Larry wrote:
Hi Nick, hope you're having a nice day!
Ok, I'm convinced the SuperConnie is officially weaker than the Fidelity
'8' level.
There is something I noticed when playing dedicated chess computers
against each other at fast time controls. The operator speed weighs
heavily on the comp's performance. If you play two comps against each
other, and to avoid time in resetting the pieces after the game by only
putting pieces on one board and playing the other comp as a bare board,
the bare board comp gets a distinct advantage. That's because it's much
quicker to press the empty squares than it is to physically press and
move a piece. So the bare board comp gets more thinking time.
When I said the operators can't be trusted, I was'nt suggesting they
were dishonest. I mean some are probably faster and more competent
than others, and have faster/slower internet connections.
I also have to concede that the Wiki ratings are for long time controls,
when in fact the vast majority of games are played quickly, when the
relative strengths are quite different.
all the best
Larry
Hi Larry:
I am sorry but I still do not know how to interpret your first line other than with a hint of sarcasm from you. Of course the CC8 is not stronger and the Elostat stats I gave do not indicate that at all:
57 Chess Challenger 8, Fidelity : 1550 0 0 6 0.0 % 2150 0.0 %
Chess Challenger 8, Fidelity : 6 (+ 0,= 0,- 6), 0.0 %
Star Sapphire, Novag : 1 (+ 0,= 0,- 1), 0.0 %
Atlanta, Mephisto : 1 (+ 0,= 0,- 1), 0.0 %
Sapphire, Novag : 1 (+ 0,= 0,- 1), 0.0 %
Genius 68030, Mephisto : 1 (+ 0,= 0,- 1), 0.0 %
Decorator, Fidelity : 1 (+ 0,= 0,- 1), 0.0 %
Mach III, Fidelity : 1 (+ 0,= 0,- 1), 0.0 %
This tells you that CC8 played 6 games. The first two zeros mean that there are not enough games played otherwise there would be a plus and a minus tolerance. The 0.0 % also tells you that it lost all 6 games. So i do not understand how you can seriously interpret the ELO at all. The ELO is still too high for the CC8 because it played too few games and the average Opponent it played was ELO 2150. So I don't even know why you are even trying to interpret the CC8 ELO differntly to what is provided above. FYI the Wiki is also too high at 1300 for a CC8 at best it would be around 1100 -1200 no more. Wiki only has reasonably accurate ELO's for machines over @ 1500 since Wiki took most of the info from SSDF and SSDF stopped rating machines at around 1550.
The same applies for the Super Constellation which has only 13 games which is still not enough for it to be considered seriously in my example world since its opponents are also rated too high as well. But still Elostat provided a good plus tolerance for the Super Conny based on the game data provided. Tolerance +320 - 41 Av opponent 2099.
58 Super Constellation, Novag : 1540 320 41 13 3.8 % 2099 7.7 %
Super Constellation, Novag : 13 (+ 0,= 1,- 12), 3.8 %
Kasparov GK 2100, Saitek : 1 (+ 0,= 0,- 1), 0.0 %
Mach IV, Fidelity : 1 (+ 0,= 0,- 1), 0.0 %
MM V, Mephisto : 1 (+ 0,= 0,- 1), 0.0 %
R30 V2.5, TASC. : 1 (+ 0,= 0,- 1), 0.0 %
Grandmaster, Excalibur : 1 (+ 0,= 0,- 1), 0.0 %
Risc 1MB, Mephisto : 1 (+ 0,= 0,- 1), 0.0 %
Lyon 68020 22 MHz, Mephisto : 2 (+ 0,= 0,- 2), 0.0 %
ChessMachine 3.1 32 MHz, TASC.: 2 (+ 0,= 1,- 1), 25.0 %
Polgar, Mephisto : 2 (+ 0,= 0,- 2), 0.0 %
Star Sapphire, Novag : 1 (+ 0,= 0,- 1), 0.0 %
As I said before in order to have good ratings you have to have enough games (at least 25-30 against at least 6 different opponents within the machines class for starters)
I know one thing in order to play in an Online World Championship, I am focusing on 1 machine that I know and therefore it is for sure that I have it set for its strongest setting because that is all I am playing with. This is why I trust a 2 player match in a World Championship to give me a best possible result for the machine. Since all these games are available I can review the games and opponents for accuracy and make a good judgment also. I do not have access to games from other lists therefore I do not know if a machine played 200 games against 1 opponent or if it played against 20 so I cannot say if the other lists are more accurate or not. I also do not know if the machine was set at its strongest level or not. Perhaps it is a mixture of Brute force and selective who knows. I do however believe that in 2hrs / 40 SSDF ratings are very good because they have high criterias that they work with.
Also bear in mind that my example has a closed world of only 3 online tournaments played at 30 sec/move. However the accuracy of the results are not in question here (I hope) these games are available for you to play back and check with your machines.
Now if you take Atlanta which by the way was the whole point of my example world to praise it's successes against other top machines who also played lots of games. And was intended to show it's live - online tournament results which also is not in dispute I hope?
Atlanta, Mephisto : 2310 70 68 81 61.7 % 2227 22.2 %
Atlanta, Mephisto : 81 (+ 41,= 18,- 22), 61.7 %
Mach III, Fidelity : 2 (+ 2,= 0,- 0), 100.0 %
Star Sapphire, Novag : 2 (+ 2,= 0,- 0), 100.0 %
Chess Challenger 8, Fidelity : 1 (+ 1,= 0,- 0), 100.0 %
Sapphire, Novag : 3 (+ 0,= 2,- 1), 33.3 %
Genius 68030, Mephisto : 1 (+ 0,= 1,- 0), 50.0 %
Decorator, Fidelity : 1 (+ 1,= 0,- 0), 100.0 %
Super Forte C (7 MHz), Novag : 3 (+ 3,= 0,- 0), 100.0 %
Berlin Pro, Mephisto : 5 (+ 0,= 2,- 3), 20.0 %
London 68030, Mephisto : 1 (+ 1,= 0,- 0), 100.0 %
Lyon 68020 22 MHz, Mephisto : 7 (+ 2,= 4,- 1), 57.1 %
Kasparov GK 2100, Saitek : 1 (+ 1,= 0,- 0), 100.0 %
Turboking II, Saitek : 1 (+ 1,= 0,- 0), 100.0 %
Elite 68060 V11, Fidelity : 11 (+ 3,= 3,- 5), 40.9 %
Mach IV, Fidelity : 4 (+ 4,= 0,- 0), 100.0 %
R30 V2.2, TASC. : 10 (+ 5,= 2,- 3), 60.0 %
Resurrection Ruffian,Phoenix : 8 (+ 2,= 2,- 4), 37.5 %
Brute Force, Saitek : 2 (+ 2,= 0,- 0), 100.0 %
Milano, Mephisto : 2 (+ 2,= 0,- 0), 100.0 %
Resurrection Deep Sjeng,Phoenix: 2 (+ 1,= 0,- 1), 50.0 %
Milano Pro, Mephisto : 8 (+ 4,= 2,- 2), 62.5 %
RISC 2500, Saitek : 6 (+ 4,= 0,- 2), 66.7 %
The 2310 ELO admittedly is higher than Schachinfo which shows it at 2249 after 439 games Atlanta/Magellan.
But with a tolerance of +70/-68 it still looks to be within acceptable limits and the above opponents from the 3 online WM's also speak for themselves.
All the best