B-Ratings, all matches/ratings recalculated
2016-01-06 04:50 CET
Matches with custom game settings and factories were removed from the standard rating.
To provide balancing data for Vampiric PQL CA, instaFreeze, instaCTF and other mods there is now a separate B-rating for all custom matches.
Support for the B-ratings on the web site is currently under construction.
The B-ratings can be used by minqlx for balancing by using the /elo_b API URL instead of /elo
With that in place, FFA, CTF, FT and TDM have been recalculated to remove custom matches from the main ratings and put them in the B-ratings.
Matches with custom game settings and factories were removed from the standard rating.
To provide balancing data for Vampiric PQL CA, instaFreeze, instaCTF and other mods there is now a separate B-rating for all custom matches.
Support for the B-ratings on the web site is currently under construction.
The B-ratings can be used by minqlx for balancing by using the /elo_b API URL instead of /elo
With that in place, FFA, CTF, FT and TDM have been recalculated to remove custom matches from the main ratings and put them in the B-ratings.
What's the server IP:port ?
There won't be any ratings for these game types though.
For testing purposes I lowered the minimum to 1 per team and reloaded that 1v1 match, which shows up in the match index now: http://qlstats.net/games?type=ad
The limit is now back to 2.
If my nickname is all uppercase, i'm not online but will receive the messages later.
I restarted it about 3h ago.
That removes all possibilities of fine-tuned customization of the servers.
The best thing I can do is look at factory names and only allow known ones to be rated.
If I throw random stuff in the rating process, the ratings become random too.
The B-rating is only listed on each player's profile page. There is no ranking for it.
The purpose of the B-rating is to provide data to minqlx for balancing non-standard game formats like 8v8 vampiric PQL CA
If the match is rated, but you have less than 50% round (or time) participation in the match, you are not rated for that match.
why do I have no rank yet?
http://qlstats.net/player/18164#ca
I can see a lot of players without rank. Why is that?
Would like to understand this :)
You are only playing on servers with non-standard settings (vampiric PQL) and those results can't be mixed with regular matches.
The purpose of the B-ratings is to provide data for balanced team shuffling for non-standard matches.
Ratings are designed for team balancing, nothing else.
On popular demand I gave in and added ranking for the main game types. Ranking means comparing and ordering rating values. For that to make sense, the rating numbers must be comparable and that means, the matches used to establish those rating numbers must be comparable. Therefore only matches with similar game settings can be used for that. If you add 5 inches and 10 centimeters you get 15 whatevers.
Weeding out the ratings to support ranking had a big side effect. No more rating data for people who play only non-standard matches, like PQL CA or iCTF, iFT, ... So I added the B-ratings to allow matchmaking for those again.
And I'm not starting it all over again with C-, D-, ... Z-ratings. It would produce shady numbers, waste a lot of my time and waste server resources tracking dozens of different ratings for each player.
Bayesian statistics is like a machine learning system that you have to train with a large amount of consistent data so the ratings start forming a meaningful distribuion.
Mixing different sorts of input data will degrade the quality of the resulting ratings, which will make the existing team balancing for A-rated games worse.
The ratings are ment for matchmaking (= team shuffling, finding similarly skilled opponents), not ranking. I'd rather remove all the rankings entirely than add more.
a little leaderbord for the capql player <3
i'm more tempted to remove them all.
First, thanks for all the work you're doing with qlstats, it's really great.
About B-ranking, I understand it would be "too much" of rankings.
My question was more, why considering the A-rating should be all weapon ? Indeed, if you look at FreezeTag, there are much more players (bigger community) in Insta Freeze Tag than in All weap freezetag. Tbh, players who feel like playing all weap play in CA, TDM or Duel.
FreezeTag is a mode which became famous and populated thanks to people who plays Instagib.
What would you think about reversing A <--> B in FreezeTag section ? So IFT players could have their ranking, with much more players than FT curently does.
Thanks for your attention,
New players (noobs and fake accounts) are thrown in the middle of the heap and then move up or down as they play matches. When someone with a fake account starts at 1000, he messes up more matches till he reaches a more accurate rating and may even overshoot it significantly when his increases are too high.