 MHz Pentium available at the time to be around 2550-2600 or so (even though the SSDF list certainly gave a lower rating) based on its results against the grandmasters of the day (the Kasparov game that you give here was a rapid game, where computers have a bigger advantage, and there will also occur situations where Kasparov will simply have a bad day), whereas today, it would be very, very lucky to break 2300.
Parent - - By George Tsavdaris (****) Date 2007-08-04 11:06

>Keep in mind that 10-15 years ago, people didn't know how to play against chess programs--they didn't know their strengths very well and they >knew even less about their weaknesses and how to exploit them.


Before 10-15 years, people were not morons and they perfectly know Chess program's strengths and weaknesses as also strong points.
It is exactly the same with today!

>For example, I'd bet that any grandmaster rated above about 2700 could now beat the second Deep Blue (1997 version) in a six-game match >without much trouble.


I disagree about beating it without much trouble. I disagree about the reason too. Perhaps it would perform better against Deep Blue of 1997, than a 2700+ grandmaster of 1997, but NOT because they know better about computers now, but only because i guess 2700+ grandmasters of today, are stronger in Chess knowledge and strength than these on 1997.

>However, the correct methods of playing such programs weren't known then, and so their estimated elo in the 1990's was always far higher >than what we'd estimate nowadays.


Why do you believe that people in 90's were idiots with no knowledge about some things?
People at 90's KNEW Chess engine's strengths and weaknesses very well, just like today.
Back then the endgame was a clearly weak point of computers as they played the endgame very very badly many times. People knew that!
Play on closed positions, was also a very very weak point of computers back then. People knew that!
Tactics was the strong point of computers even back then in 90's. People knew that again.

It is just that todays computers are so much stronger at tactics today and even a small error leads you to disaster, so that GM's HAD to switch to more quiet play. But this is because they saw computers are far superior at tactics than they were at 90's and NOT because at 90's they didn't know that tactics is the strong point of computers, while now they know it and avoid it.

You should not see those people of 90's as ignorant and without much knowledge just because it was 15 years ago.
They knew perfectly well.....
Parent - By turbojuice1122 (Gold) Date 2007-08-04 14:44
I am saying this not based on seeing things today, but based on seeing things in the 1990's, and what you're saying does, in a way, make me even more convinced of my argument by helping to bring back the memories.  Near the end of the 1990's and early 2000's, there were several tournaments in which strong grandmasters played along with a computer running Fritz, Junior, or Shredder.  They were getting completely demolished by these programs running on hardware of Pentium III 600 MHz or less.  This definitely wouldn't happen today against Fritz 6, Junior 6, or Shredder 6 on such hardware (only Shredder 6 would be rated over 2500 on that hardware nowadays).  Then Kramnik, among others, showed that it wasn't enough to blockade the position, trade down, and then try to beat the computer in the endgame--this hadn't been working very well--instead, you had to semi-blockade the position and then launch a deadly attack that would turn into a winning attack just beyond the programs' horizons.  This became the new way to go in beating the machines of the time; this method was very definitely not employed in the Deep Blue games, but it probably would have worked just as well, if not better, if it had been due to Deep Blue having had a relatively close horizon--it's just that if you go wrong against Deep Blue, you will be annihilated very quickly (e.g. 1997 game 6, though Kasparov certainly resigned too early there).
Parent - By lkaufman (*****) Date 2007-08-04 13:07
Yes. Although I'm not familiar with that exact version, all the old programs would perform much better than their low ratings on the SSDF today, especially those with ratings starting with 1. It's true that people are more familiar with programs now than then, but in terms of ratings they would get playing against an unfamiliar engine this effect is fairly small, nothing like the several hundred elo I am talking about.
Parent - - By SillyFunction (**) Date 2007-08-02 00:22 Edited 2007-08-02 00:46

> What do they mean with that statement?


I think that Magmanus means specifically Rybka.
Personally, I think that Rybka is really above the other engines
But when super GM says that it's overrated, they use their gut feeling and there should be something right in it.

Actually, if my (a) is correct, we should say that Fritz is also 300 points above the number one player, too (due to Bonn match it won without loss.) It leads to the idea that from now on we will only know that "engines are better at least 300 points" but we'll never really know exactly how many points.

> If Engine-A has for instance 70 ELO Points more than Engine-B, declared in good Engine-Elo-Lists,will Engine-A then get on avarage that much more success than Enine-B against Humans


My 2 cents, I think that the engine Elo lists tell us some correct information. Rybka should win more than the other engines. 70 points is meaningless but 100-200 points is surely meaningful.
Parent - By Vasik Rajlich (Silver) Date 2007-08-03 11:37
Fritz's performance rating vs Kramnik was 2875 or so.

Of course, 6-game performance ratings are really hard to talk seriously about.

Vas
Parent - - By Uri Blass (*****) Date 2007-08-01 11:41
a is not correct.
Even when the difference in elo is more than 300 elo I expect the weaker player to win one game if he plays enough games but it gives an idea for future handicap match.

The human play normal chess against rybka when his target is simply to get at least one win
when he has limited time of one week.

The human can choose the time control of his games.

Uri
Parent - - By SillyFunction (**) Date 2007-08-02 00:42

> Even when the difference in elo is more than 300 elo I expect the weaker player to win one game if he plays enough games but it gives an idea for future handicap match.


I realize that my (a) has many flaws but I can't give 1000 to it :-) I also realize that Rybka is not invincible (hopefully this quotation will be wrong one day soon!) We've seen that it can be beaten by its rivals from time to time. And that means some humans still have chance to win it, too. My point is, we should have a number for the present idea HOW MUCH DIFFERENCE CAN IT BE.

My 300 means what it shows in the present lists. Rybka is now about 3100, right?
Parent - - By jamerolle (**) Date 2007-08-02 01:22
" Rybka is overrated "

People that make staments such as this are not juging Rybka Holisticallly , there making jugments based on  1 % of its play which can be " less then great" of which they will never earn in a real game anyway.  Magus, lets look at it this way Fritz beat the hell out of Kramnik and Kramnik uses you as the door mat in tounaments while Rybka beats Fritz over 70% of the time. 
Parent - - By Uri Blass (*****) Date 2007-08-02 05:28
It is easy to beat a player who cannot see mate in 1 so I do not care about the match of Fritz against kramnik.
The only way to find rybka's rating against humans is to give her to play against humans.

Top humans do not need to beat rybka in a match in order to show that she is overrated and scoring 25% against her is clearly enough.

Uri
Parent - - By Quapsel (****) Date 2007-08-02 05:57

> Top humans do not need to beat rybka in a match in order to show that she is overrated and scoring 25% against her is clearly enough.


If such a top human wants to create consequently drawn matches, will he reach this aim against Rybka?
At least in half of all games?
I absolutely assume so.
What could really be shown by this?

Quap
Parent - - By M ANSARI (*****) Date 2007-08-02 07:18
I think LK estimate of a 3000 ELO rating for Rybka is accurate.  The very top GM's with ratings of 2700 ELO and above should manage to steer quite a few games into draws and maybe ... a big maybe ... win a game occasionally.  Still ... they would have much more success in scoring a draw or a win against Kasparov or Kramnik.  It would be really interesting if a tournament field of say the top 10 GM's on earth would play a tournament against an identical field of Rybka's ... with the prize going for the highest scorer against Rybka.  This might be the only way to really motivate top GM's to play Rybka without risking humiliation.  The thing is that Rybka is not static and is constantly improving.  Even if the software would remain static (which it is not) hardware is progressing at a tremendous pace.  A lot of weaknesses of Rybka will be "fixed" automatically as hardware progresses ... it will get more and more difficult to find ways of making Rybka stronger as the tiny defects are fixed.  I would guess there might still be a 100 ELO improvement there somewhere ... then the climb will be much more difficult.
Parent - By SillyFunction (**) Date 2007-08-02 08:09
interesting idea indeed.
Parent - By Quapsel (****) Date 2007-08-02 11:22

> with the prize going for the highest scorer against Rybka


I think, this is problematical. I'll explain:

A long time ago (2 or 3 centuries, it seems) there was an Mephisto-Event:
Mephistos had do play against Amateur players. In thoses times these
Humans had very good chances to win against the programs.

But Mephisto splashed out a very nice price for that Human, who would
mate Mephisto in the shortest Game.
An so many tricks were tested and many soldiers of fortune played unsuitable aggressive.
:-o

In the end the mephistos had won most Games, against all expactations,
and Mephisto had a very good headline for sales promotion.
:-D

Quap
Parent - - By SR (****) Date 2007-08-02 09:48
You write "some humans still have a chance top win it too".  Of course Rybka is not invincible. It is even possible for a human to outplay Rybka and in fact I have outplayed my Rybka (on a pentium 4 CPU 3200 GHz) so on a few occations (time control 10min+5s). Here is one clear cut win where Rybka first (I think) underestimate the black attack on the black squares, and then get caught in an very nasty hold that it never recover from....

[Event "Blitz:10'+5"]
[Site "?"]
[Date "2007.07.25"]
[Round "?"]
[White "Rybka"]
[Black "Riis"]
[Result "0-1"]
[ECO "E97"]
[WhiteElo "3318"]
[BlackElo "3007"]
[Annotator "Riis,Soren"]
[PlyCount "96"]
[EventDate "2007.??.??"]
[TimeControl "10'+5"]

{159MB, RybkaII.ctg, D1CH4T1J} 1. d4 Nf6 2. c4 d6 3. Nc3 g6 4. Nf3 Bg7 5. e4
O-O 6. Be2 e5 7. O-O Nc6 8. d5 Ne7 9. b4 Ne8 10. Nd2 f5 11. c5 Nf6 12. f3 f4
13. Nc4 Rf7 14. Qb3 Bf8 15. Bd2 g5 16. cxd6 cxd6 17. Be1 Ng6 18. Bf2 g4 19. Kh1
g3 20. hxg3 fxg3 21. Bxg3 Nh5 22. Bh2 Qh4 23. Rfe1 Ng3+ 24. Kg1 Nf4 25. Bf1 Rf6
26. Bxg3 Qxg3 27. Ne2 Nh3+ 28. Kh1 Qh4 29. g3 Qh5 30. Kg2 Rh6 31. Rec1 Be7 32.
Rc3 Bd7 33. Qa3 Bb5 34. Rb1 Rf8 35. Rbb3 a6 36. Rc2 Rf7 37. Rcc3 Rg7 38. Rc1
Bh4 39. Nxd6 Rxd6 40. g4 Bxe2 41. Rc8+  Kf7 42.Bxe2 Be1 43. Rc7+ Kf6 44. Qc1 Qh4 45. Qc5 Nf4+
46. Kf1 Qh1+ 47. Qg1 Qxg1+ 48. Kxg1 Rxc7 {and black won} 0-1
Parent - - By lkaufman (*****) Date 2007-08-02 15:21
     This is the worst possible opening for Rybka against a human. Closed position with clear attacking plan for Black. It would probably be much harder for you to beat Rybka with no book at all. For human events, the book needs are quite different than for computer events.

    
Parent - - By SR (****) Date 2007-08-02 15:58 Edited 2007-08-02 16:03
Absolutely! Of course the point is to choose openings that produce middlegames that are easy to play for the human. The "handicap" I get is that Rybka only learn relatively slowly from the opening (I am using the defalut values).  I am not sure my small "tricks" would work agaist Rybka on a 4 processor machine.
With more time Rybka would for example "punish" 13. -,Rf7 that I think is actually incorrect (I think 13.-,g5 is better).

I have learned a lot of chess from Rybka. Before I would have answered 9.b4 with 9.-a5 almost automatically, however Rybka taught me that 9.-a5 is in fact a serious positional mistake, that  easen whites play of the Queen Side considerably. 9.-Ne8 is a much better move, though Black still have a difficult position.
13.-Rf7 and 14.-Bf8 where a bit cheeky move from my part leaving Rybkas opening books, but exploiting that Rybka has to little time (on my machine) to exploit the move order.

After a loss Rybka played a different opening move (e.g. 1.c4), but with the particular default setting of the opening book it is relatively easy to trick Rybka into a similar type of closed position.   It is a huge advantage for the human always to be familiar with the early middlegame positions that arrise from the opening. Its unfair to Rybka, however the conditions teach me a lot of chess.
Parent - - By Hetman (*****) Date 2007-08-03 17:09
Yes, programm is very usefull training tool. The problem is to find the proper one.

Your achievements is big, nevertheless. The book of Rybka shall not be deciding factor. I think You showed the human power - long term planning i.e. play on the weak fields area. It is classic of the chess .
Regards
Hetman
Parent - - By killkra (*) Date 2007-08-05 01:20
"Fools rush in where angels fear to tread. " Magnus Carlsen is one of the fools.

Since Fritz can kill Kram easily, and Junior can kill Fritz, Rybka can kill Junior, So Rybka can kill Carlsen just like to kill an ant.

Parent - - By lkaufman (*****) Date 2007-08-05 05:06
Saying that Rybka is overrated does not imply that Carlsen thinks he is anywhere near Rybka's level. I happen to agree with him -- I think that Rybka would only earn a rating of slightly over 3000 if she played in the top human tournaments regularly, as opposed to her 3100+ CCRL rating. Other top programs would be in the 2900 ballpark.
Parent - - By SillyFunction (**) Date 2007-08-05 06:52 Edited 2007-08-05 06:55
When I think of Rybka, the face of Vas usually appear in my head. It may mean Rybka is Vas and Vas is Rybka for me. But, well, Vas is 2300-2400 player (I don't remember exactly) and you say that Rybka should be around 3000, then I have to change my mind. Rybka is Rybka and Vas is the creator of it :-)

If I'm not wrong, Vas used to say that he'd never played any serious games with his creation. Just test positions or something. You know, when I was a kid I once thought that computer chess is the only friend of mine. I would be certainly glad if I could create a great friend by myself and I would be proud to be crushed by it everyday :-)

One day Rybka should be able to play like God. No mistakes. No losses. And even more exciting, no draws! Someone here throw in the number 3700 for that and I will wait and see if there is any significant differences. But now I believe that 200-300 points above the top rated human player in the history is what Rybka deserves :-)
Parent - - By oudheusa (*****) Date 2007-08-05 08:20
No draws?

I happen to be of the opinion that chess is always drawn with best play, as was just proven for checkers.

So a 3700 Rybka - Rybka match will result in wildly interesting chess, but end in a draw.
Parent - By SillyFunction (**) Date 2007-08-05 15:31

> I happen to be of the opinion that chess is always drawn with best play


Yes, I think so. Don't take me wrong. I mean that it will be veryyyy exciting to see if any program (whether Rybka or not)can beat everyone and every engine in this world. Sure if it's PERFECT and play against itself every game should end in a draw. Guess at that time I and you won't watch chess any more. We're watching chess cuz there are still victories to expect, not to forget :-)
Parent - - By George Tsavdaris (****) Date 2007-08-05 10:42

>I think that Rybka would only earn a rating of slightly over 3000 if she played in the top human tournaments regularly.


That means you believe in a 12-game match, that Rybka would win against:

-Kramnik:      +7 -0 =5  or  +8 -1 =3
-Morozevich: +8 -0 =4  or  +9 -1 =2
-Shirov:         +8 -0 =4
-Carlsen:       +9 -0 =3 or +10 -1 =1

These scores seems a bit high to me and i would expect something like:
-Kramnik:      +4 -0 =8  2890 ELO performance
-Morozevich: +6 -0 =6  2950 ELO performance  
-Shirov:         +6 -0 =6 or +6 -1 =5  2926 ELO or 2890 ELO performance
-Carlsen:       +8 -0 =4  2984 ELO performance.

If the GM's are prepared normally, be in good shape and not like Adams who underestimated Hydra and he was on bad form anyway.

As for the number 3700 ELO people discuss here, against humans is not probable to happen right now, since even if Rybka wins +11 -0 =1 against the highest rated human right now Anand, she would only gain an ELO of 3282.
To have an elo of 3700 she should beat Anand with a score of 99.94% or more that means they would have to play 5000 games and Rybka to lose only 3 of them. But a human will not accept to play more than 20-30 games so such rating will never be observed.

>SillyFunction:
>But now I believe that 200-300 points above the top rated human player in the history is what Rybka deserves


No one deserves nothing, if he does not gains it. :)
300 ELO points above Kasparov's 2851 means current Rybka should be able to win that Kasparov: +9 -0 =3 or +10 -1 =1
I don't think could happen. This is way an overestimation.
Rybka is not 300 ELO points stronger than the highest rated human in history in my opinion....
Perhaps 100, perhaps 150 but not 300.
Parent - By SillyFunction (**) Date 2007-08-05 15:32 Edited 2007-08-05 15:40
What I can say is your opinion is more scientific than mine. Mine is just a dummy's gut feeling :-)
Up Topic Rybka Support & Discussion / Rybka Discussion / If it's overrated, then what's the real FIDE ELO of Rybka?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill