New SSDF rating list
Andreas Mader
I really LOVE the SSDF list!
There's always something to discuss! :-)
Best wishes
Andreas (supporter of the SSDF list)
And now I'm getting ready to be flamed (again)....
THE SSDF RATING LIST 1997-04-26 54875 games played by 170
computers
Rating +- Games Won Oppo
------ --- ----- --- ----
1 Hiarcs 6.0 Pentium 90 MHz 2465 45 265 63% 2369
2 Rebel 8.0 Pentium 90 MHz 2464 26 813 67% 2341
3 MChess Pro 6.0 Pentium 90 MHz 2438 25 796 61% 2363
4 Hiarcs 5.0 Pentium 90 MHz 2435 42 306 63% 2343
5 Genius 5.0 Pentium 90 MHz 2424 27 689 57% 2376
6 Rebel 6.0 Pentium 90 MHz 2413 31 540 60% 2341
7 Genius 4.0 Pentium 90 MHz 2412 25 809 64% 2313
8 MChess Pro 5.0 Pentium 90 MHz 2411 26 759 62% 2322
8 Genius 3.0 Pentium 90 MHz 2411 27 706 62% 2328
10 Rebel 7.0 Pentium 90 MHz 2407 26 725 60% 2334
11 Hiarcs 4.0 Pentium 90 MHz 2397 26 726 55% 2359
12 Chessmaster 5000 Pentium 90 MHz 2387 49 240 67% 2262
12 Nimzo 3.0 Pentium 90 MHz 2387 26 755 58% 2331
14 Nimzo 3.5 Pentium 90 MHz 2386 30 540 53% 2367
15 Hiarcs 3.0 Pentium 90 MHz 2380 30 545 56% 2335
16 Genius 4.0 486/50-66 MHz 2378 27 671 56% 2335
17 R30 v. 2.5 2373 45 281 66% 2253
18 Junior 3.3-3.5 Pentium 90 MHz 2372 33 436 49% 2376
19 MChess Pro 4.0 Pentium 90 MHz 2361 29 578 52% 2345
20 Genius 3.0 486/50-66 MHz 2359 24 910 62% 2272
21 Fritz 3.0 Pentium 90 MHz 2351 25 775 52% 2338
22 Fritz 4.0 Pentium 90 MHz 2337 40 324 60% 2267
22 Kallisto 1.98 Pentium 90 MHz 2337 27 657 49% 2346
24 MChess Pro 5.0 486/50-66 MHz 2336 28 625 52% 2323
25 Rebel 7.0 486/50-66 MHz 2335 26 697 51% 2330
26 WChess 1.06 Pentium 90 MHz 2333 26 738 47% 2355
27 MChess Pro 4.0 486/50-66 MHz 2314 24 846 53% 2290
28 WChess 1.03 486/50-66 MHz 2303 27 677 49% 2309
29 Hiarcs 3.0 486/50-66 MHz 2301 24 865 50% 2298
29 Rebel 6.0 486/50-66 MHz 2301 24 884 55% 2266
31 Meph Genius 68 030 33 MHz 2299 47 228 57% 2252
32 Ch.Machine 30 MHz King 2.0 aggr/R30 off 2295 21 1182 65% 2190
33 Chessmaster 4000 486/50-66 MHz 2292 34 462 66% 2172
34 Rebel Decade Pentium 90 MHz 2281 30 563 45% 2319
35 Fritz 3.0 486/50-66 MHz 2257 22 1040 50% 2254
36 Berlin Pro 68 020 24 MHz 2223 25 816 58% 2163
37 Comet32 Pentium 90 MHz 2221 32 528 32% 2350
38 Kallisto 1.82-1.83 486/50-66 MHz 2216 22 1046 42% 2273
39 Kasparov SPARC 20 MHz 2215 25 769 51% 2207
40 Saitek RISC 2500 ARM2 14 MHz 128K 2193 22 1062 56% 2148
41 Mephisto Montreux ARM 14 MHz 512K 2187 32 589 73% 2015
42 Socrates 3.0 486/33 MHz 2141 50 203 47% 2162
43 Novag Sapphire H8 10 MHz 2086 23 945 49% 2097
44 Milano Pro SH7000 20 MHz 2075 72 111 64% 1974
45 Gandalf 2.1 Pentium 90 MHz 2042 52 230 27% 2215
46 Nimzo 2.2.1 486/33 MHz 2034 47 229 42% 2088
47 Zarkov 3.0 486/25-33 MHz 2029 48 232 39% 2108
48 Kasparov Brute Force H8 10 MHz 2018 23 906 44% 2058
49 Complete Chess System 486/33 MHz 1984 47 221 47% 2007
50 Kasparov President/GK-2100 H8 10 MHz 1976 30 558 47% 1998
Best wishes
Andreas
Herbert Groot Jebbink
>I really LOVE the SSDF list!
>There's always something to discuss! :-)
.....
> 1 Hiarcs 6.0 Pentium 90 MHz 2465 45 265 63% 2369
> 2 Rebel 8.0 Pentium 90 MHz 2464 26 813 67% 2341
oohhhh, 1 point
Greetings, Herbert
---
The Trans-Siberian Railroad Page, http://www.xs4all.nl/~hgj/
Ed Schroder
From: h...@xs4all.nl (Herbert Groot Jebbink)
: 1 Hiarcs 6.0 Pentium 90 MHz 2465 45 265 63% 2369
: 2 Rebel 8.0 Pentium 90 MHz 2464 26 813 67% 2341
: oohhhh, 1 point
Congratulations to Mark Uniacke and Eric Hallsworth from the Hiarcs team.
I will be back, grrrrrr... :)))
- Ed Schroder -
: Greetings, Herbert
Hello
Congrats to Ed for being so gracious, and to the Hiarcs team doing so
well...
However, these results highlight the problems inherent in rankings...Rebel
has played over 500 more games than Hiarcs, but Hiarcs gets the number 1
ranking by 1 point. Obviously, 1 point is statistically meaningless, so the
question is: "Does Hiarcs (or Rebel, for that matter) really deserve the
number 1 ranking?" Perhaps a better presentation would be to publish these
results with a confidence interval as the last column. It wouldn't have to
be a 90% confidence interval, which might make the interval so wide as to
become meaningless, just one of known value. A 50% confidence interval
might be nice, since then we could say, "There's is a 50-50 chance that
Hiarcs is better than CM5000", or something like this. Further, all of the
programs within the confidence interval of the highest rated program could
then be grouped together, and share the position as TOP TIER
(notice I didn't say ranked) programs, since there would be no firm basis
on which to say that the skill of one really differs from another.
Roger Davis
Ed Schroder wrote in article <5k1uuq$kgm$2...@thor.wirehub.nl>...
Komputer Korner
Andreas Mader wrote:
snipped
> THE SSDF RATING LIST 1997-04-26 54875 games played by 170
> computers
> Rating +- Games Won Oppo
> ------ --- ----- --- ----
> 2 Rebel 8.0 Pentium 90 MHz 2464 26 813 67% 2341
> 4 Hiarcs 5.0 Pentium 90 MHz 2435 42 306 63% 2343
> 5 Genius 5.0 Pentium 90 MHz 2424 27 689 57% 2376
> 6 Rebel 6.0 Pentium 90 MHz 2413 31 540 60% 2341
> 7 Genius 4.0 Pentium 90 MHz 2412 25 809 64% 2313
> 8 MChess Pro 5.0 Pentium 90 MHz 2411 26 759 62% 2322
> 8 Genius 3.0 Pentium 90 MHz 2411 27 706 62% 2328
> 10 Rebel 7.0 Pentium 90 MHz 2407 26 725 60% 2334
> Andreas
Don't forget that the SSDF rating list is only computer vs computer.
A lot of people put stock into this as the be all and end all of
strength test. I put more stock into human vs computer. The only
problem is it takes lot of money to entice an IM to risk his
reputation. I was fortunate that IM Deen Hergott agreed to play for
only
$1000 if he won and only $500 if he lost . Just as some people have
criticized me for calling Hiarcs a monster, I criticize all those who
slavor
over the latest SSDF list. The commercial programmers are forced to
recognize and to program towards the SSDF list because so many people
treat it as the only thing that counts. It is only one piece of the
strength puzzle. Let us not forget this. Aegon was another piece.
--
Komputer Korner
The inkompetent komputer.
Andreas Mader
Hello wrote:
>
> Congrats to Ed for being so gracious, and to the Hiarcs team doing so
> well...
>
> However, these results highlight the problems inherent in rankings...Rebel
> has played over 500 more games than Hiarcs, but Hiarcs gets the number 1
> ranking by 1 point. Obviously, 1 point is statistically meaningless, so the
> question is: "Does Hiarcs (or Rebel, for that matter) really deserve the
> number 1 ranking?"
This is an interesting point. Has anyone ever questioned Kasparows
Number 1 ranking just because he played less games than Karpov? It is
true that the rating becomes more accurate the more games are played,
but it could become higher, too. I think there was never a number 1 in
the SSDF rating list who didn't "deserve" it!
> Perhaps a better presentation would be to publish these
> results with a confidence interval as the last column. It wouldn't have to
> be a 90% confidence interval, which might make the interval so wide as to
> become meaningless, just one of known value.
The confidence is 95%, not 90%. If there is a error probability of 5% or
less then you can statistically(!) speak of "not relevant".
> A 50% confidence interval
> might be nice, since then we could say, "There's is a 50-50 chance that
> Hiarcs is better than CM5000", or something like this.
A 50% confidence interval is sure very narrow but also says nothing.
There is a 50% chance that the rating is in the given interval and a 50%
chance that the rating is outside the given interval. It says NOTHING
about "X is better than Y", because you do not know if an "outside"
rating is higher or lower than the given interval.
Best wishes
Andreas
Jouni Uski
Please note, that match Hiarcs6 - Mchess6 is missing. The
probable big win by Hiarcs will raise it's rating by about
20 points!
Jouni
Serge Desmarais
Herbert Groot Jebbink wrote:
>
> On Mon, 28 Apr 1997 07:54:14 -0700, Andreas Mader
> <ma...@p6.gud.siemens.co.at> wrote:
>
> >I really LOVE the SSDF list!
> >There's always something to discuss! :-)
> .....
> > 2 Rebel 8.0 Pentium 90 MHz 2464 26 813 67% 2341
>
Hehehehe. In fact I NEVER thought HIARCS would be ABOVE Rebel 8! I
thought it would be #2!
Serge