TS rating on test hosts
Posted: 16 Mar 2018, 00:15
Currently, there are a lot of games going on at the BA test hosts. This autohost does not use trueskill balancing and currently Triton is spending quite some time to manually chrank players in order to get balanced teams.
The rationale for this is obvious, games on the test host are not supposed to influence the TS rating of the players as to not discourage playing around with new, potentially unbalanced, game features.
Technically, this seems to be implemented by choosing a different mod-shortname for the pre-release builds of BA. SLDB then falls back to rank-based rating in absence of any TS rating database.
I've spent some time digging through the source of SLDB, replays.springrts.com, and SPADS. As far as I can say, the simplest (and probably hackiest) way would probably be to modify SLDB to map ([:alpha:])-test to \1, e.g. in Sldb.pm:1076. bibim: Any thoughts on this?
(Not opening a github issue, but discussing it here, for various reasons. Most importantly, I'm still not entirely sure if SLDB is the right place to solve this).
The rationale for this is obvious, games on the test host are not supposed to influence the TS rating of the players as to not discourage playing around with new, potentially unbalanced, game features.
Technically, this seems to be implemented by choosing a different mod-shortname for the pre-release builds of BA. SLDB then falls back to rank-based rating in absence of any TS rating database.
I've spent some time digging through the source of SLDB, replays.springrts.com, and SPADS. As far as I can say, the simplest (and probably hackiest) way would probably be to modify SLDB to map ([:alpha:])-test to \1, e.g. in Sldb.pm:1076. bibim: Any thoughts on this?
(Not opening a github issue, but discussing it here, for various reasons. Most importantly, I'm still not entirely sure if SLDB is the right place to solve this).