As a customer, I visit the product website once in a while to see if there are any news on the 6 months long unresolved promised bug fixes -- and then I came across the following:
Best Engine for cross-checking Rybka?
Zukertort Date 2010-11-14 19:28 wrote:What is the best engine for crosschecking Rybka? I have two things in mind:
i) Looking for a promising line where Rybka thinks it is 0.00
ii) Catching Rybka's blindspots, especially its endgame issues.
I'm mostly interested in opening analysis, but often the opportunity for simplification means that endgame expertise is important. The endgame seemed to me to be Rybka's weakest area, so an engine known for good endgame feel might be exceptionally worthwhile.
I had been using Zappa, but it mostly became a waste of time because there were almost no instances where Zappa and Rybka disagreed when Zappa was correct.
I'm looking more for "promising lines that end up being good" than for "absolute best moves at every point." Rybka often shows no interest in a line [adjudging it 0.0] when it might have some undiscovered potential that another engine might find. Even if that "other engine" was only right 1 in 5 times, it would be tremendously helpful.
Thanks for any advice.
Felix Kling Date 2010-11-14 20:13 wrote:I wouldn't recommend cross checking - there could be situations where it works, but most of the time it won't work (the second opinion is worthless)
Yes, he actually said that...and I am shocked, I tell ya...
Credit to Nelson H. for at least trying to be objective. But funny his post where he said the engine he would recommend cannot be named has now been removed/edited. Lol.
Credit to Nelson H. for at least trying to be objective. But funny his post where he said the engine he would recommend cannot be named has now been removed/edited. Lol.
Seems Felix is a bit out of line with the main stream computer chess folks. Does anyone really take this guy serious.
You managed to find three positions, from three different opponents (in the same tournament, no less), all where Rybka 4 as White has Rook and bad dark-squared Bishop against Rook and Knight and goes on to lose? Pretty impressive data-mining.
FK: I wouldn't recommend cross checking - there could be situations where it works, but most of the time it won't work (the second opinion is worthless)
In general, I might almost agree on a cost/benefit basis, though the questioner specifically aimed at the cases where cross-checking was likely to be of value.
At one point, LK opined that IvanHoe/Rybka were not sufficiently different to make cross-checking worthwhile --- in the middlegame I would be more prone to agree (usually the question is one of speed), but in the endgame Rybka [and engines in general for that matter] seems a bit more woeful. In my communications with LK, when I noted that bad bishops were something that was notably different in IvanHoe (both in calculation method and in size of adjustments), he was of the opinion that Rybka's method was better, though again maybe that was more for the middlegame than the endgame.
BB+ wrote:You managed to find three positions, from three different opponents (in the same tournament, no less), all where Rybka 4 as White has Rook and bad dark-squared Bishop against Rook and Knight and goes on to lose? Pretty impressive data-mining.
Even more impressive that you observed it. But I assure those were randomly selected games just to prove the point that Rybka is not invisible -- hence the common factor was merely coincidental.
BB+ wrote:
FK: I wouldn't recommend cross checking - there could be situations where it works, but most of the time it won't work (the second opinion is worthless)
In general, I might almost agree on a cost/benefit basis, though the questioner specifically aimed at the cases where cross-checking was likely to be of value.
FK initial statement did not explain his reasoning and my reaction was similar to others who responded.
BB+ wrote:
At one point, LK opined that IvanHoe/Rybka were not sufficiently different to make cross-checking worthwhile --- in the middlegame I would be more prone to agree (usually the question is one of speed), but in the endgame Rybka [and engines in general for that matter] seems a bit more woeful. In my communications with LK, when I noted that bad bishops were something that was notably different in IvanHoe (both in calculation method and in size of adjustments), he was of the opinion that Rybka's method was better, though again maybe that was more for the middlegame than the endgame.
Maybe GM Kaufman meant middlegame because endgame wise, Rybka does not evaluate bishop as good as IvanHoe or Houdini. And As for endgame, the best engine in my view is Houdini.
Does anyone really take this guy serious. 
