As I said in another thread, I thought this might possibly be too big an elo jump for Bouquet- but I went ahead anyway. It was set up to be a 100-game match, but my first check was at 75 games- and I did not have to think- I immediately stopped it then and there. CPU/core time is more valuable than this fiasco. Anyway:
Intel i5 w/4TCs
Fritz 13 gui
1CPU/64bit
128MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/3 Repeating (Benched to adapt to 40/4) Match=100 games (aborted after 75 games!)
Look, when we speak of spark-1.0, we are not talking about a top 10 engine obviously. However, I doubt there are many, if any, engines below the Top 20 who could make spark look this bad. What can I say?
One hell of a performance by Bouquet- after the first move out of the opening in game 1- spark never had a chance!.
I have included the PGNs for this match, as well as the 3 previous matches I have run. Anyone who wants any of the PGN sets should get them while they are still available.
estimate rating for spark based on this match is even bigger than based on the match against zappa
Zappa Mexico II 64-bit 2870 performance +164 based on 50 games
Spark 1.0 64-bit 2954
2870+164=3034 ccrl performance based on 50 games
2954+147=3101 ccrl performance based on 75 games.
75 is more than 50 so my estimate now is near 3074
number 7 in the CCRL list of single engine Fritz 13 (rating 3026) is probably weaker
when
number 6 in the CCRL
Rybka 4.1 64-bit(rating 3167) is too strong for Bouquet.
You can choose something like Rybka3 32 bits(CCRL rating 3092) for the next match or Rybka3 dynamic 64 bits(CCRL rating 3087)
or old stockfish(Stockfish1.7.1 64 bits(CCRL rating 3095)).
Uri Blass wrote:estimate rating for spark based on this match is even bigger than based on the match against zappa
Zappa Mexico II 64-bit 2870 performance +164 based on 50 games
Spark 1.0 64-bit 2954
2870+164=3034 ccrl performance based on 50 games
2954+147=3101 ccrl performance based on 75 games.
75 is more than 50 so my estimate now is near 3074
number 7 in the CCRL list of single engine Fritz 13 (rating 3026) is probably weaker
when
number 6 in the CCRL
Rybka 4.1 64-bit(rating 3167) is too strong for Bouquet.
You can choose something like Rybka3 32 bits(CCRL rating 3092) for the next match or Rybka3 dynamic 64 bits(CCRL rating 3087)
or old stockfish(Stockfish1.7.1 64 bits(CCRL rating 3095)).
next thread- check!
Last edited by geots on Tue Jul 31, 2012 11:26 am, edited 1 time in total.
Uri Blass wrote:estimate rating for spark based on this match is even bigger than based on the match against zappa
Zappa Mexico II 64-bit 2870 performance +164 based on 50 games
Spark 1.0 64-bit 2954
2870+164=3034 ccrl performance based on 50 games
2954+147=3101 ccrl performance based on 75 games.
75 is more than 50 so my estimate now is near 3074
number 7 in the CCRL list of single engine Fritz 13 (rating 3026) is probably weaker
when
number 6 in the CCRL
Rybka 4.1 64-bit(rating 3167) is too strong for Bouquet.
You can choose something like Rybka3 32 bits(CCRL rating 3092) for the next match or Rybka3 dynamic 64 bits(CCRL rating 3087)
or old stockfish(Stockfish1.7.1 64 bits(CCRL rating 3095)).
I agree, Uri- except look at the CCRL 40/40 list that has single-cpu engines- 64 bit if available, and spark is rated quite higher than appa Mexico II. I will check- give me a sec. I just checked and here is the link you can see for yourself, or just take my word for it: http://www.computerchess.org.uk/ccrl/40 ... ons_only=1
spark 1.0 64-bit 1CPU= 2974
Zappa Mexico II x64 1CPU= 2890
I cannot guarantee that is an accurate rating or not- but I would have no reason to expect it not to be.
I understand what you are saying tho, but that is not what stumps me here. And I think you know what I am referring to. If you don't get my drift- PM me and give me your email address. I want to discuss something with you besides this that is extremely important- but in private. Please do this in the next hour- I have to then get some sleep.
