My chess site ...

[Marc Lacrosse]

... has been rejuvenated and I intend to add content more regularly.
Address is still http://chessbazaar.mlweb.info.
I hope computer chess fans will find some interesting info there.
There is a linked blog where comments and suggestions are welcome.

Marc

[Marc Lacrosse]

No dot at the end of the URL : should be:

http://chessbazaar.mlweb.info

[Kirill Kryukov]

Hi Marc,

Thanks for posting link to your web-site. Pages about rating estimation are particularly interesting. Do you have ideas to further improve the accuracy of fast rating estimation?

EDIT. Also, I'd be very curious to see the results of your fast rating estimation methods on weaker engines, say, down to 2000 Elo points.

Best,
Kirill

[Marc Lacrosse]

Hi Marc,

Thanks for posting link to your web-site. Pages about rating estimation are particularly interesting.

Thanks

Do you have ideas to further improve the accuracy of fast rating estimation?

As I am member of an engine "team", I am always trying to find better and faster methods for comparing two versions of the same engine.

Recently I have had a look at ultrafast games (entire game in a few seconds) but results are a bit incoherent. This approach seems better when the difference between versions relates to evaluation, and less good when the difference is mostly on search. What I am quite sure is that ultrafast testing may be good for comparing two versions of the same engine against a common set of opponents but wil not be good for establishing a complete rating scale.

The testing scheme based on fast evaluation of a large set of positions is interesting and I suppose some improvement could come from a better selection of a larger number of positions.
But I doubt it could be improved to the point where a full-scale rating list without aberrant results could be built. Here again this approach could well be more interesting for comparing two closely related versions of the same engine without trying to estimate their precise position in the full population of existing engines.

EDIT. Also, I'd be very curious to see the results of your fast rating estimation methods on weaker engines, say, down to 2000 Elo points.
Best,
Kirill

Unfortunately I have not enough free time for this; in my YAFTS attempt I already covered a rather large elo span from 2590 to 3225 but I don't intend to extend this downside myself. Moreover the engine I am interested to help contribute for is at the higher end of the list.
Anybody with a little personal and computing free time can do the job with less strong engines and I would surely be interested in seeing other's results with similar methodologies...

Thanks for your comments and for your impressive contribution to the community with the CCRL!

Marc

[Kirill Kryukov]

Hi Marc,

Thanks for posting link to your web-site. Pages about rating estimation are particularly interesting.

Thanks

Do you have ideas to further improve the accuracy of fast rating estimation?

As I am member of an engine "team", I am always trying to find better and faster methods for comparing two versions of the same engine.

Recently I have had a look at ultrafast games (entire game in a few seconds) but results are a bit incoherent. This approach seems better when the difference between versions relates to evaluation, and less good when the difference is mostly on search. What I am quite sure is that ultrafast testing may be good for comparing two versions of the same engine against a common set of opponents but wil not be good for establishing a complete rating scale.

The testing scheme based on fast evaluation of a large set of positions is interesting and I suppose some improvement could come from a better selection of a larger number of positions.
But I doubt it could be improved to the point where a full-scale rating list without aberrant results could be built. Here again this approach could well be more interesting for comparing two closely related versions of the same engine without trying to estimate their precise position in the full population of existing engines.

EDIT. Also, I'd be very curious to see the results of your fast rating estimation methods on weaker engines, say, down to 2000 Elo points.
Best,
Kirill

Unfortunately I have not enough free time for this; in my YAFTS attempt I already covered a rather large elo span from 2590 to 3225 but I don't intend to extend this downside myself. Moreover the engine I am interested to help contribute for is at the higher end of the list.
Anybody with a little personal and computing free time can do the job with less strong engines and I would surely be interested in seeing other's results with similar methodologies...

Thank you for your reply. It's understandable that you are interested in particular Elo range. Unfortunately I also have no time to do this.

Thanks for your comments and for your impressive contribution to the community with the CCRL!

Marc

Always welcome!

Best,
Kirill

[Eelco de Groot]

... has been rejuvenated and I intend to add content more regularly.
Address is still http://chessbazaar.mlweb.info.
I hope computer chess fans will find some interesting info there.
There is a linked blog where comments and suggestions are welcome.

Marc

Nice site you built Marc! So many sites have disappeared, it is good to see also some important sites remaining. I did not know the link you posted Marc, I had in Favourites http://users.skynet.be/mlcc/chessbazaar/ that is also still around, I hope you don't take it down now that I mention it...
We can never have too many computerchess-sites even if no longer active...

Still missing Bruce Moreland's old site, although I have copies now of his computer chessprogramming pages on my harddisk and I think copies are on the Wiki as well.

Regards, Eelco