Author: Robert Hyatt
Date: 07:19:36 10/10/99
Go up one level in this thread
On October 10, 1999 at 01:45:22, James B. Shearer wrote: >On October 10, 1999 at 00:10:19, Robert Hyatt wrote: > >>On October 09, 1999 at 23:26:16, James B. Shearer wrote: >> >>> I'm not playing 5 0 because I like it, this is the only time control >>>your formula allows me to play (I also played a few 2 1 games a while ago but >>>this really is kind of hopeless especially with the interface I'm using and my >>>rating got knocked out of range). I expect I would do a bit better at 3 3 or >>>1 6. >>> >>> <deletions> >> >>It should play almost anything from 5 3 on down (blitz) or any standard time >>control up to 30 30. Is something broken that I don't know about? > > I believe if you check your formula for scrappy you will find it will >play rated games if you are within 400 rating points at a variety of time >controls but will only play 5 0 unrated. Since I am a bit more than 400 rating >points lower than scrappy I can only play unrated 5 0 games. > ok... forgot you were playing unrated. I try to limit unrated games as there is never any telling who/what I am playing. Many feel that it is perfectly ok to use a computer when playing unrated, since no rating points are at stake. This makes it harder to interpret the results. My restricted unrated formula does allow most everyone to play, but discourages the above. IE on ICC I don't allow unrated games at all unless someone specifically asks. > <deletions> > >>All I can say about speed is that on the quad P6/200, it runs about 400K >>nps normally, with blips to 500K+ in endgames. However, it taking an unusual >>amount of time on one move doesn't guarantee that it 'failed low' at all >>when the machine gets loaded. Unfortunately I now can't tell exactly as the >>automatic script that cleans out old log files has run and deleted all those >>logs... > > Perhaps you should save the logs for losses (at least for a while). >This shouldn't take much space if they are really few and far between. > James B. Shearer I save them for a long while if they are important. But when I look thru a log and see the entire game played with 1% cpu utilization, or even with a few critical moves played with 1% cpu utilization, I ditch 'em as that isn't likely to be an engine problem. As far as the number of losses goes, I had a really interesting conversation with "cptnbluebear" (an anonymous GM on ICC) yesterday. He is _very_ strong vs computers, and has been one target of my aggression tuning as he is very happy to draw whenever it will allow him to. His question was "do you know how I am doing vs crafty, overall?" I didn't but did a quick search and found that Crafty had won about 400 games, had lost about 60, and had drawn almost 200. He was quite happy and said "wow, 25 percent is very good (he pointed out that this meant he was within 200 points of the computer at blitz -- he always plays 5 3 games). I was quite surprised by a GM saying that 25% is _very good_ as most don't think like that. But he pointed out that 25% was the _best_ result he had seen against Crafty on ICC, against any human player. An interesting perspective, and it also shows just how rare losses are. BTW I do try to save all computer vs crafty games as they are much more revealing since they have far fewer tactical blunders in them...
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.