Random Posts

Monday, March 5, 2012

FireBird Followup

Back in April, 2010, I reported on how FireBird fared against Houdini and that the results were pretty much even in long time controls.  Houdini gets a slight edge at faster time controls though.  It’s the long time controls that are important because that’s what you’ll use when analyzing.  Unfortunately, most engine vs. engine tests are actually at fairly fast time controls.   
Firebird has had many names over the course of its development. One of the early names was RobboLito.  People accused RobboLito of being a cloned copy of Rybka and because of that the program was banned from use on many sites and refused testing.  Despite all the brouhaha, as far as I know, its developers never had a case heard against them by the engine police.
Robbolito’s early development was fast paced and eventually the developers went their separate ways and as a result Firebird was born.  The argument was that Firebird was stripped down in its error checking code in an effort to make it calculate faster and the result was that at faster time controls, it was better that Rybka.  But in long controls, so the theory went, Rybka would make fewer mistakes and therefore be the stronger of the two.  In one independent test at a time control of 40 moves in 2 hours, after 32 games the result was tied at 16 points each.
Still, Houdini remains the most popular free engine.  The latest version now costs about $50, but as I pointed out in an earlier post, its rating is only about 25 points over the free version so right now there seems little point in parting with $50 just for 25 points.
FireBird is also considered to be one of the best programs in the world and it is free. There is also some controversy associated with Firebird as Vasik Rajlich, the founder and head of the Rybka team, has alleged that the programmers of Firebird copied nearly all aspects of Rybka in their design. It’s probably no surprise that his claims have not been been substantiated.
My own personal (and not very scientific) testing has pretty much lead me to believe that FireBird is the engine of choice when it comes to playing Advanced Chess at LSS. 
It was a little less than a year ago that I began using FireBird along with my game database and opening book consisting of top level CC and engine games.  Since that time I have completed 39 games.  Of course it is impossible to know what engines my opponents were using and how long they allowed their engines to analyze, so the results don’t really prove anything.  Besides that, you also have to take into account that probably half of my opponents, like myself, did not rely solely on engines to select moves, so that would definitely affect the outcome, too.  The point is that I have been using FireBird exclusively and I was interested in seeing if its use had a drastic effect on my results.
Prior to using FireBird my engine of choice was Houdini and I had compiled a record of 26-28 with it.  Using FireBird my record is 20-19.  The significance of the results doesn’t mean much in itself, but considering I have obtained a draw with an ICCF Senior Master and gained 76 rating points since using FireBird's assistance, I can’t be unhappy with the results.  Actually the results would have been better except they were somewhat skewed when in one event I (foolishly) used the Beginners Opening which resulted in positions so inferior no engine in the world would have saved me and ended up with the worst results I’ve ever had, losing 4 games and drawing two. 
I saw one report where the Engine Parameters were tweaked and it played even better.  I have never bothered with this though because it was done on the single processor version and whenever you tweak the parameters it seems that they always revert back to the default when you close the engine.  That means resetting them every time you restart and I’m not interested in doing that.
So, while the results using FireBird instead of Houdini haven’t been drastic, I think they are sufficient that FireBird will continue to be my engine of choice for analysis.