A few thoughts on the NFL Draft

0205bertbell-459x395

Bert Bell — the father of all drafts (and specifically the father of the NFL draft)

As a Lions fan – a team that generally picks early in the NFL draft – the draft tends to be the highlight of the NFL year. And since most readers of the WoW  Journal are NBA fans, I thought I would offer a few thoughts on this event for those who primarily watch bouncing basketballs (and not flying pigskins).

The NFL draft has seven rounds, and now takes place across three days. To help put this event in perspective – for an NBA fan – the following chart is an effort to convert each pick into its NBA equivalent.

NFL Pick

Round

Equivalent NBA Pick

Round

1-22

1st

1-5

1st

23-44

1st-2nd

6-10

1st

45-66

2nd-3rd

11-15

1st

67-88

3rd

16-20

1st

89-110

3rd-4th

21-25

1st

111-132

4th-5th

26-30

1st

133-154

5th-6th

31-35

2nd

155-176

6th

36-40

2nd

177-198

6th-7th

41-45

2nd

199-220

7th

46-50

2nd

221-242

7th

51-55

2nd

The basic idea behind this chart is that NFL teams have 22 starters while NBA teams have five. So therefore, the first 22 picks in the NFL draft are essentially the same as the first five picks in the NBA. And following this reasoning, the NFL is still making NBA-first-round picks into the 5th round.

The NBA draft – as we have noted in this forum – is not entirely predictable. What we see in college doesn’t always translate into the NBA. However, performance in college is statistically related to what we see in the professional ranks. So when teams draft players who did not perform well in college (such as Austin Rivers), we should ask questions.

Can we do the same for the NFL? Although there are many mock drafts and plenty of people providing draft grades immediately after this event concludes, evaluating college talent in football is amazingly difficult. And this difficulty can be seen by just considering how hard it is to forecast the performance of veterans who are already in the NFL.

Brian Burke – of Advanced NFL Stats — and I published an examination of various measures of performance in the NFL. Within this examination we noted the inconsistency of NFL quarterbacks. For example, for the following aggregate measures of performance, here is how much of current performance is explained by last season’s performance:

  • Success Rate: 29.0%
  • Expected points per play: 21.0%
  • Wins Produced per 100 plays: 16.9%
  • NFL Quarterback Rating: 15.0%
  • Wins Probability Added per play: 11.7%

And here is what we see when we look at the components of the NFL’s quarterback rating.

  • Completion percentage: 31.1%
  • Passing yards per attempt: 22.1%
  • Touchdowns per attempt: 10.1%
  • Interceptions per attempt: 0.6%

Turnovers are a big part of football. And as one can see, these are very hard to predict (a similar story can be told for fumbles).

Now contrast these results with what we see in baseball and basketball. In Stumbling on Wins, Martin Schmidt and I noted that you see much more consistency in the other major sports. For example, in hockey 80% of a skater’s shot on goal (per-minute) this season are explained by a skater’s shots on goal the previous season (but for goalies, only 6% of a goalies current save percentage is explained by last year’s save percentage). And in baseball (this results was originally reported by JC Bradbury), 43% of a hitter’s OPS this season is explained by the same player’s OPS last year (pitchers, though, are not quite as consistent with respect to popular metrics like ERA).

And when we turn to basketball, we see that for many measures, about 80% of what player does this season is explained by what he did last year. In sum, with the exception of goalies and pitchers, veteran  performance in the other major sports is somewhat consistent. In the NFL, though, veteran quarterbacks are hard to predict (and in Stumbling on Wins, similar inconsistencies were reported for running backs).

Now turn to the draft. In the NFL draft, decision-makers are not just forecasting what a player who has already been in the league will do next year. They are trying to forecast what a player who has never faced NFL competition will do in the future. Such a forecast requires a decision-maker to not only assess the individual player’s abilities, but also how those abilities will interact with the other players on the roster in the future. For many positions, we don’t have great measures of individual performance. And because the interaction effects are likely large, it is hard to interpret what these measures for the individual really mean.

Further complicating these forecasts are the high rate of injury we see in football. And that means that even if you nailed the forecast of ability and fully understood all the interaction effects, player injury can transform a great pick into the perception of a bust.

Starting tonight, though, we are going to hear over and over again how each drafted player is going to make a difference for their new team. And of course, similar predictions have made about drafted players in the past. But as Cade Massey has noted (see the discussion at Advanced NFL Stats), “randomness dominates the process.”

So much of what we hear this week will not be consistent with what we observe in the NFL in the future.

Of course, that is not going to stop fans – especially fans of the Lions (who really live for the draft) – from being excited. In other words, I just know that the Lions picks are going to be great.  And with the help of the players the Lions are drafting, 2013 is really going to be a great year for the Lions.

Hey, if we can’t predict well, why not just predict the best? I have been doing this for decades, and look how that has worked out for this Lions fan!

- DJ

Comments are closed.