Punch Score: Identifying the players that Speed Score doesn't

Every year, armchair analysts declare which running backs are bargains and which will be busts, based solely on one number: Speed Score.

While Speed Score enjoys moderate success - it has successfully identified bargain backs like Brandon Jacobs, Arian Foster, and Rudi Johnson, and issued warnings upon NFL duds Mark Ingram and Trung Canidate, among others - it both misses on several running back archetypes and has a shaky mathematical foundation (more on this shaky mathematical foundation in future posts).

I don't intend on quote-shaming, but here's a quote from Grantland's football resident-expert, Bill Barnwell, from when he unveiled Football Outsider's Speed Score back in 2008:

The raw numbers for the bench press, broad jump, shuttle drill and three-cone drill bear no relationship whatsoever to the future NFL success of a running back. Scouts might see things in a player's stride and cuts that fans can't, but if somebody's quoting a three-cone drill number for a halfback, it's relatively irrelevant.

I am about to show you that what that statement implies is complete rubbish: a player's bench press and broad jump are most certainly relevant, usable statistics - and that they help identify the players that Speed Score fails to identify.

Those figures from the NFL combine are made useful in two new, exclusive-to-BFTB metrics: Punch Score and Foot Speed.

Punch Score

To begin with, the whole concept of the Speed Score is to measure how well a player is making the best of use of his body, in terms of speed. The analogue here is that Punch Speed is measuring how well a player is making the best use of his weight, in terms of muscle.

The Math!

Punch score approximates the ratio of muscle for a given player, in affect measuring the punch of each pound (That's where this gets its name).

This is accomplished by measuring the z-score of a player's bench press and subtracting the z-score of his weight.

What is z-scoring? Basically, z-scoring takes everyone's measurements and scales it such that players who are above average in a given quality have positive values while those who are below average have negative values. By definition, the mean of all the players is zero.

For a more detailed explanation of z-scoring, search for it on Google or Yahoo! or Bing or something. It's a very common mathematical procedure for use with a normally distributed population dataset, which we have.

By z-scoring, we are expressing that attribute in terms of rarity, and can therefore compare the rarity of one player's 40 time versus the rarity of another player's vertical leap. We could then say "Chris Johnson's 40 time is more impressive than LaDainian Tomlinson's vertical leap" with the mathematics to back it up, by citing their z-scores in each statistic.

Note: because this statistic is going to supplement (a z-scored) Speed Score in a predictive model, I have only included players who participated in both the 40 yard dash and the Bench Press for consideration. Speed Score and Punch Score may have value on their own, but I'd prefer to focus on players who we can most accurately grade.

What that means

Basically, what we're doing with Punch Score is comparing the rarity of that player's bench press against the rarity of that player's weight.

A player who is, for lack of a better term, jacked will have a higher z-scored bench press than his z-scored weight. So, when you subtract the z-scored weights from the z-scored bench press figures, the players with the highest values are the most jacked - the ones with the lowest values have a lower percentage of upper-body muscle mass.

The reason Punch Score is important is because players who score high in this statistic have a justification for a low Speed Score - they may have exchanged some speed for bulk, upper body muscle. We should take this into consideration.

Top Punch Scores

Remember how we all used to talk about how jacked Darren Sproles was, and that he lifted more than most of us could ever dream to? Meet the math to prove it based on just his NFL Combine numbers: his Punch Score is absurd.

Player	Year	Bench	Weight	Punch Score	Speed Score
Lorenzo Booker	2007	26	191	3.06	96.54
Shane Vereen	2011	31	210	2.83	103.34
Darren Sproles	2005	23	187	2.66	93.68
Brian Westbrook	2002	26	200	2.41	91.71
Alonzo Coleman	2007	27	207	2.13	101.86
Kendall Hunter	2011	24	199	2.02	100.59
Leonard Henry	2002	26	206	1.97	94.46
Jordan Todman	2011	25	203	1.96	108.32
Thomas Brown	2008	25	204	1.89	98.62
Ray Rice	2008	23	199	1.80	104.28

It reads almost like a who's who of guys that Speed Score failed to identify. Darren Sproles, Brian Westbrook, and Ray Rice are obvious home runs, Shane Vereen looks like he's going to be a home run, and the jury is still out on Todman - although he looked good to end 2013.

And among those misses:

• Lorenzo Booker was plagued by both injuries and fumble concerns, and ultimately left the NFL due to both (a concussion and a fumble-fest in Minnesota).
• Alonzo Coleman went undrafted before retiring with knee issues after being activated from the Cowboys practice squad just once.
• Leonard Henry was a seventh round pick who, in the absence of much opportunity, went and played in NFL Europe, earning Offensive Player of the Week honors in his first game before suffering a season-ending groin injury. Henry came back to the NFL and earned three starts with Miami - rushing for 137 yards on 44 carries - before being passed on the depth chart by Sammy Morris.
• Thomas Brown lost his rookie season to a horse-collar-tackle-caused injury in preseason. Brown would bounce around practice squads once he returned, but retired having never played.

I don't know about you, but that's not a very incriminating miss list. None of those players were high selections and only one of them, Henry, ever received a true, injury-free opportunity - and a short one at that. The bottom line is that every list of eleven drafted running backs will have players on that list who never get going because of injury. None of these guys were monumental busts or a significant waste of resources.

Sign me up for Punch Score.

Bad Punch Scores

Before you get the list of the lowest Punch Scores, it's important to contextualize what Punch Score is supposed to do: basically, tell you whose weight is comprised of a higher percentage of upper body muscle. So, we'd expect the lowest score guys to be, for lack of a better term, the fatter guys.

This doesn't make those fat guys undraftable - they could still be really fast for their weight, which we already know is a positive quality - but examining those with low scores does serve to validate whether or not our score is doing what it's supposed to.

The bottom Punch Scores:

Brandon Jacobs, Gary Russell, Dicenzo Miller, Correll Buckhalter, Roy Helu, Anthony Dixon, Devin West, Dwayne Wright, Duron Croson, Michael Basnight, and LaGarrette Blount.

You're excused if you've never heard of Miller, West, Wright, Croson, or Basnight. They played in a combined 26 games - and three of them converted to fullback.

The others - Jacobs, Russell, Buckhalter, Helu, Dixon, and Blount - are mostly players who can (very) accurately be considered fat running backs.

In other words, success on this end, as well.

What's next?

Punch Score doesn't do everything on its own. Neither did Speed Score. However, when you take Punch Score in tandem with Speed Score, you have a really powerful pair of metrics.

I will present the predictiveness of this metric, along with Z-Speed Score and Foot Speed, in the final post putting this altogether.

For now, discuss! And stay tuned for the next post, Foot Speed, another metric which adds predictive information.