Author: Dann Corbit
Date: 15:20:07 10/02/02
Go up one level in this thread
On October 02, 2002 at 18:15:04, Uri Blass wrote: >On October 02, 2002 at 18:04:57, Dann Corbit wrote: > >>On October 02, 2002 at 17:33:39, Mark Young wrote: >> >>>On October 02, 2002 at 15:39:34, Dann Corbit wrote: >>> >>>>On October 02, 2002 at 10:00:45, Paul Doire wrote: >>>> >>>>> >>>>> >>>>>5 Quick 4 minute games (for what they are worth) >>>> >>>>Therein lies the rub. Similar results were had with GLC, but it turns out that >>>>GLC (too) is being run over rough-shod, like everyone else. >>>> >>>>>P3 450Mhz, crafty 24mb ram, ruffian 32mb ram, 128MB system, Win 98 >>>>>3,4 and some 5 man TB's, own books played under Winboard 4.2.6 >>>>> >>>>>Crafty_1714 vs Ruffian_101 +3-0=2 >>>> >>>>5 games have little meaning. >>>> >>>>>Does older Crafty have Ruffians number?, >>>> >>>>I doubt it. Unless you are running Crafty on a 4 CPU box. ;-) >>>> >>>>>Anyone else tried this yet? >>>> >>>>Here is a ten game match at G/60 between two titans: >>>> >>>> Program Elo + - Games Score Av.Op. Draws >>>> 1 Ruffian 1.0.1 : 2554 244 142 10 65.0 % 2446 50.0 % >>>> 2 Crafty-19.0 : 2446 142 244 10 35.0 % 2554 50.0 % >>>> >>>>[games to follow] >>>> >>>>Ten games are also totally unconvincing. However, I think we can see from the >>>>pure volume of games that have been trundled under Ruffian's hooves that Ruffian >>>>is a force to be dealt with. >>> >>>What force. I must have been getting bogus copies of Ruffian. It is good, but I >>>would not call it a force... The more I test Ruffian and at longer time >>>controls the worse Ruffian scores. >>> >>>Here is my current test..still running. >>> >>>My Crafty has been out scoring Ruffian in almost all my tests. >>> >>>P4 2.8 Ghz 1 GB Ram. >>> >>>Tiger book used for all programs. >>> >>>Each program has its own copy of the tiger book. Book learning is carried over >>>for each test. >>> >>>Blitz:15'+15" 0 >>> >>> 1 2 3 4 5 6 >>>1 Hiarcs 7.32 ** 1 00 11 1 1 5.0/7 >>>2 Deep Fritz 7 0 ** ½½ ½ ½1 1 4.0/7 >>>3 Chess Tiger 14.0 11 ½½ ** 0 0 ½ 3.5/7 15.00 >>>4 Junior 6.0 00 ½ 1 ** 1 ½½ 3.5/7 10.50 >>>5 Crafty 18.15 0 ½0 1 0 ** 1½ 3.0/7 >>>6 Ruffian 1.0.1 0 0 ½ ½½ 0½ ** 2.0/7 >> >>Your results seem to be very atypical from what others achieve. >> >>However, with only 7 Ruffian games, this particular result certainly can't be >>called surprising. >> >>I suspect you may have the hash (and possibly the EGTB) set up wrong. Or >>perhaps your tests are using Ruffian as a UCI engine instead of as a Winboard >>engine, where [as UCI] the results do not seem to come out as well. >> >>Ruffian seems to be scoring as well or better than most professional engines in >>all the tests that I have seen. It has a + score against ChessMaster 8000 and >>Chess Tiger 14 on my hardware. The only engine I have seen give Ruffian a >>negative score is Chess Tiger 15. But (of course) the results are preliminary. >>Perhaps we will learn more over time. > >Maybe ruffian does not like the p4 2.8Mh. > >I do not know but the fact that ruffian does good results on your hardware does >not contradict bad results on Mark young hardware. > >The first test to compare is number of nodes per second >If you and Mark young can post the number of nodes per second of Ruffian and >Crafty in anlysis mode in the opening position then we may know better >if my conjecture is correct. That is an interesting idea. I actually have hardware very similar to what Mark has (in addition to Athlon 950 where I do most of my tests) as well as slower platforms. It might be interesting to benchmark.
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.