Author: Melvin S. Schwartz
Date: 11:40:45 06/08/99
Go up one level in this thread
On June 08, 1999 at 06:06:34, Bertil Eklund wrote: >On June 07, 1999 at 23:37:07, Melvin S. Schwartz wrote: > >> >>On June 07, 1999 at 20:24:59, blass uri wrote: >> >>> >>>On June 07, 1999 at 19:49:17, Melvin S. Schwartz wrote: >>> >>>> >>>>On June 07, 1999 at 12:37:07, blass uri wrote: >>>> >>>>> >>>>>On June 07, 1999 at 11:16:13, Shep wrote: >>>>> >>>>>>On June 07, 1999 at 10:15:55, Melvin S. Schwartz wrote: >>>>>> >>>>>>>Since Hiarcs 7.32 and Fritz 5.32 are condidered to be two of the best three >>>>>>>prgrams available, if not the two best, wouldn't it be inteesting to run Hiarcs >>>>>>>in Fritz, or Fritz in Hiarcs, and have Fritz play tactical and Hiarcs play >>>>>>>positional positions? Of course the question then becomes, would the program >>>>>>>accurately know when to switch? Anybody have thoughts on this? >>>>>> >>>>>>Unfortunately, it is not very clear how Fritz differentiates between these >>>>>>types. I noticed that, running the Louguet suite, even most of the POS subsuite >>>>>>is considered "Tactical" by Fritz (IIRC only 5/14 are "positional" for Fritz). >>>>>> >>>>>>I have played some games with T:Fritz P/E:Junior (I called this combination >>>>>>"ClaireChess" on SCCS) and noticed that either almost the whole game was played >>>>>>by one engine alone or there was exactly one point where Fritz switched from one >>>>>>engine to the other and never switched back again. >>>>>> >>>>>>I think Chessbase will sooner or later have to make this function modular and >>>>>>programmable because it is still very difficult to decided which positions are >>>>>>tactical and which aren't. >>>>> >>>>>There is a simple solution. >>>>>look for the first 5 options for 1% of your time. >>>>>If the difference between the best move and the 5th best move is more than a >>>>>pawn then the position is tactical otherwise it is positional. >>>>> >>>>>You can change the numbers but the idea is clear. >>>>> >>>>>This is of course not a perfect solution but does someone has a better idea? >>>>> >>>>> >>>>>I do not know if Fritz5.32 is better in tactical positions. >>>> >>>>Whether they are right or wrong, I'll present my sources for saying Fritz is >>>>better in tactical positions. Mr. Giehring at Chessbase sent me a message, which >>>>I have saved, stating Fritz 5.32 is slightly better in tactical positions than >>>>Hiarcs 7.32. Also, in a review by Claudio Bollini, which can be read at this >>>>site, he shows a diagram based on a particular test done to evaluate different >>>>programs that I believe indicates Fritz to be better at tactics. >>>> >>>>>I read that Hiarcs7 is a good solver so it is good at tactics and I know about >>>>>ssdf games that Fritz5.32 won Hiarcs7 because of better endgame play and not >>>>>because of tactics. >>>> >>>>This is very surprising in light of the review of Hiarcs 7 where in the "endgame >>>>hits test" Hiarcs was excellent along with MC8. I believe he rated those two as >>>>tops in endgame play according to the "hits"test. >>>> >>>>>Does someone has a proof that Fritz is better than Hiarcs7 or Junior in >>>>>practical tactical positions? >>>> >>>>I believe you think quite highly of Junior. For that reason I am baffled by >>>>Junior's poor showing in SSDF testing. How do you compare Junior 5 against >>>>Hiarcs 7 and Fritz 5.32 in overall playing strength, and by what criteria do you >>>>base your conclusion? >>> >>>I know that James walker did a match between Junior5 and Fritz5.32 >>>(90 minutes per 40 moves) and the result was 6:6 >>> >>>I know that Junior5 lost 28.5:11.5 against Hiarcs7 but the problem is that the >>>games are not public. >>> >>>Junior5 did better results in public games against Hiarcs7. >>>I know that it won Hiarcs7 6:4 in enrique games(40 minutes per 40 moves on a >>>fast pentium) >>> >>>I also read that Hiarcs7.32 had problems against Junior5 in the test games >>>and I did not read about results like 28.5:11.5 >>> >>>Junior5 earns more speed from the fast computers relative to Fritz5.32 or >>>Hiarcs7 >>> >>>Here are the numbers: >>> >>>AMD K6-2 450 vs P200MMX >>> >>>Speed improvement: >>> >>>Fritz3 130% >>>Genius2 119% >>>Gandalf3 84% >>>McP6 117% >>>Rebel8 148% >>>Hiarcs7.01 137%=2.37 times faster >>>Fritz5.32 114%=2.14 times faster >>>Junior5 142%=2.42 times faster >>>Shreddr3 106% >>>Nimzo98 147% >> >>I know that at 40/2 with a 200MHz processor, Fritz needs 72 megs of RAM. At 40/2 >>with a 400 MHz processor, Fritz needs 144 megs of RAM. Without knowing more >>details from your posting above, it is impossible for me to come to a valid >>conclusion. I don't know if Junior uses Hash Tables in the same way Fritz does, >>but I do know Hiarcs does not. If you could provide the time control used for >>the above test and Fritz's allocated Hash Table usage, then I could evaluate the >>above with more authenticity. Unless of course, the time control and Hash Table >>total are both irrelevant? >> >>Mel> >>>Uri > >Hallo! > >All tests is done with the same hash-table size. Typically a doubling of the >hash-size gives around 7-8% extra speed. The only exception I know is Fritz. >Fritz5 +20-90%. This according to several tests I have done on tournament level >3-10 min/move. Hello Bertil! Thank you for the info above. I would be grateful if you could tell me if Fritz 5.32 had 144 megs of RAM with your AMD K6-2 450 processor what the difference would be in speed versus a P200 at tournament level of 3 minutes a move? Or if you must calculate at 3-10 minutes per move as with the other programs, then we have to re-calculate the Hash-Table requirement for Fritz based on the formula recommended by the manual: 2 X processor speed X seconds per move = Hash-Table size needed. Regards, Mel > >Regards Bertil SSDF
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.