View Single Post

Ramalina's Avatar


Ramalina
07.03.2021 , 09:26 AM | #31
As far as matches going into averages 20 was basically the smallest number I figured might give a decent answer. Data storage is cheap so there wouldn't really be a problem with going up to 50 or 100 matches. There's sort of a tradeoff between accuracy of the long term average and the speed with which the average updates for recent changes in skill. A smaller pool of matches makes the rating more responsive to changes in player performance, a bigger pool is less sensitive to variation due to randomness or due to "going to fly this lame build for the lulz," that might not represent true skill. I tend slightly to the quicker change side of things in preference.

I'm also somewhat influenced by having seen a fair bit of team vs team play back in the day. If you have a couple of SRW groups going up against Drakolich and friends, you get extremely high skill, but the kill numbers tend to be not that impressive. With enough good players on each side kills per player go down steeply. For one thing people are much more likely to take survivable damage to shields more often because they go evasive before a kill is landed, and there are also enough skilled players around that instead of one ace picking up all the kills whoever happens to be close and more or less pointed at the target gets the kill, because basically everyone is skilled enough to get the kill. High kill counts are just as much a symptom of poor matchmaking and an abundance of low skill players in a match as they are an indication of the skill of the player getting the kills.

In a best case scenario, where GSF population and average skill go up, maybe even to the point where team vs team matches become somewhat common again, ideally the matchmaker doesn't start getting thrown off by solo vs team fliers or by time of day someone typically is flying. Basically if I have an option to make the matchmaker blind to things that are less reliable indicators of skill then I'm happy to do that.

Personally, at the end of a match if I'm looking for whether or not there are names that I should look for in the future my ranking of stats goes: accuracy, medals, damage, combined kills and assists. Any stat can be cheesed of course, but accuracy gets at fundamental flying and shooting skills and medals gets at playing objectives (which includes kills and assists). Damage gets at how actively engaged the player is in combat. I tend to regard the kills and assists as byproducts of those factors, byproducts that are more vulnerable to variation in conditions than the underlying factors that produce them. I don't think I've ever seen someone with over 20 kills do poorly on accuracy, medals and damage, but there are plenty of times when I've seen someone that I know is very good get 6-12 kills, but still have have excellent accuracy, medals, and damage. That's sort of typical for a strong player in a well used support ship.

I do also have in the back of my mind the scenario, what if Drako decided to do a hard support "no kills challenge." What sort of win rate would that produce if solo queuing for a bunch of games? I'm wagering well over 50%. There would still be accidental kills I'm sure, but a scoreline like: 2 22 0 108570 52% 278 17, would be perfectly reasonable. Likely game-winning as well.
"A padawan's master sets their Jedi trial, Rajivari set mine."
- Zhe Lian, Sage.