Author: Stephen Ham
Date: 14:12:41 02/27/04
Go up one level in this thread
On February 27, 2004 at 10:28:57, Geert van der Wulp wrote: >On February 27, 2004 at 10:20:08, Stephen Ham wrote: > >>On February 27, 2004 at 08:46:30, Geert van der Wulp wrote: >> >>>David, Keith >>> >>>Thank you for your replies. This is the information that I was looking for. >>>Fortunately there are people here who can still read. >>> >>>Regards, >>> >>>Geert >> >>Dear Geert, >> >>While we can all read, it is unfortunate that your message was ambiguous, to say >>the least. For example, your subject line addressed strictly evaluation. However >>your posts shifted the focus to human intuition and style and then later shifted >>again to anaylsis - three entirely different things. >> >>Here's my opinions, for whay that's worth: >> >>In general, Shredder 8 is probably the strongest engine, in computer versus >>computer action. Is it the best against humans? Who knows? My perception is that >>Shredder 8 is indeed very strong, but its evaluation is not to be trusted and it >>doesn't have a particulary human playing style. So for analysis of middlegame >>positions, it will indeed find strong moves, worthy of further investigation, >>but it probably won't evaluate the position accurately. >> >>Shifting to one of your other topics, human-like style, then I think the clear >>preference is for Rebel 12. While it won't deeply probe tactical positions like >>Shredder 8 can, it seems to produce very human-like moves and almost gives the >>feeling that it understands the position and plays accordingly. Hiarcs isn't bad >>either in this respect. Rebel 12 is also a fine analysis machine if you alow it >>an extended analysis period (it doesn't seem to perform well if it's allowed >>brief amounts of time). >> >>Shifting to your other topic, evaluation, Shredder 8 is a failure here. My >>favorite is Fritz 7 (I don't have Fritz 8) for the most accurate evaluations of >>positions. Fritz is generally an accurate evaluator throughout the game >>(opening, middlegame, and endgame). >> >>But Geert, the above is all relative to the position at hand. Some engines are >>particularly strong/weak in certain types of positions. They each have their >>peculiarities, so you need to experiment to find what gives you the greatest >>confidence. For me, it's the experimentation in various positions that I enjoy >>doing the most, with my engines. >> >>All the best, >> >>Stephen > >Stephen, > >Thank you for your reply. Your points make sense to me. Even the ones about the >chaotic shift in the nature of my questions ;-) > >But I have been given the link > >http://www.computerschach.de/test/index.htm > >where a lot of positions from World Chess Championchips (human) are given to the >programs for evaluation. So I can see for myself in which positions a certain >program is strong and what his (her?) weak points are. > >By the way, in my questions I always made it clear that I am looking for an >engine which has a good "tactical feeling", because I know that every discipline >will give different results. > >Regards, > >Geert Hi Geert, Thanks so much for that link. That was revelation to me! I must confess that I'm surprised by the reported results. For example, I'd rate Rebel 12 and Yace Paderborn MUCH higher in all three categories. Also while it's natural to expect that Gambit Shredder 8 would be the top King attacker, it's a shocker to learn that Gambit Shredder 7.04 is the #1 "Positional" player. I've been disapointed with Shredder 7.04 and Shredder 8 regarding positional play, finding them to be exclusively attackers. For more on that, here's my review at ChessCafe.com: http://chesscafe.com/text/review365.pdf And then to see "Chess Academy 6.0 middlegame" as #3 in the endgame skill test is the biggest shock. What do you and the other readers think of these test results? I'm not saying they're wrong. I have the very highest respect for Manfred's work, and so I believe they are indeed legitimate figures. I just found some of them to be a surprise because they differ from my tests. But my manner of testing may be less scientific than Manfred's. All the best, Stephen
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.