Author: Robert Hyatt
Date: 12:44:02 11/29/97
Go up one level in this thread
On November 29, 1997 at 14:42:47, Leon Stancliff wrote: > >I wish to thank both Mike Byrne and Robert Hyatt for responding to my >message regarding the strength of Hiarcs 6 for MacIntosh. I also wish to >express my gratitude to Bob for the impetus he has provided to the ICC. > >Mike quoted information on Hiarcs 4 for MacIntosh rather than for Hiarcs >6. Bob followed Mike's line of thought. However; Hiarcs 6 is much >stronger, perhaps as much as 150 points. > >I would like to keep my games using Hiarcs 6 under wraps until it has >reached the established rating level of 20 games. Then the results will >be made available to each and all who may be interested. At this point >Hiarcs 6 Mac has played 14 rated standard games. > >Mike challenged Hiarcs 6 for any standard time control up to about six >hours for the match. I would rather play a four game match with Fitter >at a time control of 15 15. Mike, as soon as Hiarcs 6 for Mac has >reached the 20 game established rating I will be making contact to see >if you are still interested. > >I might add that my remark about Hiarcs 6 being significantly stronger >than the version of Crafty playing on the ICC was limited to Crafty and >not Moonshot or Fitter. Both of the latter programs are rated higher in >standard time limits than Crafty is. > I'm not sure exactly what you mean. I can interpret your remarks two ways: (1) you are simply comparing ratings; (2) you have played lots of games against crafty. I haven't seen but a couple of games vs crafty, so I assume (2) is not your point. (1) is not a reasonable comparison, because "crafty" is the development account for crafty. It often plays badly, sometimes plays well, because this is where I do all of my testing. Fitter, Moonshot, Data and others are simply older versions (excepting fitter usually) that were tested before being released. Crafty's rating on ICC fluctuates by at least 200 points regularly as I break and fix things... >A figure which might be revealing is the results of the LCII test. >Hiarcs 6 for MacIntosh (operating at 132 Mhz) received a rating of 2515. >Do we have similar ratings for the version of Crafty playing on the ICC, >or for either Moonshot or Fitter? These tests are "interesting" but not revealing. The point is, how well do the programs do against each other, and against strong humans. Either of these can be answered quite easily of course... IE the last time I ran Win at Chess on crafty v14.2, at 1 minute per move, it missed only #141, getting 299 correct. Is this better than Hiarcs? Probably. Does this mean Crafty is better? Hardly. So throw out the test suites, unless you want a program that solves that kind of position... > >Remember also that Hiarcs 6 will be operating without the benefit of an >automatic interface. Perhaps the strength of this program will inspire >someone to write one. > I ran Crafty for a long time on ICC manually. I don't believe the auto interface means anything at standard time controls. For good operators it doesn't mean anything much except at bullet time controls. I'd personally prefer to play any program using an auto interface on both ends, solely to keep the human operator "out of the game"... which is a big advantage and a common problem on ICC... >Thanks again. It is easy to strike a nerve without intending to... > >Leon Stancliff I only want to see "correct" results. If you can actually beat Crafty most of the time, the it is obviously better. But this can easily be settled on the servers. Just watch out "who" you play, as some crafty's are intentionally dumbed down, some play with gross opening books, etc...
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.