An Unexpected Rybka vs. Robbo result

[lkaufman]

I got different code from another source so probably it depends on which version we are talking about. Anyway in view of both your comments and this other code, I'm convinced now that the results are probably due to Robbo handling very short time limits differently than normal time limits, for whatever reason, so I would have to conclude that this particular finding of Rybka's superiority over Robbo at such hyperspeed chess has no larger meaning.

[Albert Silver]

I got different code from another source so probably it depends on which version we are talking about. Anyway in view of both your comments and this other code, I'm convinced now that the results are probably due to Robbo handling very short time limits differently than normal time limits, for whatever reason, so I would have to conclude that this particular finding of Rybka's superiority over Robbo at such hyperspeed chess has no larger meaning.

Have you checked the type of TC? Meaning game in X, or X moves in Y, and finally X moves in Y with Z increment. Some engines have been known to suffer specifically in certain TC formats but less than others even if the overall time was similar.

[lkaufman]

No, I suppose it's possible that at 40/x instead of increment Robbo would not misbehave at super-short controls, but I'm convinced that such super-short levels are only valid if you know that the time management code of both engines being compared is the same at all levels. So I think we'll continue to test Komodo with other strong engines this way, but just leave out Robbo.

[Mincho Georgiev]

5. Other ideas??? .

Here is one idea. Every posted results before was too preliminary and based on not-enough games to confirm with confidence that robo is really stronger than rybka. How about that ?

[Albert Silver]

5. Other ideas??? .

Here is one idea. Every posted results before was too preliminary and based on not-enough games to confirm with confidence that robo is really stronger than rybka. How about that ?

I don't buy it. I tested them against each other with decent hardware, ponder on and off, at fast and slow TCs, and there is compelling evidence that Robbo is stronger at faster TCs than Rybka, but not at slower ones (considerably slower to be clear). I am not basing this on pure numbers, but watching many of the games. In fact, Robbo scored so well at fast TCs in spite of being swindled quite a number of times in balanced endgames. I suspect that there are diminishing returns as the time control increases, and since they are both 1st cousins to say the least, this is exacerbated when playing each other.