Winnipeg Jets 2013 Lineup and Statistical Analysis (See Post 1 for reference)
View Single Post
02-05-2013, 04:47 PM
Join Date: Jan 2012
Originally Posted by
is this for 5v5 only? if so, then it would seem to be appropriate if starts and finishes are approximately 50/50 overall (probably pretty close). if a player is, on average, ending a higher percentage of shifts in the OZ than they've started, then they're doing something right.
The main issue is the division by 50 - OZS. The closer someone is to 50-50, the screwier the stat would get due to possiblity of a small denominator. or conversely, the farther someone is form 50-50. imagine someone at 51 OZS, and someone else at 52 OZS. Let's say they both finish 10 points better then they started (61 and 62). so you get 10/1 and 10/2. half the stat for something pretty identical. somebody else could start at 51 and finish at 56 and get the same as the guy at 52 and 62. or someone coudl start at 60 and go to 100 and still not appear as good as the guy at 51 - 61.
so this stat would still have its issues.
I'm not sure what you mean by
'remove the integer
there it is.
thats what i was looking for. I think the "right" way to do it is as i said in my last post, would be to apply diminishing returns on movement away from 50, but i don't have the math skills to do figure that out.
as for removing the integer i meant turning all negative numbers into a positive, as otherwise you'll get incorrectly "flipped" +/-
EDIT: also regarding 5v5, i did not due 5v5...orI may have...i'm not sure now. Nto overly important at this tage as it's more about the equation then the data, but yes- when doing it for realy you'd want to be using 5v5 only.
View Public Profile
Find More Posts by Grind