A Ranking Mean vs. Ranking Median Experiment (1 Viewer)

[freshly_shorn]

Inspired by the contentious 'Why do rankings use average for consensus" thread. I decided to try small experiment on the affect of the sorting the rankings by mean/alphabetical and median/alphabetical. The alphabetical sort was used only where average or median were tied.

I used the overall rankings provided today, and used the fantasy points forecast by Dodds to provide fantasy points. I assumed a 12 team, sepentine draft of seven rounds (84 players). No kickers or TD/IDP were included. No regard was given to positional need- best player was taken. I summed the FP value of each drafter's picks.. here are the results (negative difference favors median, positive difference favors average):

Average Median Diff % Diff

Player 1 1413.8 1374.9 38.9 3%

Player 2 1325.1 1321.4 3.7 0%

Player 3 1235.7 1314.4 -78.7 -6%

Player 4 1366.1 1198.5 167.6 12%

Player 5 1071.1 1191.8 -120.7 -11%

Player 6 1297.6 1082.5 215.1 17%

Player 7 1150.5 1159.7 -9.2 -1%

Player 8 1112.1 1485.1 -373 -34%

Player 9 952.2 953.5 -1.3 0%

Player 10 1353.9 1189 164.9 12%

Player 11 1180.2 1210.8 -30.6 -3%

Player 12 1073.6 1122.4 -48.8 -5%

Mean -6.01 -1%

1210.991667 Teams built off average rankings

1217 Team Based on Median rankings

So, as you can see, there is a small difference between the two overall, but there are big swings from team to team. This experiment would have to be repeated and applied to historical rankings/projections.. which I don't have access to at FBG. But it *may* be worth exploring, based on this result.

 

[JayMan]

MT and Chase... that statistical experiment has to be meaningless, right?

 

[freshly_shorn]

MT and Chase... that statistical experiment has to be meaningless, right?

Funny.

 

[videoguy505]

Now I'm confused. I don't think you can do this with Overall Rankings and then ignore positional needs. Reason being is that Peyton Manning is ranked #12 because RBs are worth more than QBs in VBD, but if we were sorting rankings based soley on FPs he'd be #2 overall. In a draft such as the one you constructed, no staffer would put Manning's ranking outside the top 3.

To run this you'd have to run it for one position only, instead of Overall rankings, right?

 

[hove]

 

[AB in DC]

No regard was given to positional need- best player was taken.

You need to define this.

 

[freshly_shorn]

No regard was given to positional need- best player was taken.

You need to define this.

I basically drafted straight from the rankings list, just to see what the effect was whether you draft from a mean-sorted list or median-sorted list. I didn't look at who was 'drafted', so it is NOT a true test of whether drafting by a mean ranking is different from a median ranking. I was curious at *how* different each slot could be, and what the effect would be overall, in this very small sample.I just wanted to see, in one experiment, how different it was to see if it was worth studying. I think it is, but you would have to draft according to need at some point to make it a legit test, and apply it to past years.