Yesterday I did my first rating estimate of approximately 3033 Elo when Komodo had scored 1982.5 points out of 2687 games, so I expected that it will finish with (2700/2687)*1982.5 ~ 1992 points. It finally scored 7/13 in the last 13 games instead of 9.5 that I supposed yesterday. Now my rating estimate with my strange interpolation method is 3032 Elo more less (1 Elo less than yesterday), but remember that games versus Fritz are still missing.
I was much faster today in my calculation (I had it recent from yesterday...). I am waiting for know how much wrong I am in my rating estimate.
Ajedrecista wrote:
Yesterday I did my first rating estimate of approximately 3033 Elo when Komodo had scored 1982.5 points out of 2687 games, so I expected that it will finish with (2700/2687)*1982.5 ~ 1992 points. It finally scored 7/13 in the last 13 games instead of 9.5 that I supposed yesterday. Now my rating estimate with my strange interpolation method is 3032 Elo more less (1 Elo less than yesterday), but remember that games versus Fritz are still missing.
I was much faster today in my calculation (I had it recent from yesterday...). I am waiting for know how much wrong I am in my rating estimate.
Ajedrecista wrote:
Yesterday I did my first rating estimate of approximately 3033 Elo when Komodo had scored 1982.5 points out of 2687 games, so I expected that it will finish with (2700/2687)*1982.5 ~ 1992 points. It finally scored 7/13 in the last 13 games instead of 9.5 that I supposed yesterday. Now my rating estimate with my strange interpolation method is 3032 Elo more less (1 Elo less than yesterday), but remember that games versus Fritz are still missing.
I was much faster today in my calculation (I had it recent from yesterday...). I am waiting for know how much wrong I am in my rating estimate.
Thank you very much for your work, Ingo.
Regards from Spain.
Ajedrecista.
8 Elo off, the automatic estimation was better
Bye
Ingo
True, but I had in mind my estimated rating not in the section of full list but in the top-20 (excluding CCT version and replacing it by 5.1). For example, Quazar 0.4 has a difference of 8 Elo depending on which list it is (top-20 or full list); Houdini 3 ratings differ in 5 Elo, etc. A while ago, I did other rating estimate for SF 3: I said 2979 and the official result was 2978 IIRC... later it has dropped a little due to Komodo CCT and Booot 5.2.0.
Again, I repeat that the rating of Protector was 30 Elo higher than its real rating in the live transmission (2867 instead of the real 2837), which undoubtedly made the live rating a bit inflated (1 or 2 Elo).
