CCRL engines with more than 500 games

[Dirt]

I stopped at 2465 and it seems that something is wrong at the rating of weak programs because I cannot believe that they are stronger at long time control.

Wouldn't you expect the increased number of draws at long time control to reduce the rating differences between engines, and thus make the weak ones score relatively better?

[Uri Blass]

I stopped at 2465 and it seems that something is wrong at the rating of weak programs because I cannot believe that they are stronger at long time control.

Wouldn't you expect the increased number of draws at long time control to reduce the rating differences between engines, and thus make the weak ones score relatively better?

I expect the worse programs to perform worse at longer time control because I believe that they tend to have worse order of moves that cause them to earn less from time.

Uri

[Uri Blass]

Thinking about it I think that it may be interesting to calculate rating list when you ignore games when there is a big difference in rating

I see the following results for Hermann 2.0

Naum 2.1 32-bit 2834 +12
−12 (+361) 1 − 31
(+0−30=2) 3.1%
1.0 / 32 0.0% −167 52.0%
1961 1.05
1020
– Spike 1.2 Turin 2826 +7
−7 (+353) 2 − 30
(+0−28=4) 6.3%
2.0 / 32 0.0% −59

I doubt if these type of results are productive for more accurate rating
They may be even counterproductive because a strong program that has some type of bug may lose rating because of playing against weak opponents and the weak opponents that are lucky to play against that program may earn rating.

I suggest to try the following for calculation of rating.

1)delete the match between programs with the biggest difference and calculate the rating again.
2)repeat step 1 and stop only when there is no match when the difference between programs is more than 200 elo(you can also use a different number).

I think that it may be interesting to see if the weak programs tend to have higher rating at long time control after these changes.

Uri

[Graham Banks]

I tried the following link that has reference list from 09.11.2007 and I see no number for Movei 00.8.438

Maybe the reason is that Movei is written as Movei 0.08.438 in the 40/40 list

Uri - what would you like us to name it?

[Uri Blass]

I suggest 00.8.438 that is more consistent with the name of the exe
(00_8_438)

[Graham Banks]

I suggest 00.8.438 that is more consistent with the name of the exe
(00_8_438)

We'll change it to 00.8.438 in all of our lists.

Thanks, Graham.