Author: Djordje Vidanovic
Date: 18:00:42 02/27/04
Go up one level in this thread
On February 27, 2004 at 19:16:10, Eduard Nemeth wrote: >On February 27, 2004 at 17:25:15, Djordje Vidanovic wrote: > >>On February 27, 2004 at 15:19:39, Dann Corbit wrote: >> >>>Started a little contest at G/10 on AMD 950 MHz, so far this is the output: >>> >>> Program Elo + - Games Score Av.Op. Draws >>> 1 List512 : 2629 296 207 8 68.8 % 2492 37.5 % >>> 2 Ruffian_105 : 2611 360 201 7 64.3 % 2509 42.9 % >>> 3 Rebel12_CB : 2601 378 251 7 57.1 % 2551 28.6 % >>> 4 Ruffian_210 : 2595 310 249 8 62.5 % 2506 25.0 % >>> 5 Ruffian_101 : 2568 310 387 8 62.5 % 2480 0.0 % >>> 6 Aristarch 4.41: 2541 325 292 8 56.2 % 2498 12.5 % >>> 7 Delfi-440 : 2523 325 292 8 56.2 % 2479 12.5 % >>> 8 Gothmog : 2500 251 378 7 42.9 % 2550 28.6 % >>> 9 Ruffian_202 : 2485 343 343 7 50.0 % 2485 14.3 % >>>10 Glc300 : 2466 343 343 7 50.0 % 2466 14.3 % >>>11 DeepSjeng : 2406 207 296 8 31.2 % 2543 37.5 % >>>12 Quark-232-net : 2399 249 310 8 37.5 % 2488 25.0 % >>>13 ELChinito 3.25: 2392 333 360 7 35.7 % 2494 14.3 % >>>14 Crafty-1910 : 2290 270 284 8 25.0 % 2481 25.0 % >>> >>>Initial indications are that all of the Ruffians are about the same strength. >>>But more data may change that appearance significantly. >> >> >>In case you wonder, I think that I might be of some help. You listed almost all >>the available versions up there, almost all but one: Ruffian Leiden 2003 BTW, >>the version that won the Dutch Open 2003. >>And that one is the strongest of all, at least in my testing: more than 5000 >>games at different although fast time controls (longest was G/45, played on two >>fast machines: P4-2.4 and Athlon 2800+ Tbred). >>Of course, I am aware of some other credible data (cf. Sarah Bird, for example) >>that point in a different direction. However, none of the testers has done so >>much testing with the Leiden version which is simply neglected for reasons >>unknown to me. Even the mammoth effort of Leo Dijksman, a person of absolute >>integrity, has somehow "missed" to test the Leiden version. Most probably due >>to the _unnecessary_ and very much _premature_ hype about Ruffian 2.0.0 (almost >>untested by me, at least). >> >>I am as sure as one can reasonably be that the Leiden version of Ruffian would >>perform about 30-40 Elo better than _any_ other version of Ruffian in SSDF >>testing and that it should place about as high as CM 9000. >> >>Djordje > >Hello Djordje! > >Can it be the case, that Ruffian 2.0.0 plays not good with the Leiden BOOK? > >I seems that this book is optimized for Ruffian _leiden but not for Ruffian 2.0 > >Or? > >Best, >Eduard Hello Eudard, this was a very intelligent question. Yes, I believe that you may be quite right. I was trying to gear Ruffian (Oct. 10 version - Leiden) to play stuff it felt OK with, such as semi closed and closed positions and it appears that Ruffian 10-10 was at home with that book. However, the later versions were more tactical (faster at tactics) and more aggressive and the book surely had to be changed accordingly. I had no time for that as the new Ruffian 2.0.0 just popped up from the blue and I thought (mistakenly) that the Ruffian Leiden book could still be the best choice. The other books on the Arena CD were all for fun and experimentation against human players. Briefly, the Ruffian Leiden book is surely OK for Ruffian 10-10 (Oct. 10), i.e. the Ruffian Leiden 2003 version on the Arena CD, but I am very suspicious about it as being suitable for the sharper new versions of the Ruffians on the Arena CD. Still I believe that the Arena CD offers more than the Lokasoft product -- mainly due to the fact that the customers get the Leiden version as well. (Saying this, I do not promote the product as I am not commercially involved in the least in either of the Ruffian productions nor have I earned a cent from my work, which was a "labour of love"). That's how things are viewed by the eyes of this beholder. To reply to your additional query below about Per-Ola making a new bug-fix for Ruffian Leiden: I am very sceptical about fiddling with stuff that runs fine most of the time. By doing this and that you usually ruin things... Ask some other programmers or even Vladimir Vuckovic and myself about our own experiences with Axon... The moment we think that we made a successful fix it turns out that the program is doing something really stupid elsewhere. I would leave Ruffian Leiden as it is: a smart and strong program, performing great at tournament time controls and scoring _always_ over 65% against a very well-known program even at blitz. Djordje
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.