Author: Robert Hyatt
Date: 11:37:56 01/25/00
Go up one level in this thread
On January 25, 2000 at 13:09:52, Ernst A. Heinz wrote: >On January 24, 2000 at 23:10:12, Robert Hyatt wrote: >> >>On January 24, 2000 at 16:02:48, Ernst A. Heinz wrote: >> >>>>Sure, but it did full singular extensions and the rest of the things we were >>>>all doing then. >>> >>>I do not remember having heard or read Hans Berliner talk/publish >>>about singular extensions in "Hitech". >>> >>>Please enlighten me where you got that information from. >> >>From sitting across the board from them at an ACM event. Murray was at >>the table as well, and we had a long discussion about singular extensions >>after deep though was using them. They implemented them in HiTech very >>quickly after the idea was published... That was the reason I added them >>to Cray Blitz... I had used this sort of idea a long while back, at the >>suggestion of someone whose name I don't recall. It didn't work well in >>1978 and I removed them from the program just prior to the ACM event that >>year, because it costs us well over one ply, and going from 5 to 4 was a >>killer. >> >>They were using them in at least the last few ACM events... I don't recall >>when they added them. You could probably ask Murray since he probably wrote >>the code for adding them to HiTech. They were also using Ken's databases in >>the search (not only at the root) way back then too... > >Interesting. > >Hans Berliner visited us in Karlsruhe in summer 1996 for >almost one week and I also had considerable email contact >with him about "Hitech". > >He never said anything about singular extensions then. > >Moreover, we let the old "DarkThought WMCC'95" running on >a slow 275MHz Alpha-21064a (ca. 30K nps) play some test >games against "Hitech B*". This old and slow version of >"DarkThought" won all the games although Berliner claimed >that "Hitech B*" was even stronger than the normal "Hitech" >(and yes, in the endgame, Berliner manually switched to the >normal "Hitech" during these games). B* certainly didn't use SE. But it also was later in the development. The last ACM event was 1995. I don't recall when I had this conversation, but it was definitely prior to then, as by that time Murray was an IBMer with Hsu. When we had this conversation, Murray was explaining what they did and how they did a couple of things 'cheaper' as (in his words) "Hitech just doesn't have the horsepower to do everything that is done in Deep Thought. (with respect to SE). Generally SE is dangerous in endgames as well, as too much gets extended, and the overhead isn't worth the effort spent... so it is likely what you played had no SE of any kind... I don't remember when he switched to B*... I do remember that he originally asked some of your group to develop a fast engine for him to produce the bounds that B* needs, didn't he??? That should put a time-frame for dropping the original hitech with SE in favor of B*. > >Of course, we did not play enough games to gain any >statistically significant insight into the relative playing >strengths of both programs. But I was extremely disappointed >by the performance of "Hitech" at that time -- especially >because it was roughly 5x faster in terms of nodes searched >per second. > >=Ernst= I didn't think the B* version played well at all. It might have been better than his normal HiTech (in head-to-head, somehow) but it seemed to play badly at ACM events, _never_ doing well after it was implemented.
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.