Ye, the Equalizer mutator is something that I can continue working with. I need to first make myself comfortable with UT99 crowd again because it can and will go meta (meaning algorithm affecting players and the backreaction from players).
The ut2004 equalizer, you correctly identify, is fledgling stage, in the sense, we made the solid enough framework (mutator gauging CTF skills, storing over SQL database, and performing PHP query to retrieve the information), which worked gracefully for player base of 30,000 players or so, however we couldn't figure out the criteria for midway balancing, with players leaving (or "specting") just because they have been equalised and they feel like they'd be doing more work instead of doing their habitual cakewalk.
The balancing algorithm, suggested by some previous ut2004 admin, along with refinements, takes the form
At the beginning of match
- Seperate Bots and Humans
- Sort the EQPlayerInfo of the Human players with some basis parameter (total score vs rating (computed from caps, seals, and covers along with score) decision)
- FillUp teams like so
- Highest rating player on red
- Next two on blue (ye I never understood the reason of two :) )
- Fill the rest alternatively
- Restore bot-crowd at the end of the routine
There were heavy discussions from the ut2004 crowd about two chief areas
- Rating generation mechanism (I see UT Stats here could be of tremendous help)
- Mid-way balancing (heh, the crucial factor)
IMHO, issue 1 is solved for UT99. Issue 2 is, well, to be tackled gracefully!