According to my Calculations

[timeconsumer]

Welcome back to the newest installment of According to My Calculations where we come to take a slightly statistical look at NSFL data.

In case you missed the last edition on Running Backs in the NSFL you can find it here: http://nsfl.jcink.net/index.php?showtopic=1227 We discussed how running back YPC ratings correlated with their individual attributes for week 1 NSFL data.

As a brief followup I would like to touch on this subject again with data up through week 8 to see what a larger sample size of games does to the analysis.

We'll open up with a quick look at the yards, carries, YPC, and average attributes of each running back. (The attributes are averaged because they change, with updates, I account for that in my raw data)



Not much to see here. Lots of names, lots of numbers, no clear idea. Now let's look at our original methodology to see what the data tells us with correlation coefficients.



Now if you recall from our last discussion this indicates whether or not we have a linear relationship with the data. From these coefficients through week 8 I can tell you that I am not seeing anything even close to a linear relationship here between weekly attributes and weekly YPC. Basically, this analysis is flawed. So I will not attempt to do my regression analysis from the previous article.

Instead let us try to approach this from a different angle. We know that attribute increases are non-linear, going from 50 to 60 strength is not the same benefit as going from 90 to 100 strength. So in that case we need to weight or attributes in order to find out what benefit we might find in increasing each attribute. Unfortunately for you, I will not go into what my weighting method is, it's secret. This is no peer reviewed piece of research so I'm okay with that. If you don't like it, go kick rocks.

So with my weighting in place let's see what happens to our correlations.



Can this be possible? Can intelligence really be THAT highly correlated with a higher YPC than just about anything else (besides pancakes?) Let's dig further to see how significant our findings are.



And here we have the big table. So what I'm seeing here is that 60% of the variance in YPC can be explained by all of the stats I have listed. However I am not showing any level of significance that I can point to and say yes, this is a good representation of a strong relationship. It is mere okay. From the conclusions I can draw from this really the only significant finding is that for each additional pancake block per game, the YPC increases by 0.02. A small amount, but if a team averages 10 pancakes more per game it begins to show its value.

The other stats that while not showing a strong relationship are at least somewhat reasonable are speed and endurance. Endurance is showing a slightly negative relationship, which is contrary to the first article. Indicating that putting TPE resources into endurance rather than other attributes could be holding your player's progression back. Speed on the other hand is showing a somewhat positive relationship, which indicates that for every TPE point spent in speed a player is getting roughly 0.01 increase in YPC.

Unfortunately without a good strong relationship for the other stats I have to call the entire analysis inconclusive as we simply do not have enough data. The only thing I can see as being very important is the performance of the offensive line. So if you want to run better, block better. Shocking....