Rybka and FIDE

[swami]

I was talking to this young lad the other day, I told him that Rybka on his own computer could defeat Anand, he looked sceptical and thought I was kidding. I showed him some rating list to prove that Rybka is rated at 3200 something. He suddenly became a Rybka fanatic and began to consider using the engine for his game analysis.

So that leads me to the real question, What exactly would be the approximate strength range of Rybka relative to Topalov's Fide Rating of 2813?

Assume the hardware used is Quad core - 4 cores available to many users.

[swami]

Whoops, While editing and reposting, I missed adding "Rybka is about 100-200 elo higher" option. I can't edit anymore now that some people have voted.

Anyway if its about <150, you could choose "less than 100 elo higher", and if it's >150, choose the 2nd one "about 200-300 elo"

[BubbaTough]

Are you assuming that opponents are studying and preparing for Rybka with the same seriousness they do for each other...and that they have for several years? Are we also assuming Rybka has the same level of opening preparation is does now?

Under those conditions I do not expect Rybka to have a huge advantage...opponents can prepare too well against it. On pure chess strength it is much stronger of course, but at the highest level the preparation level should not be underestimated. Of course, if you want to assume Rybka just appears on the scene without opponents doing serious preparation, or that it has its own world class seconds helping with openings or something that is another matter.

-Sam

[swami]

Are you assuming that opponents are studying and preparing for Rybka with the same seriousness they do for each other...and that they have for several years? Are we also assuming Rybka has the same level of opening preparation is does now?

Under those conditions I do not expect Rybka to have a huge advantage...opponents can prepare too well against it. On pure chess strength it is much stronger of course, but at the highest level the preparation level should not be underestimated. Of course, if you want to assume Rybka just appears on the scene without opponents doing serious preparation, or that it has its own world class seconds helping with openings or something that is another matter.

-Sam

Yes, I'm assuming that Topalov, Anand or any other player closer to their rating get exactly an year of preparation against a Rybka 3.0 4 cpu and its own secret opening book. I'm still skeptical about the claim that human GM's with strong preparation could match Rybka on 4 CPU. They may play some elo lower, they could manage a draw, they may find some kind of anti-Rybka stuff. But with the anti-human book, I doubt they would have a chance. There's always a chance they make mistakes from tactical oversight.

Recent GM games gives us an indication they did make a lot of tactical oversights under time pressure.

[M ANSARI]

It is extremely hard to compare top human GM ELO versus a top engine ELO. Humans are probably stronger in static non tactical situations while engines are dramatically more superior in tactical situations. My guess is that Rybka 3's tactical strength far outweighs its static weaknesses versus humans and is overall about 200 to 300 ELO stronger.

[James Constance]

Let's remember that ratings are just statistical tools. If we say that we think Rybka is 200 points higher than X, all we are saying is that we expect X to score 25% against Rybka.

[BubbaTough]

Anand and others at the very top are using Rybka to prepare lines to the point that the game is practically decided. Rybka's ability to vary from their preparation will not be great. Thus, games will likely go quite far into their preparation, and the number of moves they will have to find at the board will be limited compared to against human opponents. I would not be surprised if entire games were regurgitated from memory in this scenario. Thus, in my opinion fatigue would not be as great a factor as it is normally (normally people are not preparing fanatically for computer matches like they do against each other at the top).

-Sam

[swami]

Of course time pressure is a factor for engines as well. Given less time they play worse moves.

Engines with good time management will never run into time problem in the first place.

The factor that cannot be taken into account is fatigue (stamina). This comparison between human and engine would make sense only if the engines could be programmed to lose concentration due to tiredness. How much Elo is fatigue responsible for? I don't think anyone examined this matter in detail.

Interesting idea. Specific rule can be implemented in such a way that engine stops pondering on opponents clock after move number 50, or the engine should think at depth 12 from move number 60 and so on... any sensible limitation/restriction could substitute the tiredness factor.

[Leto]

Have Rybka playing alongside the super Grandmasters and I'd expect her to perform 200-300 elo higher. The chance that they'd allow her to play would be highly unlikely seeing as how they really hated playing against previous engines under similar conditions.