I have analyzed the 12 games between Kasparov and Deep Blue played in the two matches in 1996 and 1997.
Arena 3.0 and Rybka4 was used, and the time spent was 30 seconds pr move.
Is this powerful enough to make a judgement of the games?
Well....
Philadelphia 1996 - average performance rating:
=================================
Kasparov = 2661
Deep Blue = 2713
Although Kasparov won the match 4-2, Deep Blue has a slightly better performance rating.
This is mainly due to game 1 with a huge difference between these scores:
Kasparov = 2168, Deep Blue = 3123.
New York 1997 - average performance rating:
===============================
Kasparov = 2658 *)
Deep Blue = 2837
Deep Blue won the match 3.5-2.5.
Kasparov got the same score (-3) as the year before.
But Deep Blue improved with 124.
It is something like that I had expected...
*) But we must also remember, that Kasparov made a very bad "move" in game 2 - he resigned in a drawn position.
I can simulate that by adding a bad move 45...Qa7 to the game, and assign it a score of 20 (which is used for mates).
Code: Select all
----- Error -------- Score IF This line was played
44. Kf1 ooooo. 0.63 2.20 Kg1-h1 Rb7-b8 Ra8-a6 Qb6xc6 d5xc6 Rb8-e8 Ra6-a5
44. .. Rb8 x . 0.07 1.57 Rb7-b8 Qc6-d7+ Kf7-g8 Ra8-a7 Bd6-f8 Qd7-f7+ Kg8-
45. Ra6 ooooooo 1.74 1.64 Qc6-d7+ Kf7-g8 Ra8-a7 Bd6-f8 Qd7-f7+ Kg8-h8 Kf1-
.....Qa7 xxxxxxxxxxxxxxx 20.10 -0.10 Qe3
20.00
White Black
Total error 4.33 24.09
Evaluated moves 45 45
Error/move 0.096 0.535
Performancerating 2758 1026
But his average of 2658 is certainly too high if the end of this game is considered.
See the additional posts:
Analyzed games (some are modified a bit)
Original games (the original versions of the modified games)
How the performance rating is calculated: http://74.220.23.57/forum/viewtopic.php ... ht=dabbaba