geots wrote:You should test RobboLito 0.09- it is latest and strongest. But be aware there will be no more RobboLitos. Development is over. All effort goes now to one engine- Ivanhoe.
OK I'll test IvanHoe 46a only then (when my test of Daydreamer 1.75 is finished). I'll post results on my Open Source Blitz Rating List (cf. my thread in this forum). How does it compare to StockFish 2.2.1, you reckon ?
Not sure in Linux, but in Windows 32 bit any of the 46 versions and some of the older 47 versions will walk thru Stockfish 2.2.1, Fritz13, Komodo 4, Rybka 4.1 like they aren't even there. They have no chance. I have, however only tested 32bit. If you want to see many results, start back on page 3 or 4 here and come forward. You will see EVERY ONE of my Ivanhoe match results ag. above engines. But you will also see that PeterPan's compiles are by far the strongest, and you have only KLO's in Linux. So you cannot run 46a- you have to run 9.46a, and that is KLO's- as I told you Izak (PP) made no Linux compiles.
Only 2 engines in existence that can stay with the Ivanhoes are Critter and Houdini. THAT IS IT!!
IvanHoe 999946a is definitely very strong. Whether or not he will beat Stockfish 2.2.1 in my Open Source Blitz Rating List is unknown so far. Only 15 games out of 250. But so far, IvanHoe is slightly ahead of SF.
lucasart wrote:IvanHoe 999946a is definitely very strong. Whether or not he will beat Stockfish 2.2.1 in my Open Source Blitz Rating List is unknown so far. Only 15 games out of 250. But so far, IvanHoe is slightly ahead of SF.
IH 47-46 (Peterpan or other good compiles) are just a bit stronger than SF2.2.1, depending slightly on the time control. Their best qualities are in analysis, you can dig deeper into the source, their GUI, their pretty unique endgame bases. By "their" I mean those funny "comrades".
lucasart wrote:IvanHoe 999946a is definitely very strong. Whether or not he will beat Stockfish 2.2.1 in my Open Source Blitz Rating List is unknown so far. Only 15 games out of 250. But so far, IvanHoe is slightly ahead of SF.
IH 47-46 (Peterpan or other good compiles) are just a bit stronger than SF2.2.1, depending slightly on the time control. Their best qualities are in analysis, you can dig deeper into the source, their GUI, their pretty unique endgame bases. By "their" I mean those funny "comrades".
Kai
Forgot to add: IH is optimized to 32bit (all ippo-related are), and seems stronger by ~25 Elo points compared to non-ippo-related 32bit engines. Therefore, if you run 32bit engines, it's significantly stronger than SF2.2.1 32bit.
OK I've finally got the best linux version which is 999946f (46h doesn't work as it requires glibc 2.14 that even the most up to date debian based distribitions don't have).
I'm running Pepito 1.59 now, and will test 46f next. I'll delete the games of 46a and rerun the same test with 46f.
How much elo do you reckon I should expect between 46a and 46h ?
lucasart wrote:OK I've finally got the best linux version which is 999946f (46h doesn't work as it requires glibc 2.14 that even the most up to date debian based distribitions don't have).
I'm running Pepito 1.59 now, and will test 46f next. I'll delete the games of 46a and rerun the same test with 46f.
How much elo do you reckon I should expect between 46a and 46h ?
Hard to tell with KLO's. I don't have as much to go on. Keeping in mind I have not run KLO's compile of "f" or "h" in 46, i would say I would be surprised if either were as strong as 46a that you ran. I know that 46f and 46h are newer, but I don't think either is as strong as 46a or 46e. But you have thrown a wild card in the deck. Figuring Linux is hard enough, but now I gotta allow for a comparison between your compiles and his. Some jobs are just too tough.
lucasart wrote:OK I've finally got the best linux version which is 999946f (46h doesn't work as it requires glibc 2.14 that even the most up to date debian based distribitions don't have).
I'm running Pepito 1.59 now, and will test 46f next. I'll delete the games of 46a and rerun the same test with 46f.
How much elo do you reckon I should expect between 46a and 46h ?
Hard to tell with KLO's. I don't have as much to go on. Keeping in mind I have not run KLO's compile of "f" or "h" in 46, i would say I would be surprised if either were as strong as 46a that you ran. I know that 46f and 46h are newer, but I don't think either is as strong as 46a or 46e. But you have thrown a wild card in the deck. Figuring Linux is hard enough, but now I gotta allow for a comparison between your compiles and his. Some jobs are just too tough.
george
i bet my compile is faster. I'll show you some evidence of it when my computer is free (currently running multithreaded testing with cutechess-cli, that is two games in // on 2 CPU, one thread per engine)
but really linux is easier than windows. micri$oft wants you to believe the contrary, or couse, but it's just capitalist propaganda
Laskos wrote:Forgot to add: IH is optimized to 32bit (all ippo-related are), and seems stronger by ~25 Elo points compared to non-ippo-related 32bit engines. Therefore, if you run 32bit engines, it's significantly stronger than SF2.2.1 32bit.
Kai
Haven't seen any evidence of that in my testing so far, but I'm testing at a slower time control.
lucasart wrote:IvanHoe 999946a is definitely very strong. Whether or not he will beat Stockfish 2.2.1 in my Open Source Blitz Rating List is unknown so far. Only 15 games out of 250. But so far, IvanHoe is slightly ahead of SF.
IH 47-46 (Peterpan or other good compiles) are just a bit stronger than SF2.2.1, depending slightly on the time control. Their best qualities are in analysis, you can dig deeper into the source, their GUI, their pretty unique endgame bases. By "their" I mean those funny "comrades".
Kai
Forgot to add: IH is optimized to 32bit (all ippo-related are), and seems stronger by ~25 Elo points compared to non-ippo-related 32bit engines. Therefore, if you run 32bit engines, it's significantly stronger than SF2.2.1 32bit.
Kai
i DO NOT run 32 bit engines, only 64 bit running with 1 thread. and the reason is that there is no good reason to double the testing work by 32 and 64 bit versions, and that all computers have 64 bit CPUs since 2003... now some people may run a 32 bit operating system on a 64 bit machine, typically capitalist windows users, but it's their problem...
and ivanhoe 999946h definitely beats StockFish whether on 32 or 64 bits. Note that SF is also very well optimized for 32 bits. that's the first thing that marco did when he worked on SF 1.0 from the code of glaurung
