Comparision between IPON and CEGT/CCRL

[IWB]

Hi

All single Engines nomalized to 2800 for S12 32bit. Data taken today, 2010.02.07.

It is interesting that CEGT/CCRL are rating some engines much higer than IPON. The difference is very often much bigger than the 95% margin and much more often than on 1 out of 20 engines.




Bye
Ingo

PS: Some results are entered manualy, I appologize if I made a mistake. In general it should be fine.

PPS: Personal remark (my opinion!) The CCRL single data is a mess, with a lot of holes and too few games! It would be better to leave some engines completly out than to publish these results. I left some results out or the Delta would be even bigger!!

[Kirill Kryukov]

Hi

All single Engines nomalized to 2800 for S12 32bit. Data taken today, 2010.02.07.

It is interesting that CEGT/CCRL are rating some engines much higer than IPON. The difference is very often much bigger than the 95% margin and much more often than on 1 out of 20 engines.

Hi Ingo,

Before this comparison begins to make any sense you need to normalize using multiple engines. Using just S12 32-bit is very bad, as can be seen in your "Difference" column: most of the numbers are positive. "Average difference to 0" is 15.9 and 28.25 - this is how much S12 32-bit rating is varying in one list from another.

I suggest to use weighted average of all engines you are comparing (that are present in all lists).

Best,
Kirill

[IWB]

Yea, I see that there is a problem!

But it is a problem to do that with all three lists. To CEGT I should use something around +16 to CCRL close to 28. Average 22 ... all not perfect.

I have to ponder about this

Thx for the tip
Ingo