Author: Mogens Larsen
Date: 03:51:24 03/06/01
Go up one level in this thread
On March 05, 2001 at 19:56:03, Fernando Villegas wrote: >Hi: >I was not very satisfied with Gandalf f and g -and I had sometimes a somewhat >harsh discussion about that with Mogens- but the new update, "h", seems to be a >very different animal. Just as a little note. We didn't argue about the playing style, because that is a very subjective topic. But I objected to nonsensical idea of (relative) uniform program performance disregarding effects from changing hardware/timecontrol. Generally, I don't care too much about other opinions than my own when it comes to playing style and I think most people do the same. Not a fruitful topic for debate IMO. >I have not masive data to support my impression. Just >some games, but, as Thorsten, I believe that after many years playing programs a >chess computer nerd can get some accurate or at leats not very distorted >perception with very few elements in hand. Thorsten has said that he just need >some moves: I need some games. Then you're both wrong. Chess programs aren't deterministic enough to reveal identifiable patterns of correct or incorrect play within a subset of related positions from a few games. The more the merrier. Fewer games just mean personal or psychological preferences and not analytical observations. >To begin with, it seems to me that this update is more agressive and >enterprising, a lot more gifted with the so called "killer instinct". With >Gandalf F and G I felt I was playing a somewhat pasive program waiting for your >mistakes; H, on the contrary, goes for your king a lot quicker. >If I am right, this could be the result of a huge rewriting or just some touches >here and there capable of important different outputs. This last posibility is >not so imposible, as the case of Gambit Tiger shows. According to Christophe, he >just made some changes in the original code of Tiger to give his cousin a >greater inclination to attacking moves to the king, but nevertheless, as >everybody here knows, those changes produced a great differential in playing >style. I would like to know something about that from the authors of the program >or from Mogens, if he does not mind. I'm not sure if the program author follows this forum that closely, which is probably a sensible move. And since I'm not associated with the Gandalf Team as such, except for the occasional testrun, there isn't a lot of information to be revealed. However, I'm quite certain that the "h" version isn't a rewrite of the previous versions. The emphasis were on bugfixing, eg. learning and removing the extension responsible for the loss against DF at Paderborn, and some tweaking of the evaluation. I imagine that the tweaking of various parameters was based on experinces gained at chess servers. According to Frank Q. its also a bit faster, but I haven't tried to check that. So even though I'm tempted to say that nothing has changed to make you look silly :-), I really can't say for sure. The timespan also suggests that it would be difficult to implement and test new ideas en masse. I think that this is reserved for Gandalf 5. Mogens.
This page took 0.01 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.