Author: John Merlino
Date: 12:24:46 02/22/01
Go up one level in this thread
On February 22, 2001 at 10:35:12, Elizabeth Schwartz wrote: >On February 21, 2001 at 21:12:40, John Merlino wrote: > >>Thank you for saying so. However, there were only five personalities that were >>used in the test: Chessmaster, Josh Age 9, Willow, Sonja and Skippy -- basically >>personalities with ratings of approximately 2500, 2000, 1500, 1000 and 500 >>respectively. > >That's good to know! I wish there were one at my level. I'm currently somewhere >between 600 and 900, but I don't know where! BTW Skippy shows up as "643" on my >PII-266 laptop and "Sonja" shows up as 824. I have 10 600-level players, a >707, a 790, and two 800-level. > >I suspect that processor speed doesn't make a huge difference with the beginning >players just because they aren't looking ahead that much anyway and they're >still making dumb mistakes like hanging major pieces. If you're going to hang >your queen for no good reason, it doesn't matter how much you think about other >stuff. > >I don't feel that I have a good sense of what my "real" rating would be, at >least partly because the low-rated players seem erratic. Now that I know that >"skippy" I'm starting to play a lot on WCN but I don't really know how good >those ratings are (or where a good place to *ask* is). I got my first rating at >905 and am dropping steadily. I'm winning maybe one game in four against other >newcomers with high 3-digit ratings. > >At my level, with CM8000, the big temptation is to treat it like a video game >and go for the "high score" by playing tournament games mostly against the lower >rated players. But, it is more educational to play games that I am guaranteed to >lose against the higher-rated players. At this point, CM8000 still has me as >600-something, but when I play the 600-level opponents I get bored and sloppy. >Playing humans on the online servers keeps me honest. > >If "Sonja" is really accurate I think I will spend more time playing her! Yes, it can be inferred from the testing that those five personalities have the most accurate ratings. The reason that Sonja shows up as 824 and Skippy at 643, even though they were SUPPOSED to be 1000 and 500 respectively, is simply because that was their ratings (approximately) in CM7000, which shows just how far off the ratings were without "real-world" numbers thrown in. CM7000's ratings were strictly based on comp vs. comp games (almost 23,000 of them). But, all those games were thrown out for CM8000 because of the new engine. So, the ratings in CM8000 are based on about 3,000 human vs. comp games (all with those five personalities), and about 10,500 additional comp vs. comp games to fill out the ratings. Each personality that was not in the human vs. comp testing played about 100 test games, resulting in the final ratings. jm
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.