Quote:
|
Originally Posted by thebest
regrading the consistency index I have following observation
1. it gives too much emphasis on the two most extreme outliers i.e. highest and lowest. But the outliers would have same impact if the second highest and lowest are close that is this formula would make very big difference between two similar performer. Even making the better player worse off If sample size is small i.e. less than 30. For example,
Player A has following score
100, 0, 0, 100, 20, 25, 20, 25, 20, 25, 20, 25, 20, 25.
He has an average of 30.35 and AC of 27.28
Another player B has same score but his second highest and lowest two are 20 and 25. Thus he has the following score
100, 0, 20, 25, 20, 25, 20, 25, 20, 25, 20,25, 20, 25
Thus B’s average score is 26.3 and AC of 22.5 though is more consitent than A, but the index show A is more consistent than B. I would rather prefer Rajesh’s consistency index which give importance to all score but deflate the extreme value by considering standard deviation in the folrmula. Rajesh one is average/ SD. B’s consistency is rewarded here as he has better Consistency index of 1.12 where A has 0.99
|
Excellent post and thank you for it. I see you point very clearly indeed ...
I took such a small number of matches to define a particular peiod simply because of the frequency with which we tend to shuffle players. I chose the number 10 in that light and also because I'd personally like to see a player get at least 10 matches on the trot before we appraise him either way over a fixed period of time, unless of course we're practically FORCED not to on account of his performance.
Also, there's something inherently simple about my AC that makes "ballpark" sense to me. For example, if a player scores 100-10-30-30-30-30-30-30-30-30 or 350 runs over 10 matches, his API will be 35.0. But once we drop the top and bottom scores of 100 and 10, and divide the remaining 240 by 8 (=10-2 matches with scores of 100 and 10), we get 30. Since the batsman has scored 30 many more times than 100 or 10, 8 out of 10 times to be exact, AC can tell me that he gets 30 CONSISTENTLY, not 35.
Just keeping it simple I guess.