Leaderboard

The Etpub server is back up on a new ip 37.187.79.49:27990 and new name Teammuppet Etpub. It is running 9 maps Frostbite, aldernest, braundorf, bremen, caen2, sp_delivery_te, supply, goldrush and oasis.

Skill rating is visible on the legacy server since 2.76. Some players ask what it is on server, and some discussion in #freenode#etlegacy (the respective discord channel) showed, that the documentation on https://dev.etlegacy.com/projects/etlegacy/wiki/Bayesian_skill_rating and https://dev.etlegacy.com/issues/403 is not a short answer. So i try to summon up my understanding and questions from a users view, and hope that spyhawk from ETL will setup a FAQ with competent answers with developers truth. Calculation with some samples can be followed best imo with the paper: http://www.moserware.com/assets/computing-your-skill/The%20Math%20Behind%20TrueSkill.pdf Skillrating calculation and presentation can be activated serverside. Skillrating is calculated from the statistical propability to win or loose a game. Every player has a mean and standard deviation of this propability. Calculation of skill from this point is on page 15. That is why everybody starts with 0. It is in no way influenced by stats like XP, Kill/Death ratio, time to finish a map. So if that values dont matter what does? Win or loose. It is only possible to increase SR when in the winning team (or reducing the standard deviation over-proportionally by loosing, but then mean is reduced too). Amount of SR adjustment depends on the overall SR of the team and the SR difference of the teams. (see page 29, mean and standard deviation are treated separately). Dropping, joining players are taken into account. How map is taken in account is not clear for me, maybe as evidence (see page 16). What i like: - completely new ranking system that only rewards winning a map, so no necessity for spree hunting, XP collecting as long it is not useful to reach the objective - Admins can choose to use it or not - "simple" system for a complex situation - should have quick convergence at fixed player setup (see page 38) - starting from zero avoids disappointment, if SR ends up below 25 - it is bound to server, so players might be more loyal to find their real SR - cant stop admiring the ETL-People that implemented such a modern system in that Oldie What i dont like: - When is convergence reachable if you have constantly joining players with high standard deviation (SR0)? - The deviation for bots is very small (they play all day), so if you play alone with them and win your deviation decreases fast - Not sure if the mean for bots is comparable, as they tend to play alone (guess this might even lead to wrong Map evidence, as win/loose depends very much on path setting, is this contradictory?) - I will never be General again What is your opinion on the system? @Kemon, Spyhawk please correct misleading information in this post.