ELO progression measured by year

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Rebel
Posts: 4663
Joined: Thu Aug 18, 2011 10:04 am

ELO progression measured by year

Post by Rebel » Wed Dec 13, 2017 4:11 pm

ELO progression measured by year based on CCRL 40/20 games which were played that year. ELO numbers are based on the average rating of the 100 best engines and splitted into a (S)ingle (P)rocessor and (M)ulti (P)rocessor list, follow the orange links to view the details. Furthermore the statistic of the opening lines shows the quality of the randomness.

...

http://rebel13.nl/misc/stats.html

duncan
Posts: 10206
Joined: Mon Jul 07, 2008 8:50 pm

Re: ELO progression measured by year

Post by duncan » Wed Dec 13, 2017 6:58 pm

Rebel wrote:ELO progression measured by year based on CCRL 40/20 games which were played that year. ELO numbers are based on the average rating of the 100 best engines and splitted into a (S)ingle (P)rocessor and (M)ulti (P)rocessor list, follow the orange links to view the details. Furthermore the statistic of the opening lines shows the quality of the randomness.

...

http://rebel13.nl/misc/stats.html
very useful. thanks.

were there hardware changes during those years?

mjlef
Posts: 1424
Joined: Thu Mar 30, 2006 12:08 pm
Contact:

Re: ELO progression measured by year

Post by mjlef » Wed Dec 13, 2017 7:51 pm

Do you have the same thing for the top 1 to 3 engines? I think averaging over 100 engines is going to make the progress look much slower than it was, at least for the stronger programs.

User avatar
Rebel
Posts: 4663
Joined: Thu Aug 18, 2011 10:04 am

Re: ELO progression measured by year

Post by Rebel » Thu Dec 14, 2017 10:01 am

duncan wrote:
Rebel wrote:ELO progression measured by year based on CCRL 40/20 games which were played that year. ELO numbers are based on the average rating of the 100 best engines and splitted into a (S)ingle (P)rocessor and (M)ulti (P)rocessor list, follow the orange links to view the details. Furthermore the statistic of the opening lines shows the quality of the randomness.

...

http://rebel13.nl/misc/stats.html
very useful. thanks.

were there hardware changes during those years?
That's a good question that only can be answered by the CCRL folks. If they provide information I could add an extra column to the table.

User avatar
Rebel
Posts: 4663
Joined: Thu Aug 18, 2011 10:04 am

Re: ELO progression measured by year

Post by Rebel » Thu Dec 14, 2017 10:05 am

mjlef wrote:Do you have the same thing for the top 1 to 3 engines? I think averaging over 100 engines is going to make the progress look much slower than it was, at least for the stronger programs.
I made a start doing so, it's only for the years 2006-2008 currently, it's a lot of work. But interesting enough to finish it. Also the progress for the best engine will be measured.

http://rebel13.nl/misc/stats.html

User avatar
Ozymandias
Posts: 1104
Joined: Sun Oct 25, 2009 12:30 am

Re: ELO progression measured by year

Post by Ozymandias » Thu Dec 14, 2017 10:19 am

Rebel wrote:That's a good question that only can be answered by the CCRL folks. If they provide information I could add an extra column to the table.
They adapt the TC to even out HW speed, so what machine each tester is using, becomes irrelevant.

duncan
Posts: 10206
Joined: Mon Jul 07, 2008 8:50 pm

Re: ELO progression measured by year

Post by duncan » Thu Dec 14, 2017 10:58 am

Rebel wrote:ELO progression measured by year based on CCRL 40/20 games which were played that year. ELO numbers are based on the average rating of the 100 best engines and splitted into a (S)ingle (P)rocessor and (M)ulti (P)rocessor list, follow the orange links to view the details. Furthermore the statistic of the opening lines shows the quality of the randomness.

...

http://rebel13.nl/misc/stats.html
the increase for best program for the period

single
2921 - Rybka 2.2 64-bit
3352 - asmFish 051117 64-bit

431 increase

multiple
3002 - Rybka 2.1 64-bit 4CPU
3426 - asmFish 051117 64-bit 4CPU

424 increase


for top 3

single

2921 - Rybka 2.2 64-bit
2779 - Fritz 10
2776 - Hiarcs 11

average 2825


3352 - asmFish 051117 64-bit
3341 - Houdini 6 64-bit
3327 - SugaR XPrO 1.3 64-bit

average 3340

increase 515

multiple

3002 - Rybka 2.1 64-bit 4CPU
2864 - Hiarcs 11 4CPU
2855 - Rybka 1.1 32-bit

average 2907


3426 - asmFish 051117 64-bit 4CPU
3414 - SugaR XPrO 1.2 64-bit 4CPU
3411 - Houdini 6 64-bit 4CPU

average 3417

increase 510

User avatar
Rebel
Posts: 4663
Joined: Thu Aug 18, 2011 10:04 am

Re: ELO progression measured by year

Post by Rebel » Thu Dec 14, 2017 11:50 am

Ozymandias wrote:
Rebel wrote:That's a good question that only can be answered by the CCRL folks. If they provide information I could add an extra column to the table.
They adapt the TC to even out HW speed, so what machine each tester is using, becomes irrelevant.
I see, so basically they are running a 40/40 list based on the hardware of 2006?

User avatar
Ozymandias
Posts: 1104
Joined: Sun Oct 25, 2009 12:30 am

Re: ELO progression measured by year

Post by Ozymandias » Thu Dec 14, 2017 2:54 pm

"Time control: Equivalent to 40 moves in 40 minutes on Athlon 64 X2 4600+ (2.4 GHz), about 15 minutes on a modern Intel CPU."

Note it's either "40/40" or "40/4", "40/20" would be CEGT.

User avatar
Rebel
Posts: 4663
Joined: Thu Aug 18, 2011 10:04 am

Re: ELO progression measured by year

Post by Rebel » Thu Dec 14, 2017 4:24 pm


Post Reply