I've been running several 15 10 games between Rybka 1.2f and Rybka 2.1c and the results are surprising. Rybka 1.2f is actually leading the way. Of course, so far, only 7 games have been played and it could just be bad luck for Rybka 2.1. ;p However, could it be possible that Rybka has grown weaker with the new version?!
Originally posted by exigentskyIt depends....
I've been running several 15 10 games between Rybka 1.2f and Rybka 2.1c and the results are surprising. Rybka 1.2f is actually leading the way. Of course, so far, only 7 games have been played and it could just be bad luck for Rybka 2.1. ;p However, could it be possible that Rybka has grown weaker with the new version?!
as black? as white?
and also....
Read the version history, see what they changed...
if for example the new version uses rather agressive pruning (and the older one doesn't) then that may help explain it.
....(if that is the case) then that means the new version can search deeper, BUT, there is bigger chance it may miss an exceptionally clever move...the older version being less "selective" ((and because it has the time)) may be able to spot some clever, rather obscure move...which its "successor" didn't
and another plausible possibilty...
check the engine settings...both EXACTLY the same??, same books, same tablebases, same hash sizes, etc, etc
Everything is the same except the engine.
- No books
- Same tablebases
- 200 MB each
- 15 + 10 sec. increment
- Alternate colors
- Parameters are identical/default and they are turned off anyway.
From what you say, it sounds like Rybka 2.1 may be better in blitz. Very possible, but when i ran 50 games at 5 5 settings, Rybka 1.2f was leading by about 5%. This leads me to think that the blitz settings must be very low, perhpas 3 3 or even faster.
All I can say so far is, I'm really dissapointed.
Originally posted by exigentskyTry changing the actual engine settings....
Everything is the same except the engine.
- No books
- Same tablebases
- 200 MB each
- 15 + 10 sec. increment
- Alternate colors
- Parameters are identical/default and they are turned off anyway.
From what you say, it sounds like Rybka 2.1 may be better in blitz. Very possible, but when i ran 50 games at 5 5 settings, Rybka 1.2f was leading by ...[text shortened]... t be very low, perhpas 3 3 or even faster.
All I can say so far is, I'm really dissapointed.
Many engines (of course not all) have parameters you can tinker around with - and 15-10 controls try turning down the selectivity and prunning a little (if possible)
Originally posted by exigentskytrue....
But wouldn't doing so actually make the match rather pointless? I thought both should have the same settings for fairness.
anyway, I think I have found the awnser to your question.....
http://www.rybkachess.com/index.php?auswahl=FAQ+for+v+1.2f
There it is, clear as day....
Is the playing strength changed from Rybka 1.2?
No - unless losses on time, or forfeit losses, are counted as a part of playing strength.
Originally posted by lafredoOh.....so it is - in which case I stand correct.....
I may of missed somthing but i think that link compares Rybka 1.2 and 1.2f not 1.2f and 2.1 as in the question
well from what I read 2.1 was for multi-processors, --- in which case the increase in playing strength is not from imporved evaluation functions but rather, more hardware.
Yes, 2.1 is for multiprocessors/dual core processors but it is also for the average 32 bit single core computer. (mine) Also, according to the help file, it should also have had some strength improvement.
Here are the differences between 1.2 and 2.1: http://www.rybkachess.com/download/Rybka_2_1.htm
In any case, you may be right that the search improvements actually lead Rybka to perform worse with longer time controls because it misses the less obvious moves. Thus, I have cancelled the 15 10 matches and started another tournament but with 5 0 time settings. I'll wait until at least 100 matches before commenting on the results.
Even if the strength improvement between 2.1 and 1.2 is minor, I can't understand why there would be a downgrade in strength.
I don't think you're alone in having some doubts about the strength improvement of 2.1. I've read similar posts elsewhere.
However, it's still early days. With more test data things will become clearer and more definite. Also, regarding your own tests, it's not always best to just test various versions of the same engine. e.g. it may be that 2.1 performs better against Shredder, Fritz, etc. than 1.2.
Originally posted by VarenkaAlso, anyone who bought 2.1 automatically gets 2.2 and 2.3 later this year.
I don't think you're alone in having some doubts about the strength improvement of 2.1. I've read similar posts elsewhere.
However, it's still early days. With more test data things will become clearer and more definite. Also, regarding your own tests, it's not always best to just test various versions of the same engine. e.g. it may be that 2.1 performs better against Shredder, Fritz, etc. than 1.2.
In other words, Rybka 2 is still a work in progress.
I bought Shredder 10 recently and having watched the free beta version of Rybka repeatedly kick its ass, now have no idea why.
Originally posted by exigentskyyour right 2.2 isnt due out for another two months
Rybka 2.1 is stronger if you have teh right hardware as it supports multiple processors. However, on equal hardware, I have my doubts. So far, 70 game s have been played. I'll give the results in a few more hours.
Rybka 2.2 does not yet exist.