Visualization of all NFL Kick Returns from 2015

In the interest of seeing how recent changes to the NFL kickoff rule may change the return game, I acquired play-by-play data of all NFL kickoffs from 2015.  After excluding touch-backs, returns with penalties, onside kicks with no attempted return, and end-of-game laterals, we were left with 1040 kicks returns, which I sorted by starting position then yardage to plot each return.  Each line above corresponds to a single return from the starting point (catch) to finish (end of return).

I'll analyze this data in greater depth in the near future, but some observations from the figure:
-Approximately half of the kicks returned started in the end-zone
-Kicks 9 yards deep in the end-zone never made it past the 25 yard line last year.
-Returns for touchdowns are shaded toward better starting position(-1 or better), but only slightly.

Adjusting the Dalton Scale

Andy Dalton has long been identified by NFL analysts as the dividing line between passable and non-serviceable quarterbacks in the league.  However, some stellar play from Dalton in the past year may suggest that he has surpassed his own standard as the definition of QB mediocrity (though it should be noted that NFL mediocrity is something to which many QBs aspire and have never attained).

To this end, let's break down Dalton's numbers from the most recent season and compare them to his past statistics.  Using data from Foxsports.com and a little mathematical manipulation (addition and subtraction to ensure mutually exclusive events), we arrive at the following table:

Years	Incomplete, INT	Incomplete, no INT	Complete, < 20 yds	Complete, 20-39 yds	Complete, 40+ yards
'11-'14	66	788	1127	130	44
2015	7	135	203	41	11

While the counts of events is useful for transparency and ensuring these results can be reproduced, it is not very helpful in conducting an apples-to-apples comparison of these events, which we'll accomplish by way of reducing these numbers to relative percentages before displaying them:

The associated chi-square statistic (chi-sq=12.73, df = 4, p-value = 0.013) that tests if these two distributions are statistically different was significant (p < 0.05), meaning that it unlikely that random variation is driving the observed differences here.  The most notable difference is the 4.3% improvement of completions in the 20-39 yards category that actually contributes the majority (72.5%) to the significant chi-square statistic of 12.73.  So perhaps we can conclude that Andy Dalton's improved deep ball has moved him past the status of a merely passable signal-caller.

Still, I'm skeptical that this comparison is for a full 16 games from 2011-2014 but only the  first 13 games of 2015.  As such, let's explore if this may be a confounding factor in the previous analysis.  Just to vary our methods a little, let's compare Dalton's passer rating from the first 12 games of every season, as he was actually injured early in the 13th game:

	Games Played	Avg QBR	SD of QBR	p-value of t-test
2015	12	111.6	26.8	0.007
'11-'14	48	86.2	28.2

So the observed difference of 25.4 (±18.4 for 95% Confidence interval) in Passer Rating in 2015 pushes Dalton well past the mediocre, even after adjusting for potential differences in early season performance.  So we'd still conclude that Andy Dalton played significantly better in 2015 than in 2011-2014 using Passer Rating.  While this 2015 performance may still turn out to be an anomaly, perhaps we should start looking for a new standard of merely passable quarterback in the NFL.

Weapons of Mass Destruction: tight-end depth charts

While the media was buzzing in a pre-NFL draft frenzy, Ozzie Newsome was calmly and quietly stockpiling weapons of mass destruction by adding a couple tight-ends to the roster. The most notable addition was wooing the 35 year old Ben Watson away from the Saints, with the second most intriguing prospect being Darren Waller, a converted wide-out from Georgia Tech, who gained 15 pounds in the off-season.

A retrospective analysis of combine participant tight-ends suggested that four key variables interact to affect size adjusted athleticism in terms of the High Point Potential (Height + Vertical) and Average Momentum in the 40 yard dash (Weight/Dash).  These two factors were found to strongly correlated with pro-bowl selection in the data.  While this model is still being validated to confirm its predictive value, I'm optimistically christening it with a convenient and appropriate acronym, BEAST, for Better Estimate of Athleticism by Size Transformation.  Here is the current state of the Raven's already deep TE rotation (depth chart courtesy of Rotoworld).  Athleticism is presented in terms of 40 yard dash and vertical leaps alongside height and weight.

Name	Age	Height	Weight	Dash	Vert	High-Pt	Avg_P	BEAST
Ben Watson	35	76	258	4.57	35.5	111.5	56.5	17.6%
Crockett Gillmore	24	78	270	4.89	33.5	111.5	55.2	9.9%
Maxx Williams	22	76	250	4.78	34.5	110.5	52.3	1.8%
Dennis Pitta	30	76	245	4.72	34	110	51.9	1.3%
Darren Waller	23	78	260	4.46	37	115	58.3	58.0%

The aforementioned retrospective data analysis suggested that 6% is the optimal cut-point differentiating eventual pro-bowl caliber athletes in the tight-end position from their more work-a-day counterparts.  By this criteria, it looks like the Ravens roster has 3 potential pro-bowlers!  However, some caution should be taken with Gillmore, as the 10 pounds he's gained since joining the league is the difference between exceeding 6% or not, and we don't have updated data on his vertical leap or dash-time since gaining this weight.  A similar caveat may apply to Darren Waller, though his score far exceeded the 6% cutoff (~18%) before the 15 pounds he added recently.

Baltimore wasn't the only city stockpiling weapons of mass destruction.  The Patriots acquired Martellus Bennett from the Bears and signed Clay Harbor, the free agent out of Jacksonville.  The year before (2015) they made a similar acquisition of converted offensive tackle, Michael Williams.  Based on these recent moves, here's the current state of the Patriots depth at TE:

Player Name	Age	Height	Weight	Dash	Vert	High-Pt	Avg_P	~Prob(PB)
Rob Gronkowski	26	78	264	4.68	33.5	111.5	56.4	17.2%
Martellus Bennett	29	79	273	4.68	34	113	58.3	45.9%
AJ Derby	24	77	255	4.72	?	?	54.0	?
Michael Williams	25	78	304	5.19	25.5	103.5	58.6	8.0%
Clay Harbor	28	75	250	4.69	40	115	53.3	8.8%

It looks like the Patriots may roster as many as 4 potential pro-bowler tight-ends based on the previously established 6% rule.  Again, these numbers may be somewhat misleading, as the computations driving the estimated pro-bowl probability are built upon combine numbers where Bennett and Williams have both put on significant weight since joining the league.  In both cases, this weight made the difference in exceeding the 6% threshold, again based upon the sketchy assumption that weight gain didn't affect speed and jumping ability.  In Bennett's case, his estimate is so high that some sensitivity analysis suggests that he's likely over the 6% threshold even if both his speed or vertical has suffered as much as a 5% reduction (Update: Michael Williams suffered a torn ACL in OTAs in June and is still waiting for swelling to reduce before attempting corrective surgery).

In conclusion, the Ravens and Patriots appear to be stacked in the tight-end position.  However, due to non-synchronicity in the timing of available data, it may be only the top and the bottom of the depth chart that exceed the 6% threshold for both teams.  This is a less exciting prospect, as holding a potential top 10 tight end with an athletic project tight-end waiting in the wings is pretty much the goal of every organization.