Some more ideas for the Banjo chess server

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Some more ideas for the Banjo chess server

Post by sje »

Some more ideas for the Banjo chess server

1) The magically extendable kibitz object

The use of XML for object externalization allows for good inter-version compatibility for evolving data representation. One example is the "kibitz" sub-object that can appear in certain Banjo objects. The idea is that there are perhaps a dozen different data items that could be used in a kibitz and perhaps more that haven't yet been considered. Furthermore, it's unlikely that all engines will use all of the possible attributes. So a kibitz object allows for any subset of tags and the support libraries (either/both possibly not up to date versions) will simply pass the values of unrecognized tags.

A kibitz object will have both a "score" tag and a "pov" tag and so be able to accommodate whatever the end user's preference with respect to point-of-view score reporting.

The actual human readable representation of a kibitz object is entirely up to the particular client, so each user can specify the tag subset, interpretation, and presentation.


2) The rating reference standard

An instance of a Banjo server can have one of more designated semi-permanent program players available for matches, one opponent at a time. Each of these special reference programs would have a constant rating that is established by testing and mutual consent. All other players derive, directly or indirectly, their ratings from these reference players. For example, I could connect my Novag Citrine dedicated unit to a Banjo instance, give it a fixed rating of 2050 Elo, and leave it running and available for all challengers. Other possible reference opponents could be connected, although if they aren't dedicated units then a separate machine would have to be dedicated for each so as to reduce load variance that could affect performance.

Different Banjo sites could exchange rating data so a player would have a portable rating that's valid across all such sites.

And to avoid unnecessary server load, the rating lists would be automatically and periodically posted to a web site hosted on a separate machine.


3) Alternative rating systems

In addition to the traditional Elo system, the BCF system could be used for the convenience of UK authors. The Glicko modifications to Elo as seen at some ICS sites might also be made available.

When enough players are present to form an appropriate distribution, a version of the Binet/Terman IQ measurements could be employed with a mean of 100 and a standard deviation of 15. These would be relative ratings as the interpretation of a particular value would change as the player pool changed.