Rating list history

[IanO]

I would like to construct a chart of the top ten engine ratings over time, but need access to rating list history. Do any of the rating lists maintain historical data? Monthly or quarterly resolution should be sufficient. SSDF keeps old lists, but they have not kept up with the pace of change in new engines recently.

[JuLieN]

SSDF list through time:
http://wayback.archive.org/web/*/http://ssdf.bosjo.net/

CCRL 40/40 list through time:
http://wayback.archive.org/web/*/http:/ ... t_all.html

CCRL 40/4 list through time:
http://wayback.archive.org/web/*/http:/ ... /ccrl/404/

Blue dots mean updates. Just click on them. Hope that helps

EDIT: you'll have to copy/paste the URLs by hand, as Talkchess cuts them when there's a "*"...

[Adam Hair]

Thanks, Julien. I never knew that the waybackmachine indexed its snapshots.

[jplchess]

I hope this helps; this is SSDF from 1996 until 2000.

www.reocities.com/capecanaveral/​ ... /chess.htm

Jonathan Lee

[jplchess]

http://www.reocities.com/capecanaveral/ ... f/ssdf.htm

Jonathan Lee

[IanO]

Thanks! I used the wayback machine to get approximately quarterly history for the CCRL 40/4 list and compiled this graph using gnuplot. (Big chunks were missing in 2009 and 2011.)



This incorporates some name changes over time: Glaurung to Stockfish, Bright to Spark, and Doch to Komodo. The jump on the most recent datapoint is from testing on 6CPU machines (about +40 from 4CPU machines).

I was surprised to see so little rating drift. Programs flatline when the same version is tested for a long time, and only jump when a new version is released. I'm not sure if this relates to the CCRL testing methodology or rating system used.

[JuLieN]

Thanks for this patient work, Ian, that's very interesting.

Now to answer your question I think it is a granularity problem related to how new versions of engines are made. For instance between, Pred 0.1.5 and Pred 2.0 I produced like 200 beta versions, during a year and a half, each worth a few Elo points better only. As a result pred 2.0 was about 400 points Elo stronger, and on your graphics you would see a big jump when actually it was a very smooth curve if you had access to the betas versions. I think it's the same thing for most of those engines.

My 2 cents.

[Graham Banks]

Thanks! I used the wayback machine to get approximately quarterly history for the CCRL 40/4 list and compiled this graph using gnuplot. (Big chunks were missing in 2009 and 2011.)

We can give you a complete day by day history if you'd like.