Evaluating Rookie QBs

Math for Evaluating Rookie QBs

(Photo Credit: wallpoper)

Wouldn’t it be nice to have a crystal ball that told you which rookie quarterbacks will turn into elite NFL starters? For every Drew Brees, we have all drafted our fair share of Ryan Leafs. So how can we tell them apart in advance? Put another way: Should you draft Teddy Bridgewater or Johnny Manziel this year?

To answer that question, I looked at the last decade of NFL drafts and cataloged every college quarterback that was drafted to an NFL team. After that I created a list of all the college stats for those players. I hoped to find the magic formula for what college stats translated into NFL greatness.

It was hard to quantify “NFL greatness,” so I focused on how many NFL games each quarterback started and how many fantasy points they averaged per start. Rather than take games started as an absolute value, I kept track of it as a percentage: how many games did each quarterback start, divided by how many total games their teams have played since they were drafted.

Once I had those two lists established, I used the power of Excel to help me calculate the correlation between various college stats and subsequent NFL production. My goal: learn a little about evaluating rookie QBs.

First, Some Math

To help you understand my findings, let me first summarize how correlation works. The important thing to know is that correlation is a way of measuring how closely two lists are linked together. It is measured from 1 to -1.

A correlation value of 1 means that the two lists are perfectly linked – another way of saying you’re actually comparing a list to itself. A value of -1 means that they two are perfectly linked in an opposite fashion: As the values in one list get larger, the values in the other list get smaller.

So the closer a correlation value is to 1, the more we can trust one value to be a rough approximation of the other. For example, if we compared a list of daily temperatures to a list of how many people went swimming that day, we’d likely see a high value because as temperatures go up, so do attendance numbers at your local pool.

In contrast, a correlation value close to -1 means that an increase in one is often followed by a decrease in the other. Your typical supply-demand graph fits this model: As a store increases prices, sales go down.

So now let’s turn this thing to football. Here’s what I found:

We Can Actually Predict Who Will Get NFL Starts

I started by asking myself what college stats are important to focus on. I cast a wide net, hoping to find some needles amidst the haystack of stats we now have on any given college player, to help me in evaluating rookie QBs.

I started with interception percentage, thinking that a college quarterback who protected the ball would likely turn into a solid NFL starter. As it turns out, there is only a 0.16 correlation value between college interception percentage (in a QB’s final year in the NCAA) and what percentage of his NFL games that quarterback went on to start.

Well how about height? Much is made in this numbers-obsessed era of a quarterback’s height. Although you might want your quarterback to hit a minimum height, there is an insignificant correlation between increased height and greater odds of starting in the NFL. The actual value is 0.14, a slightly worse indicator than we saw with interception percentage.

Well if neither of those helps, how about completion percentage? Is that any better? The answer: it’s actually worse. There is only a 0.13 correlation between college completion percentage and how likely the quarterback is to become an NFL starter.

Finally, I began to find some better indicators. Draft spot actually has a 0.35 correlation to NFL starts, which makes intuitive sense. The higher a player is drafted, the more likely he is to start in the NFL.

Best of all, red zone completion percentage in a player’s final NCAA year has a 0.54 correlation to NFL starts. That’s an incredible number. What it should tell you is that you need to pull up some advanced stats on this year’s rookies and see how they performed in the red zone last year. For your own edification, here are a few: Manziel (63.3%), Carr (61%), Bortles (60%), and Bridgewater (55.8%).

Math Won’t Help You Predict Fantasy Points

As a final note, I have yet to find any college stat that has any sort of real correlation with how those quarterbacks score in terms of fantasy points per start in the NFL. While NFL starts may be based more purely on talent, fantasy points are much more tightly linked to the situation a player is drafted into. Even the most elite talent would struggle in a terrible offense.

Here are the correlation values: Completion % (0.17), Draft Spot (0.13), Height (0.08), Interception % (0.002), and Three-Year Starter (-0.25).

Conclusion

My goal in this article was to take you behind the scenes in how I constructed the QB metric I referenced in earlier articles. I took these values we’ve been talking about — along with a few others — weighted them appropriately, and patched them together to create a metric for valuing incoming rookies. If you want to see how I applied this information to rank some of the 2014 rookie quarterbacks, you can find it in this article about drafting the 2014 rookie quarterbacks

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>