Author: Stephen A. Boak
Date: 12:43:44 01/29/00
Go up one level in this thread
Here are the Rebel Rating results (TPR calculation) for 23 games, broken into Color and Hardware Speed categories-- REBEL 40/2 GAMES AND RATING WHITE WHITE BLACK BLACK TOTAL TOTAL QTY RTG QTY RTG QTY RTG > 425MHz 6 2442 8 2594 14 2529 < 425MHz 5 2505 4 2511 9 2508 ALL 11 2471 12 2566 23 2521 Note the low TPR for Rebel in 6 games with White pieces on 'Fast' (over 425 MHz) PCs. But realize that those 6 opponents included Anand, Rohde and Hoffman and some special problems. Anand, of course, is an exceptionally strong player, and he is well-versed against PCs. Rohde won a game where I believe Rebel was doing reasonably well (i.e. not losing) through the opening and into the middle game--until it made some weak moves allowing a successful attack on its king. During subsequent analysis it came to light that Rebel definitely required some improvements to its king-safety algorithms, were quickly made by Ed Schroder. Hoffman won a Benko gambit game in which Rebel crashed numerous times, made at least a couple of non-reproducible moves, and from an overall viewpoint apparently could not fathom the long-term strategic considerations and maneuvering that underlies that gambit opening. I'm not trying to make excuses. But there are just a few White games (6) on Hardware faster than 425MHz, and the next 6 such games might well be much better for Rebel--we will have to watch as the data comes in from additional Rebel 40/2 experiments. Note also the excellent overall Rebel TPR for 12 games with Black--2566! This is some evidence and support for the argument of Rebel GM playing strength at 40/2. If this TPR with Black holds up, it conceivably might not be too long before such results could be boosted to 2600, by either a 50-60% improvement in hardware speed or by programming improvements to minimize weaknesses or search faster. There aren't, however, very many games in that calculation. On my Excel spreadsheet, I keep running TPR averages for Rebel's last 5 and last 10 consecutive games. I graph these, game by game, along with Rebel's game by game and running/current TPR figures. Over the 23 game sequence, in order of date played, Rebel had a running average for the prior 10 games of over 2540 on three occasions. This occured 3 times (approx. 21.4% of the time) out of 14 total such calculated 10-game runs (i.e. 1-10, 2-11, ... 10-19, 11-20, 12-31). If these Rebel 10-game measured TPR streaks rise near to 2600 over some 10-game sequences, we might consider that Rebel would be able to achieve a FIDE-type GM norm, if it played some series of games in a single FIDE-rated tournament of, say, 7, 8, 9 or 10 rounds. At the moment, we might say that Rebel has a 21.4% chance of having a 10-game streak with TPR performance level higher than 2540 FIDE rating. --Steve Boak
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.