Saturday, March 19, 2016

Saturation Drafting in the NFL

The NFL draft is coming up, and as we approach the draft, it's great to debate all kinds of theories on how teams should maximize their resources.

One theory I've debated with @EaglesRewind is the idea of saturation drafting. What we mean by saturation drafting is when a team focuses most or all of their draft resources on one key position.

For example, let's say your favorite team had terrible wide receivers (Eagles fans from the McNabb era can probably imagine such a scenario). The team is actually pretty well situated at most positions, by WR is just a black hole. Saturation drafting would be that team drafting a ton of receivers, hoping that most will work out and you'll find an absolute stud somewhere in those picks.

It's an interesting idea, although the GM would need to believe a couple things:

- The team is really well situated at all other positions
- You can accept drafting for explicit need rather than best-player-available
- Drafting a bunch of players at one position is going to make us more likely to get good player(s) (either by maximizing draft probability or under the theory that multiple draft picks will help push each other to develop better/faster)

It seemed like something that a team may pursue, even if it's extremely unconventional. But I wasn't sure if it had ever happened and even less sure that it would actually work.

So I checked - gathering data from all NFL drafts since 2000 (thanks to Pro-Football-Reference).

I defined a 'saturation draft' as a year where a team drafted four or more players at a single positional group. Now, I generalized positional group a bit, in part because some positions can be interchangeable (e.g. tackles drafted can often play guard)

I looked across the following positions:

- DL
- LB
- DB
- OL
- WR
- QB
- RB
- TE

I ignored kickers/punters - for obvious reasons. But in the rest of those cases we looked across all the drafts since 2000 to see when, if ever, teams had used 4 or more picks on one position. The only exception I allowed was for quarterbacks - in that case - I set the saturation drafting threshold at 2 picks (no team since 2000 has ever drafted three QBs in a single year).

So - first thing - does saturation drafting ever actually happen???

Well, there are 512 team-draft seasons since 2000, across eight position groups. Since 2000 - there have been 34 instances where a team has taken four or more players at a single positional group (or two QBs).

On average it happens almost twice a year, at least the way I've defined it here. Below are those 34 instances broken down by position group:

DL - 11 times
DB - 10 times
QB - 8 times
OL - 3 times
LB - 2 times

So when teams do end up spending a bunch of draft picks in one area - it's typically on the defensive line, the secondary, or because you've taken two quarterbacks. I was a little surprised teams' have taken two quarterbacks eight times - because that was something that seemed pretty unlikely to me.

Now it's possible these teams were all pursuing best-player-available strategies, and the BPA just happened to be at the same position groupings multiple times. But I've always been skeptical that teams ever pursue pure BPA strategies. I think need is always taken into account to some degree - so many of these instances could very well be considered saturation drafting rather than inadvertent.

So, if that's the case, we then have to evaluate whether this strategy is actually effective.

Well - if saturation drafting WAS a better strategy, if it did help you improve your odds at finding/developing good players, we would expect the share of Pro Bowlers created through saturation drafting to be higher than just plan old drafting under other strategies. I'm using Pro Bowl designations because it's the easiest proxy for very good players and games started is a messy metric when evaluating drafts.

Across over 4000 draft picks, the overall odds of drafting a player who makes at least one Pro Bowl is 9%.

Across all the saturation drafting scenarios we identified, the overall odds of drafting a player who makes at least one Pro also 9%.

(And for the record, those totals are weighted based on number of picks in each round - because saturation drafting instances have disproportionate picks in the later rounds)

In total, saturation drafting doesn't seem more effective at taking/developing Pro Bowl players (some of you might quibble because of the first round data that suggests saturation drafting is better, 50% to 39%, but that's off a sample of only 12 picks and skewed by saturation draftees Andrew Luck and RGIII making the Pro Bowl)

With that data - I think it's very hard to argue that saturation drafting is an idea worth pursuing. It's not any more effective at leading to finding great players, and let me offer a couple other reasons why it's not a great idea...

1 - Throwing Good Resources After Bad: If a team is considering saturation drafting because a position is a glaring weakness, it's likely that the weakness is not by chance. What I mean by that is, it's likely the team has invested in players at that position (either through earlier drafts, free agency, or trades), and things haven't worked out. Well, that suggests that maybe this team, for whatever reason, can't successfully evaluate talent or develop talent at that spot. It's not impossible to imagine a front office that's very good at identifying/developing offensive lineman but terrible at defensive backs. That would set you up to consider saturation drafting -- but if you can't evaluate the position relatively well -- the last thing you'll want to do is invest disproportionate draft resources at that spot. Better to draft what you know and maybe pursue trades for proven NFL players at your positional weakness

2 - Playing Time is a Zero Sum Game: This is an even bigger practical limitation to saturation drafting. Imagine you're a GM in the 2012 draft with the first overall pick. You take Andrew Luck (good choice), and then in the 3rd round, you take Russell Wilson. Well, that's great - only unless you try some new offense no one's tried before, you probably aren't going to play two quarterbacks at once.

Quarterback is the most extreme example, but this would hold for all positions. There's only so many game snaps to go around, and only so many first-team practice reps, and only so much positional coaching time. If you draft five defensive backs like the Jaguars did in 2013, simply put, not all of them are going to be able to play. So even if you ARE able to hit on a bunch of saturation picks, you're not going to maximize that value.

But even if it turns out saturation drafting isn't a good idea, at least we've got some data on it and have looked at it objectively. Sometimes proving which ideas won't work is just as helpful as finding ones that do. And honestly, if the Eagles took 4 OL this spring, even notwithstanding everything I've said, I'd still probably be OK with it.

Appendix (Saturation Drafting situations by team since 2000):

No comments: