Wednesday, April 6, 2016

Should NFL teams draft developmental QBs?

Matt Barkley

Mike Kafka

Andy Hall

All of these guys were at one time the Eagles 'developmental QB of the future.' It's a standard NFL draft narrative for teams to use a mid/late round draft pick on a quarterback when they already have confirmed starters on the roster.

When that happens, pundits love to talk about how the pick is a plan for the future, how the young quarterback will come in and learn at the foot of the experienced starter and be groomed to eventually take over.

The problem I've always had with this narrative is that I don't think it really ever plays out, and that those picks are generally a poor use of draft capital.

But I've never seen the data - does it make sense to use late round draft picks on quarterbacks? Do they succeed more often than I think they do?

So in the interest in destroying cliched narratives - I decided to take a look.


I started by pulling data on all quarterback seasons since 2000 (because Y2K is a nice round time bound). For each quarterback-season, I included the way in which that quarterback was acquired, and broke it down across the following buckets:

- First round pick
- 2nd/3rd round pick (Now Day 2 draft picks)
- 4th/5th/6th/7th round picks/Undrafted FA (Now Day 3 picks)
- Free Agency (excluding undrafted guys)
- Trade (yes, there are a few of these guys too...)

A quarterback's acquisition definition held as long as they didn't change teams, but once they move teams, their next season starts a reclassification. For example, Peyton Manning is defined as a First Rounder for his tenure with the Colts, but defined as a Free Agent for his tenure with the Broncos.

We also needed some measures of performance - so I'm leveraging Pro-Football-Reference's 'Approximate Value' metric. That's a metric created by the PFR guys to try and generalize performance over the course of a season - and I think it's better to use this type of measure rather than something like passing yards, QB wins, or something else that has some very tricky flaws.


The first look I put together was a view of Approximate Value distribution across our acquisition methods for each season. This was to see how much 'value' was generated by each group of quarterbacks, and see how that shifts over time. The year over year mix is illustrated below:

The first thing that you'll notice is that first round quarterbacks have typically generated 40%+ of all quarterback value over the last decade. If you include Free Agents, that jumps to about 60% of all value.

Late Round picks? Less than 20%.

A final observation from this data is that the mix has definitely shifted over time. The share of value created by first round picks has increased, while the share of value from picks made in the second round or after has declined from well over 40% in 2000-2001 to 25% in 2016.

Interesting, and we'll come back to that.

But you could look at these distributions and raise the question of playing time as a driver of that mix. First round quarterbacks are likely to get more starts than quarterbacks acquired in later rounds, so what happens when we adjust for that?

Great question.

I went back to the data on each quarterback season and adjusted the Approximate Value to a per game basis. This way, we can control a bit for playing time and the natural tendency to start first rounders won't skew our data.

Once we created our per game approximate values, I graphed the distribution of results by acquisition method.

We can see that, on a per game basis, the middle 50% of first round quarterbacks is significantly higher than all other acquisition methods. The remaining methods are relatively close to each other, but the group with the lowest middle 50% is our set of late round/undrafted quarterbacks. When you adjust for playing time, late round guys do significantly worse (except for the maximum value which is quite high -- we can call that the Tom Brady factor).

One last quick measure I looked at, almost as a check, was Pro Bowls. I know Pro Bowls aren't the best metric to judge true performance, but it's correlated with performance and I wanted to check what we've already seen here. So, below, is a quick list of the number of quarterbacks in each acquisition method bucket, the number of Pro Bowlers, and the share of quarterbacks in that bucket who have been to at least one Pro Bowl. (Note: players are unique by acquisition method, so Peyton Manning is in here once for the Colts as a First rounder and once in here for the Broncos as a FA)

- First Round: 58 quarterbacks, 19 pro bowlers, 33%
- 2nd/3rd Round: 55 quarterbacks, 8 pro bowlers, 15%
- Late Round/Undrafted: 168 quarterbacks, 8 pro bowlers, 5%
- Trade: 47 quarterbacks, 8 pro bowlers, 17%
- Free Agent: 125 quarterbacks, 12 pro bowlers, 10%

Again, this shouldn't come as a surprise. Late round picks at quarterback rarely do much of anything. They're extremely unlikely to make the pro bowl -- whereas a third of first round quarterbacks since 2000 have made it.

But let's bring back a point our earlier analysis showed -- that the value created by late round quarterbacks has declined dramatically over the last 16 years. I think that's extremely interesting, particularly given the fact that NFL offenses have shifted more towards passing and placed more of a premium on quarterback play. Few would argue that quarterbacks are more valuable than ever.

But quarterback value is being created more and more by those taken at the top of the draft. The Pro Bowl data confirms this a bit -- the most recent late round quarterback draft pick to make at least one Pro Bowl was in 2006, ten years ago (and it was Derek Anderson!). Maybe Kirk Cousins changes that someday, but less and less value is being generated by quarterbacks taken in the later rounds.

I thought part of it might be, that with increased emphasis on quarterback play, teams are reaching for QBs earlier and earlier in the draft and that's had a natural inflation effect on where they're drafted.

But I checked the data and it's just not true.

Are these newer first round quarterbacks just better than they used to be?

I don't think that's the case either.

At this point, my theory as to what's driving the shift is QB longevity. Quarterbacks drafted in the first round are more likely to be good, but with both improved medical technology and tweaked rules to keep them safe, quarterback tenure should be increasing. So I'd hypothesize that good quarterbacks are keeping their jobs longer because they aren't getting hurt, and the longer they play the more the mix shifts to more first rounders as teams without them bring them in.

We've seen the data that shows just how much more value first rounders create than other acquisition methods -- the implication there is clear -- the best way to get a good quarterback is to take one in the first round. And as those teams find good quarterbacks in the first round (unless you're in Cleveland), they hang on to them.

But the other implication is that taking 'developmental quarterbacks' in the later rounds is generally a waste of time. I'd have to check with other analysis on other positions' success in the later rounds, but I'm fairly confident the opportunity cost of taking a quarterback vs. another position is non-trivial. I'd suggest that teams looking for a developmental QB are better served signing a free agent that's been in the league, and use their picks on other positions with a higher likelihood of success.

And if you find yourself in need of a quarterback -- remember the best place to find them is in the first round of the draft. The idea that you'll get a Super Bowl QB in the later rounds is pretty much a fantasy.

No comments: