Author: Mike Byrne
Date: 18:31:43 02/01/05
Go up one level in this thread
On January 31, 2005 at 20:53:41, Earl Fuller wrote: >Hi everyone, just a little history on Dedicated Chess Computer ratings. >In the middle of the year 1994 Computer Chess Reports (CCR) came out with these >ratings on dedicated chess machines. >CCR MID 1994 40/2 against humans > >1) Tasc R30 King 2.2 with a mean rating of 2498 >2) Mephisto Genius 68030 with a mean rating of 2476 >3) Meph. Risc 2 (1 Meg) with a mean rating of 2471 >4) Mephisto Vanc.(36 MHz) 68030 with a mean rating of 2453 also a (CRA) game > in 30 min. 2491 >5) Meph Lyon 68030 with a mean rating of 2453 >6) M.Berlin Pro 68020 with a mean rating of 2443 > >And no, i do not have any exact numbers as to what these ratings equate to in >respect to the SSDF or ELO, Fide or USCF. Although remember the CCR was the >rating agency of the USCF back in those days. >Anyway lets look at the SSDF today__(2005) > >1)Tasc R30 2.2 is rated at SSDF 2274 and the Mephisto 68030 Genius is rated at >2196 SSDF. Some say you add 150 points for USCF___i don't know___ > >Anyway the reason i've brought this up is because there has been talk of >"perhaps" some new modul coming out that is going to be the strongest ever. >Well lets add 150 points to the Mephisto Genius 68030 and lets see,___ >2196 + 150 = 2346 USCF (perhaps) and the Tasc R30 would be (2424) USCF. >Note, that in Mid. 94 the CRA rated the Mephisto Vanc (68030) 2491 at 40/2 >against human players over the board, with a CCR game in 30 min. rating at 2486. >Interesting____ >earl The are two main drivers that lower SSDF computer ratings as time moves on. First we have the Bloodgood effect. I have written about this many times here a(search "Bloodgood" in the archives), The second problem is that SSDF ratings are predominately computer vs computer witha limited pool of players (other computers) and with new influx of players that are constantly stronger ( computer programs runnings on faster hardware). Saldly, with this combination, there is nothing that can turn the tide. In summary, the Bloodgood effect has an inflationery impact of a rating pool for the strongest players as compared to the strongest players of a much larger pool. The strongest play in a small pool of players will obtain inflated ratings because there is no one good enough to beat them consistently. While in prison for murdering his mother (sounds like a nice guy eh?), Claude Bloodgood became the second hightest rated player in UCSF. A master level player, no other prisoner could lay a glove on him (figuratively) in chess and he became second highest rated player in the USCF (second to Bobby Fischer , if I remember correctly). All of these grand old dedicated computers like the R30 and the 68030 Genius, are basically running 1991/1992 hardware when a 486 66 Mhz computer was considered dynamite. They simply do stand a change against today's 3.2 Ghgz machines. The difference in some cases is 500-1000x or more. From to 2 to 3 Knps to 2 or 2 million NPS. There is no way that a TASC R30 will maintain its ratings against these fast computers -> it has to go down. You combine these effects and the results no longer make sense when you factor in human play. The evidence points to situation where faster and faster hardware will achive measurable improvements against older hardware - no matter how high ratings are. But the evidence also points to less measurable returns against the stongest human players, even when you are doubling the hardware speed time and time again. So my hypthosis is that if one were to play all these "old" top end computers in a human tournamnets, they would achieve something close to where they were once before, e.g TASC R30 might be ~ 2475. Take the current top end programs and play them in the same tournament and they might achive say 2775 USCF. Thus, human component factored-in compresses the rating difference between computers , or said another way, computer vs computer (only) play exaggerates the differences rating differences, similiar to how Bloodgood's USCG rating was exaggerated in his limited prison pool. Just my $.02
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.