0

Lies, Damned Lies, and Cornerback Stats

Normally I’d have a piece up on another position group and three new players to break down the stats of. I actually have a considerable amount of data on cornerbacks Jason Verrett, Kyle Fuller and Darqueze Dennard, but I’m not posting it quite yet. Stats for CBs are rather nebulous. Whereas you can look at a wide receiver’s yards after the catch and visualize him running for 7.45 yards, you can’t do the same with burn, target, and passes defensed rates. You need a large group for comparison, so until I can get data on a few more CBs I’m going to hold off posting the entire breakdowns.

I want to take a break from posting stats for you to interpret and talk a little about how to interpret stats and some of the inherent fallacies involved in that process. Cornerbacks with their vague stats epitomize some of the pitfalls in statistics and football. My intention is not to dissuade you from using stats in your evaluation – that’s the entire point in my venture, but to show why they are a tool to complement scouting rather than replace it. That’s an important distinction that some people haven’t quite gotten.

What makes a good play?

Take a look at this play from Kyle Fuller against Devin Street and Pittsburgh.

 FullerStreet

Arguably this isn’t a positive play for Kyle Fuller. For whatever reason he doesn’t move until well after the ball is snapped. That’s a mental mistake and should not factor into his evaluation positively.

However, because Devin Street effectively runs a 5 yard out behind him, he has time to recover. He makes what I would consider a “deflection” which counts as a positive in the statistics for him. A deflection lowers his ‘burn rate’ (num passes beat/ num targets) and increases his overall pass deflection rate. If Devin Street runs literally any other route, Fuller gets burnt badly and it reflects poorly in his numbers.

Unlike ProFootballFocus, I’m not in the business of grading plays to go with my stats. I just don’t have the man-power or ability to go through every play, collect the relevant data and then assign it a grade. So while this play may have given Fuller an overall negative grade per PFF, Fuller receives a deflection and better numbers.

Small samples

Football is notoriously hard to use statistics to analyze due to small sample sizes. There are anywhere between 12 and 14 games in a college season, and even less if the player has missed any games due to injury. For some positions that’s not a big problem. If I have 250 rushing attempts for Bishop Sankey, I’m pretty comfortable that a minor change in a few runs won’t affect his overall yards after contact.

This isn’t true with CBs though. For Jason Verrett, arguable the top SR CB, I have charted 269 snaps against the pass. That in itself is a good size, but within that he only received 40 targets. 15 were completions, 11 defensed, 2 intercepted, 8 missed, and 4 pass interference. Personally, I include pass interference as a ‘completion’. The DB gives up yardage by committing the penalty and it effectively moves the chains in the same way a completion would.

So if I calculate burn rate (# passes completed/ # targets) including PI it comes out to 47.5%. That would be among the worst in last year’s CB class. You read that and suddenly you’re thinking twice about Jason Verrett as the top senior CB. However, take away the PIs and his burn rate drops down to 37.5% which would be up there with Dee Milliner and Xavier Rhodes from last year’s class.

Now of course I include PI in all calculations for all corners, so there’s a modicum of standardization – but it shows you how something as small as 4 pass interference penalties can affect a key stat.

The fallacy of a ‘target’

Here’s a play I always like to illustrate the point about what is considered a target for CBs. This is a play from last year of Johnthan Banks and an Alabama TD.

 BanksTD

Targets make up the crux of metrics for any player dropping into coverage. They’re going to affect overall target percentage, burn rate, passes defensed rate, among others. So does the above play count as a target against Banks?

While you can’t see it on the gif above, Banks clearly looks at the LB in confusion after the play. Did Banks screw up or did the LB? Do you attribute to Banks a target, completion and a touchdown?

In this case, I’d say you attribute the target/TD to both Banks and the LB because of the clear mental mistake. However, it’s not clear by any means. This same issue comes up when a CB has safety help over the top. Who gets the target when the corner and safety are equidistant to the WR on a go or corner route?

_______________________________

I of course try my best to dole out the appropriate blame and reward, but you have to remember that football stats aren’t always binary. That’s what makes this whole thing so uniquely frustrating and wonderfully challenging at the same time.

Those working on football stats, especially in relation to game play are in a position where they’re finding new things to chart, different metrics to create, and interesting ways to display that data. There will always be problems, but the benefits of having this extra data far outweighs ignoring it due to some small issues.

nugap

Leave a Reply

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