Hi Theo,
When I wrote that I think I added a
<----------- which for me is acting tongue in cheek or being naughty. Meaning you don't have to read into it with any seriousness.
I really do not know what the end ratings are going to end up being until after the 16 games are played since all these games are showing different situations in chess. There will be some end game games and defensive game, even some closed games.
So computers will do well in some games probably bad in others. After 16 games we will know better if the rating is ok or not. But for a reference it will be ok even now because for example and I am not really that mathematically minded, but if you can measure two distances accurately (ie. two stars positions in the sky) and use them as a bearing or point of reference, then you should be able to very accurately calculate the distance to a third star in the sky.
In my test I am using this same logic in a way. 1st accurate reference is doing exactly 21, 22 or 23 etc ply deep. evaluations with a top engine. The top engine in its way has provided that accurate measurement and not me I am just trusting its results. 2nd accurate measurement is the Grandmaster, we know his ability range because it is recorded so using Nick's logic
(tongue in cheek) If the GM rates for example in my test 2640 ELO for that game and this looks within reason to what his highest rating ever recorded was and within reason of his opponent, then I have my second point of reference. I also have a 3rd point of reference because the same engine Tested is also evaluated at 30 seconds per move and shows a rating ie 2800 ELO. So I think it is reasonable to speculate that if the dedicated chess computer performed at say 2300 ELO, that the computer in this game was either 500 ELO below the Engine standard of 2800, 700 below the test max of 3000 and 340 ELO below the GM standard played in this game. If I do this 16 times with games that I am not picking and just taking out of the book, I should have a fairly unbiased end score average to compare and then maybe fine tune with a different rating idea.
Now again for the tongue in cheek bit.
MFR rating for Jade 2 can be argued at either 2294 ELO or 2320
TC2100 speaks for itself
RS2200 speaks for itself
RS2250 speaks for itself
CC9 was 1721
I don't know but I am guessing that MK12 Trainer was probably rated at 1650 or 1600
The rest I don't know without some further digging. So a Test to you or anyone else. So far which ratings are closer to the manufacturer claims!!
1) Schachcomputer.info Active List
2) Schachcomputer.info Tournament List
3) Selective Search List
4) SSDF
5) Nick's tongue in check Test !!
Below is the 5 game average repeated again for the above test purpose!
1 Critter 1.6a 64 Bit - AMD Phenom 2 Core 2.8GHZ 373.7 93.43% 2803
2 Grandmaster Performance Standard 339.6 84.90% 2547
3 Tasc CM 512K – 15 MHZ – KING 2.54 301.1 75.28% 2258
4 Saitek Travel Champion 2100 299.4 74.25% 2246
5 Mephisto TM Vancouver 68030 36 MHz 295.9 73.98% 2219
6 Nova Jade 2 294.2 73.55% 2207
7 Radioshack 2250XL Brute Force 289.6 72.40% 2172
8 Radioshack 2250XL Selective 288.5 72.13% 2164
9 MChess Pro 5 - P75 278.9 69.73% 2092
10 Saitek Corona 277.4 69.35% 2081
11 CXG 3000 249.8 62.45% 1874
12 Fidelity Sensory 9 231.8 57.95% 1739
13 Saitek MK 12 Trainer LV 5 90S/Move 219.5 54.88% 1646
14 Saitek MK 12 Trainer LV 4 15S/Move 204.5 51.13% 1534
15 Novag Constellation JR 203.4 50.85% 1526
Best regards
Nick