Saturday, May 21, 2016

Please - don't draft a kicker early

My brother (@EaglesRewind) commented on the other day about this tweet, referring to the Tampa Bay Buccaneers drafting strategy:

The Bucs reportedly had a first-round draft grade on kicker Roberto Aguayo, who they actually drafted in the second round of this year's draft.

So under their logic, the Bucs got a steal by snagging a kicker in the second round!

This almost made my head explode - and then I read this article which evaluates whether the Raiders 2000 drafting of Sebastian Janikowski in the first round was a good idea or not.

The article doesn't come down one way or the other, the writer sees both sides of the 'debate.'

That did actually make my head explode.

I wanted to dig into it - and I did - and the answer is Janikowski pick was NOT a good one.

Now, I don't have to waste time talking about the case against drafting a kicker (even one as successful as Janikowski) in the first round. Anyone who's reading this blog is likely already familiar with all those thoughts (e.g., massive opportunity cost to pick at a low-value position). So we'll focus instead on the arguments 'for' drafting a kicker high.

The primary argument I've seen, is that Janikowski has been a good kicker for so long, that he was worth the investment. He played from 2000 to 2015 -- which is admittedly an absurdly long career for the NFL.

And when someone wants to make that argument, they'll point to something like PFR's Approximate Value metric. And it's true, Janikowski generated an career approximate value of 53 in aggregate across his career. The average of all other first round picks output from 2000 through 2015 is only 50 in career approximate value (this includes seasons when these players didn't play).

OK, by only that one measure, you could argue Janikowski was worth it -- but it's tremendously flawed logic, because it ignores one simple idea.

The time value of 'approximate value'

There's never any talk about a discount rate for this value, and that ignores the reality that performance generated for the team today is much more valuable than performance generated 15 years from now.

Talking about discount rate is usually reserved for financial discussions - but it's pretty straightforward. What's worth more to you, a dollar today or a dollar given to you in fifteen years?

It's also the reason why a general NFL draft rule of thumb is trading for a draft pick this year will cost you a pick in a higher round next year.

So if anyone claims that Janikowski was a good pick -- it's because they effectively have a discount rate of practically zero.

I'll show you.

I took data from the first round of the 2000 NFL draft - including approximate value generated by year by player. First, let's look at cumulative AV generated by Janikowski over his career vs. every other player in the first round.

You'll notice that Janikowski starts off behind and the gap grows very quickly - until it's highest point in 2008. Keep in mind the average of all first round picks includes both those who played well (Brian Urlacher, Career AV of 118) and those who, well, didn't (R. Jay Soward, Career AV of 2). For the first nine years in the league, that non-Janikowski average pulled away from the Raiders kicker.

That's because, as you'd expect, Kickers don't generate a ton of value by themselves. But in the world where you're trying to rationalize taking a kicker first, you'll point out that because kickers can play forever (and the Raiders happened to find a good one), that eventually Janikowski matches and passes his first round peers by 2015.

That's great, but we're evaluating a decision made in 2000. So if you were back at the draft in 2000, fresh off a sigh of relief that Y2K computer malfunctions didn't make all our planes drop out of the sky, and if you had perfect information on how Janikowski and his peer group would perform, would you still make the pick?

That's where the discount rate comes in.

Now the typical NFL discount rate seems like it would be pretty high. The NFL is a win-now league and most QBs/Coaches/GMs don't keep their jobs past a few years. That means you'd willing to give up a lot of future value for production today. No one (at least that I've seen) has put a good handle on it, but it's a relatively high number.

In the corporate world, you typically have a longer time horizon and things are a bit less focused on immediate return. Often you'll see discount rates between 10-20% for the valuation of internal projects (though of course, this can vary)

So, what I thought would be helpful is to run scenarios and see what discount rate you'd need to have in order to make Janikowski's long term AV return 'better' than his peers. If that was your personal discount rate as an NFL GM, you'd be thrilled to take Janikowski. If your personal discount rate is higher, you'd want to take the peer group, who will generate more value sooner, which is what you want.

Below is an illustration of the valuation curves for Janikowski and all the other first round picks averaged together.

Just to clarify, these 'valuations' are basically how much AV it's worth on draft day in 2000, given a varying discount rate.

If you squint, you'll barely be able to see the point at which the Janikowski curve crosses the peer group curve. Because once your discount rate exceeds 1.6%, the peer group average is worth more to you.

The only situation where you'd ever value Janikowski more than the rest of his draft group is if you have a discount rate extremely close to zero.

The implication here, to me, is that if you took a time machine and with all of today's data went back to the Raiders draft room in 2000, you would never make this pick again (you'd probably take Tom Brady instead).

And so when I see the Bucs take a kicker in the second round, and I try to think through their case that it's not an idiotic decision, I just can't do it. You basically have to assume that Aguayo will play as long as Janikowski, be significantly better than Janikowski, or you have to have a discount rate so small you must be planning to hold your GM job for the next 15 years.

Call me skeptical.

Friday, May 6, 2016

Do Small School QBs Underperform?

Welcome to the Carson Wentz era!

Now that the Eagles draft is behind us, we can start thinking about what's in store for us in the future.

The Eagles have a shiny new QB in North Dakota's Carson Wentz -- and while they still haven't figured a way out of the messy situation San Bradford has created (yes, I'm putting the blame on him), that shouldn't stop us from wondering what a Wentz' led Eagles team could look like.

I have to admit, I wasn't following all the pre-draft hype that closely, but once the Eagles moved up and were in a position to draft Wentz, I started wondering about one specific narrative that would obviously become attached to him.

Wentz played at North Dakota State, which is about as far away from a college football powerhouse as you can get unless your college has the word 'barber' or 'clown' in front of it.

I'll be honest, that concerned me. Would Wentz be at a significant disadvantage because he played in college football's lower level?

My first thought was that this will be easy - let's just look and see all the other lower division quarterbacks that have been drafted and see how they've done.

Only problem...look at this list

That's a link to all non D-1 quarterbacks drafted over the last twenty years.

You'll notice that only one lower level QB in 20 years has been drafted in the first round. Joe Flacco.

Not exactly a robust set of comps. It felt very unsatisfying.

But then I had a thought, that just looking at D-1 vs. non-D-1 QBs might be to strong of a distinction.

What I'm trying to see is whether quarterbacks at lower level football programs perform disproportionately better or worse than those who play at higher level programs. It's easy to see why the narrative might logically apply -- a higher level program is one with more pressure, more fans, more talented athletes -- so it seems plausible that QBs exposed to that crucible and perform well are more likely to succeed in the NFL than those who don't (and ostensibly, might not have their weaknesses exposed).

Yet, D-1 vs. Non-D-1 QBs is only one way to get at that distinction - it's a binary one and as we can see, there are near zero observations.

So we need another basis for comparison to measure how 'big-time' a college football program is.

What I thought would be just as good of an indicator of how 'big-time' a college football program is, is the capacity of its football stadium.

A college with 100,000 fans in the bleachers every Saturday is going to be one with a lot of exposure, pressure, and talent.

A college with 20,000 fans in the bleachers, is going to have less of all of those things.

If we use that capacity as our metric, all of a sudden we have a much broader set of observations -- you can play at a small school in the highest level of college football (e.g., Tulane, which has a capacity of only 30,000). Now those small stadium teams may still play against better competition -- but we have to use what we can.

I'd argue that going a QB from Tulane might be better compared to other small school QBs than a QB who played at Notre Dame or USC.

So I took every quarterback, from all schools, drafted in the first round since 2000, and charted them on two dimensions:

1 - The capacity of their college football stadium (thanks Wikipedia)
2 - Their career Approximate Value (Pro-Football-Reference's stat) per Game

If highly drafted quarterbacks from small schools (e.g., Carson Wentz) are more likely to underperform, we'd expect to see QBs from smaller schools have a lower approximate value, while the QBs from football factories will outpace them once they get to the NFL.

That red line is the trendline, which shows a slight negative correlation. Across our 41 quarterbacks, the data does not suggest that those from smaller schools are any worse than those from big schools. If anything, it suggests the opposite might be true.

Now this isn't a perfect methodology, but it's enough for me to feel comfortable that smaller school QBs don't clearly have a handicap relative to guys from bigger schools. But you'll also notice that the sample size of small school QBs is small -- like REALLY small.

Carson Wentz's stadium at North Dakota St. has a capacity of only 18,700, smaller than any QB drafted that high since 2000. The only guys even really close are the aforementioned Flacco and Big Ben (Miami of Ohio had a capacity of under 25k).

Even if you go back 40 years - you only have a few more first rounders from small schools.

Steve McNair - 1995 (Alcorn State)
Ken O'Brien - 1983 (Cal-Davis)
Phil Simms - 1979 (Morehead State)

So while the data suggests there's not an inherent issue with small school QBs taken in the first round, these cases just don't come along very frequently. There have been six (including Wentz) in the last 40 years.

Of course, looking across that list, all but O'Brien led their teams to at least one Super Bowl appearance.

So we got that going for us, which is nice.