How good is the program vs the man

  

Back to forum


Benjamin Block    (2008-08-30)
How good is the program vs the man?

I want more test game in corr man vs computer. To know how high elo the computer have? It was a long time ago the Hydra played vs Nickel i want to see Rybka vs GM Nickel or some one other.


Thibault de Vassal    (2008-09-02 15:14:21)
How good is the program vs the man?

True, I don't know about any recent experience like Hydra-Nickel... We may organize something like that with Rybka 3.0 and why not a strong Go engine too, any idea (rules, time control and so on) ?


Benjamin Block    (2008-09-02 18:58:21)
organize?

On this site? Yes it would be funny. I think it will be hard to ask the rybka team?? Have they time? But if some one that we trust can use the computer 30+1 to don´t get to much electricity. How are inveted to play? The openingsbook Random?


Thibault de Vassal    (2008-09-05 19:17:59)
Rybka playing correspondence chess

We may try anyway... I'll post something in the Rybka forum soon.


Thibault de Vassal    (2008-09-09 23:39:07)
Rybka forum

You may support or discuss the idea of a correspondence chess match between Rybka and a centaur (human + engines) here :

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=7307


Thibault de Vassal    (2008-09-12 14:39:49)
Rybka 3.0 about 2300 at FICGS ?

According to Larry Kaufman from Rybka team in the discussion linked above :

"If we assume that both sides have the same opening book, then I think two things are fairly safe to say: 1. A good human chessplayer (or even a bad one with good centaur skills) + same Rybka will win a long match from unassisted Rybka. 2. In any individual game, the chance of a draw is fairly high. (...) I mean more than half the games, but not way more. The actual draw percentage depends very heavily on the opening book used."

I agree with that, so I assume that Rybka 3.0 thinking at least 24 hours per move would have a correspondence chess rating of about 2300 at FICGS.

Any opinion ?


Benjamin Block    (2008-09-12 15:55:58)
Godd to know!

Thanks! Ficgs rating is around the same as iccf?


Thibault de Vassal    (2008-09-12 16:12:04)
ICCF ratings

Quite hard to say, the rating calculation is not the same, but many players here who started with a strong established rating came from ICCF...


Wayne Lowrance    (2008-09-13 00:47:03)
Rybka 2300 @ FICGS

That is interesting Tribault. Do you mean the program running unassisted, no player help, Rybka choose own book moves ? If that is your basis then I say no Way Rybka on FICGS get this rating. There are very many sharp Centaurs playing here. With excellent tuned books. That is the main thing. CC games are won/lost on opening book. I am of the opinion that centaur + program is too strong for Program itself. Results on ICC have demonstrated that. Wayne


Thibault de Vassal    (2008-09-13 13:45:25)
Correction

Hi Wayne, yes I mean Rybka unassisted and playing with the best book possible. But this assumption of rating was based on games played with a ~2450-2500 centaur... But in my opinion, Rybka would not outplay a centaur rated 2200 as well, so her rating may be closer to 2200-2250.

Finally, maybe the most interesting part would be a match between Rybka 3.0 and a correspondence chess player rated 1900 :)


Benjamin Block    (2008-09-13 14:01:21)
Rybka vs 1900 man.

Rybka will win that game if the player not are under rankad. But what about 2100 vs computer?


Benjamin Block    (2008-09-15 16:32:01)
Asked Rybka!

Why they not wanted in corr. The answer was. "Becase the man will also use rybka".