weighting doesn't really brings the score to average scoreScarlet Evans wrote:
There can be many different ways to define "average performance", but I think that weighted mean is a very good solution. With some other options, harmful things could happen, like a vast amount of weak scores dominating the better scores and bringing the total performance down; or things like making "farming" performance too easy; or something else.
Your best plays kind of indicate what you are capable of doing and the scores that are not added automatically, but you acknowledge as very good, can be added manually now <3
But you are right about the first line, indeed, it can be tedious to manually upload all the best chokes >,.<
Let's do some math When the points for a certain skill are calculated with a weighted mean, where the weights are 0.95^(n-1), then for all scores being equal, we have:
# first 14 scores include more than 51% points (weights),
# first 32 scores include roughly 66% points,
# first 32 scores include more than 80% points,
# first 45 scores include more than 90% points,
# first 59 scores include more than 95% points,
# first 76 scores include roughly 98% points,
# first 90 scores include more than 99% points.
calculated using (1-q^n)/(1-q), where q =0.95
What I mean is, that even if we include only our peak plays, having enough of them makes the rest almost negligible. It can don't look like that, if we rely on intuition instead of numbers, but you need quite many of these best scores to dominate the rest, so I think it's kind of a good way to represent an average performance.
It can be easily changed to give the lower scores a greater weight, but I don't think it would be a good solution. Players keep improving, so the best performance is rising and it would be too tedious to farm enough of maps to represent your skill
If it's not what you meant, feel free to correct me^^.
weighting stops people from farming the scores map by map (same reason was applied from ppv1 to ppv2)
it doesn't really calculate average when we can have people who can't duplicate their top plays (or at least somewhere close to it)
For example, we have players who have their peak performance to perform 99% FCs but their everyday average performance on maps of same difficulty is like bad 95% acc bad combo clear (believe it or not but this happens more frequently on a lot more people than we thought)
this actually makes the "top play only" calculation an extremely inconsistent way to measure average skill
they are good in calculating your peak performance stats, but not your average.
Also, having enough top play doesn't make the rest negligible, it makes the rest incompetent enough to show it as one of your better performance which might or might not be crucial for average performance calculation.
But then, a system that can't have human or smart AI to check every play manually can't really calculate average performance
The closest thing is to probably register every top play and have their score to decay over time, but that's highly subjective