Evaluating Kicker Performance since the PAT Change

Main Takeaway: A probability-based model as in the one described here for pass rushers appears to offer a sound framework for the evaluation of kickers. 

METHODOLOGY

Sample Size Requirement: Minimum of 30 FG Attempts in 2015 and 2016 combined

Step I: Establish the league average success rate for each distance chunk (1-19, 20-29, 30-39, 40-49, 50+) to create a performance baseline. In order to do this, data on all field goals attempted from 2012 to 2016 was used. A downward adjustment was made to the raw league average success rate for 50+ yards given that above average kickers tend to take the vast majority of such attempts, biasing the raw success rate upwards.

1-19: 0.9950
20-29: 0.9700
30-39: 0.9100
40-49: 0.8000
50+: 0.5000 (down from 0.6200)

PAT: 0.9100 (assumed to be equal to the 30-39 success rate)

Step II: Impose a “uniform kicking distribution” on all kickers given that differences in kicking distribution between kickers can significantly muddy the waters when it comes to performance evaluation. The uniform kicking distribution took the following form:

1-19: 1 Attempt
20-29: 8 Attempts
30-39: 9 Attempts
40-49: 10 Attempts
50+: 5 Attempts

PAT: 40 Attempts

RESULTS

Top 5 in Points Above Expected per Season

1) Justin Tucker (BAL): 13.03 PAE
2) Ryan Succop (TEN): 10.55 PAE
3) Adam Vinatieri (IND): 10.17 PAE
4) Dan Bailey (DAL): 9.46 PAE
5) Josh Brown (FA): 7.74 PAE

Bottom 5 in Points Above Expected per Season

27) Dan Carpenter (FA): -3.85 PAE
28) Connor Barth (CHI): -3.90 PAE
29) Mike Nugent (FA): -4.38 PAE
30) Andrew Franks (MIA): -6.31 PAE
31) Randy Bullock (CIN): -7.87 PAE

Crafting a Contract for Adrian Peterson

Main Takeaway: A unique clause included in a 2009 MLB contract could have important implications for the types of agreements that clubs reach with injury-prone players.

For nearly a decade, few have embodied the concept of an explosive workhorse back like Adrian Peterson has. Entering the 2017 season, Peterson has averaged 266 carries, 4.9 yards per attempt and 11 touchdowns per campaign (excluding 2014 due to suspension).

However, the combination of a season-ending meniscus tear in 2016, an incredibly expensive 2017 club option and the availability of several cheaper and younger free agent running backs culminated in his recent release from the Vikings. Burdened in addition by high salary expectations, Peterson remains without a contract more than two weeks after the start of free agency.

On the one hand, age and injury risk are top of mind for teams contemplating the prospect of signing Peterson. Who would want to commit any substantial amount of money to a 32-year old running back who could easily have less than 40 carries in 2017, as he did last season?

On the other hand, one could argue that Peterson’s meniscus tear has had a disproportionate impact on the league-wide perception of his abilities. Despite his injury history, Peterson is likely to leverage the sterling reputation that he possessed heading into the 2016 season to command a contract structure entailing significantly more than bare bones guarantees. Ask Eric Wright about Peterson’s willingness to compromise.

In light of these conflicting stances, can a contract structure that suits the needs of both parties be devised?

A contract signed by a fellow Texan, in another sport, could hold the answer to this question.

In 2009, free agent pitcher John Lackey signed a 5-year, $82.5M contract with the Boston Red Sox. Despite having a minor history of arm injury, Boston was so concerned with the possibility of Tommy John surgery that they mandated that Lackey agree to the following clause: should he miss an entire season due to an elbow injury, the Red Sox would gain the option to extend Lackey at the league minimum for a sixth season. Both parties ended up happy: a brash Lackey got a rich long-term contract, and the Red Sox protected themselves from a lost season.

This unique clause would appear to be ideal for teams interested in signing free agents with significant injury history, such as Adrian Peterson. Below is an example of a possible Peterson contract that could suit the needs of both club and player:

  • Term: 1 Year
  • Guaranteed Base Salary: $4M
  • Games Played Bonus: $100K per game
  • Clause: If Peterson spends 10 or more weeks on either the Regular Season PUP List, the Injured Reserve List or a combination of the two, the club will acquire an option to extend Peterson’s contract at the league minimum salary for the 2018 season

General comments:

  • The key tradeoff is between guaranteed money and exposure to a lost season
  • While the contract does not guarantee that Peterson will be productive, it does alleviate the team’s most pressing concern (exposure to a lost season)
  • The amount and guaranteed nature of the salary may appear rich at first. However, it is important to remember the potential caliber of Peterson’s performance and the fact that he does face a substantial risk of triggering the club option given his injury history.

What are your thoughts on this type of contract structure?

A reasonable argument could be made that the club option would have limited value should Peterson get injured in 2017. After all, he would have missed the majority of two straight seasons heading into 2018. While I agree with this line of reasoning, the type of contract discussed above would still have the potential to yield much more value to a team, in the event of a Peterson injury, than a standard one-year contract entailing no control over Peterson after 2017.

Shorting Salary Cap Conservatism

Main Takeaway: Unorthodox strategies are key to allowing teams to more efficiently manage two of their most scarce resources, cap space and draft picks.

Building on the piece that I wrote about the draft pick value of cap space, I sought to come up with a couple of under-explored trade scenarios that creative teams could employ to either minimize their salary cap obligations or maximize their draft capital.

I) “Sign and Trade”

  • Team Red could sign a player to a one-year contract featuring a $3M signing bonus and a $7M guaranteed base salary
  • In this case, the cost of acquiring the player would be equal to $10M in cap space
  • Team Red could also ask Team Yellow to sign the same player to a one-year contract featuring a $3M signing bonus and a $7M guaranteed base salary, and then trade a draft pick to Team Yellow for said player (the player would know prior to signing with Team Yellow that his ultimate destination would be Team Red)
  • Under such a scenario, Team Yellow would retain the signing bonus obligation on its cap, and Team Red’s cost of acquiring the player would be equal to $7M in cap space and the value of the traded draft pick
  • For a contending team with limited salary cap availability, the latter scenario would be appealing as it would enable Team Red to acquire a “$10M player” at a salary cap discount in exchange for draft pick compensation
  • Engaging in the “Sign and Trade” would also benefit a rebuilding team with excess cap space and a desire to add more draft assets

II) “Injury Guarantee Play”

  • Team Blue has an expendable player on its roster with one year left on his contract and whose base salary is not guaranteed
  • Under normal circumstances, Team Blue would simply cut the player
  • However, the player is currently injured and his contract features a significant injury guarantee of $20M which is triggered if he is still injured on “Day X”
  • At the beginning of the offseason, the probability that the player will be healthy enough to pass a physical on “Day X” is estimated to be near 50%
  • Team Blue could either accept to face this significant exposure, or seek a trading partner to eliminate it
  • In the latter case, Team Green would be willing to acquire the player (and take on the substantial financial risk related to his injury) in exchange for appropriate draft pick compensation
  • For a contending team with limited salary cap availability, this trade scenario would be enticing since it would enable Team Blue to completely eliminate its exposure to a potentially calamitous salary cap event
  • The “Injury Guarantee Play” would also serve the interests of a rebuilding team with excess cap space that is seeking to bolster its draft capital

P.S. Thank you to the 49ers for critical feedback

Expected Sacks

Main Takeaway: By capitalizing on a larger sample size, a probability-based metric may offer a much more accurate picture of a pass rusher’s performance than traditional sacks.

Although they have not yet found a solid foothold in football, probability-based metrics are popular across other major sports. In short, such metrics involve multiplying a relatively high frequency event (such as a shot) by its probability of success based on a given variable (such as location) in order to yield an “expected success total” (such as expected goals). In soccer, the infrequency of goals created a demand for a metric based on a larger sample and expected goals provided a natural solution.

A natural application of probability-based metrics to football involves pass rushing, given the infrequency of sacks. Expected Sacks would involve two components:

  • Knockdowns broken down by time elapsed between snap and knockdown (“K”)
  • While some may justifiably prefer Hurries to Knockdowns due to their greater sample size, I opted for the certainty that the pass rusher made contact with the quarterback and took him to the ground.
  • Probability of a quarterback holding onto the ball broken down by time elapsed since snap (“P”)
  • As is customary with probability-based metrics, K would be on an individual player basis while P would be based on league-wide performance on passing plays

Expected Sacks would be computed by (I) multiplying a player’s K and the respective P for each time bucket and (II) adding up all of these expected values. Expected Sacks could then be divided by the pass rusher’s number of pass rushes on the season to produce a much more comparable rate metric in Expected Sacks per Pass Rush.

Applied Example (Theoretical Data)

Von Miller (2017)

  • Pass Rushes: 320
  • Knockdowns (1 Second After Snap): 1
  • Knockdowns (2 Seconds After Snap): 7
  • Knockdowns (3 Seconds After Snap): 3
  • Knockdowns (4 Seconds After Snap): 6
  • Knockdowns (5 Seconds After Snap): 2
  • Knockdowns (6 Seconds After Snap): 2
  • Probability (QB Holding Ball 1 Second After Snap): 95%
  • Probability (QB Holding Ball 2 Seconds After Snap): 85%
  • Probability (QB Holding Ball 3 Seconds After Snap): 75%
  • Probability (QB Holding Ball 4 Seconds After Snap): 50%
  • Probability (QB Holding Ball 5 Seconds After Snap): 20%
  • Probability (QB Holding Ball 6 Seconds After Snap): 5%

Expected Sacks

= 1*0.95 + 7*0.85 + 3*0.75 + 6*0.50 + 2*0.20 + 2 *0.05

= 12.65

Expected Sacks per Pass Rush

= 12.65/320

= 0.0395

Should I Trade the 30th Pick for $8M in Cap Space?

Main Takeaway: A pick for cap space trade can definitely benefit a team if the freed-up cap space can be redeployed towards a much more productive player.

How should teams go about determining the relationship between draft picks and cap space? One possibility lies in surplus value. A measure developed by Professors Massey and Thaler in their seminal “Loser’s Curse” paper, surplus value represents “the difference between what you pay for performance and what you’d be willing to pay for performance” as Brian Burke so aptly put it.

The method I propose to assess the desirability of a pick for cap space trade would simply involve converting both assets into their surplus values and then comparing them. One half of the puzzle has already been solved, as Mr. Burke’s research has yielded surplus value figures for every pick in the draft under the new CBA. Determining the surplus value of a given amount of cap space is a much more tricky endeavour. A possible approach involves a team’s current cap structure and free agency.

A team would identify an expendable player on its current roster with a fully guaranteed salary. It would simultaneously identify a desirable free agent that could be signed for the same annual cap hit as the expendable player. Both players’ surplus values over the entirety of their respective contracts would be compared and the difference would effectively represent the surplus value of that cap space.

Below is a theoretical (read: DePodestian) example of the method proposed:

– Player Beta’s contract projects to provide Team Blue with $2M in surplus value
– Team Blue currently has $8M in cap space committed to Player Beta

– Team Blue has the option of making a trade with Team Orange: the latter would acquire Player Beta (and his $8M cap hit) and the 30th pick in the draft
– With this newly freed-up $8M, Team Blue could sign free agent Player Alpha to a contract which would be anticipated to provide a surplus value of $12M
– The 30th pick in the draft has an estimated surplus value equal to $8.5M

– Since the upgrade from Player Beta to Player Alpha gives the team an extra $10M in surplus value ($12M – $2M), Team Blue should execute the trade for any pick with a surplus value under $10M
– Therefore, Team Blue should deal the 30th Pick to Team Orange for $8M in cap space

Clearly, the proposed approach is highly simplified. Important considerations in a deeper analysis include:

  • The importance of Player Beta and Player Alpha’s exact contract structure. While Player Alpha is projected to provide greater surplus value, completing the trade could be significantly less appealing to Team Blue if Player Beta only has one year left on his deal and Player Alpha would command a 5-year, $40M 70% guaranteed contract. Two measures that could certainly shed more light on the attractiveness of the deal are surplus value per contract year (Player Beta = $2M, Player Alpha = $2.4M) and surplus value per dollar of salary (Player Beta = 0.25, Player Alpha = 0.30).
  • The difference in the value of a given amount of cap space based on the team’s total available cap space ($8M is a lot more important when a team has $0M left than when it has $50M at its disposal)
  • The changing value of a given amount of cap space over time due to increases in the salary cap

That being said, the model does provide a foundation for an intelligent analysis and illustrate a critical reality: the value of a given amount of cap space is not constant, but entirely dependent on what can be done with it.

P.S. Thank you to the Broncos, Falcons and Seahawks for critical feedback.

Framework for Evaluating Receivers

GENERAL COMMENT: While I believe in a 50/50 split between data and video to assess a player’s performance, this post will primarily focus on the former.

The Dolphins’ re-signing of Kenny Stills for $32 M understandably left some puzzled. While thinking about the contract, I began to contemplate the type of framework teams should use when evaluating a wide receiver. Below is an attempt at a starting point:

CONTEXT

I) Quarterback Skills: Not only will the overall level of ability of the quarterback affect the receiver’s performance, but the extent to which their strengths match is also crucially important. Brian Hoyer is an objectively decent quarterback, but pairing him with Torrey Smith would be disastrous for the wide receiver.

II) Offensive Line Pass Blocking Performance: If an offensive line performs poorly by providing little protection time to its quarterback, the latter will surely throw many more short passes and many fewer medium to long balls than he would have desired. This outcome must be adjusted for since it unjustly benefits short pass receivers and penalizes medium to long ball receivers.

III) Running Game Performance: A high-performing running game will typically force the defense to load up the box, significantly facilitating a wide receiver’s task of getting open.

IV) Opposing Defense Performance: Not only will a strong secondary affect a receiver’s performance, but a high quality pass rush will have similar effects on his output to those of poor offensive line pass blocking.

V) Double Coverage% (Double Covered Routes/On-Field Passing Snaps)

VI) Routes Breakdown by Starting Position (Outside or Slot) and Support (Island or Other Receiver Near): A deep threat has the ability to stretch an opponent’s defensive coverage and provide a receiver that lines up next to him with greater openings than if he were on an island.

VII) Routes Breakdown by Distance (Short, Medium or Long): Since long passes are correlated with lower Reception%, an adjustment for route distance must be made so as to not unjustly penalize deep threats.

RECEIVER PERFORMANCE

I) Target% (Targets/On-Field Passing Snaps): Merely being targeted is a decent indicator of success since it implies that the Quarterback determined that you were the best receiving option on the play.

II) Reception% (Receptions/Targets)

III) Drop% (Drops/Targets): Although Drop% represents an imperfect gauge of a receiver’s responsibility for an unconverted target (not reaching the spot of the catch in time is not recorded in Drop%), it does at least provide the evaluator with a baseline for receiver fallibility.

IV) Average Yards per Target (Yards/Targets)

V) Average Yards After the Catch (Yards After the Catch/Receptions)

VI) Avoided Tackles% ((Broken + Missed Tackles)/Total Tackle Attempts Against): While both broken and missed tackles represent positive outcomes for the offense, the latter are much more desirable since they do not typically involve as significant a decrease in the player’s forward momentum as the former. Generally, a team would also prefer that a player avoid contact as a means of maximizing his health.

VII) 1st Down%: While somewhat controversial, 1st Down% can potentially capture a receiver’s ability to “mold his skills to the situation” in order to convert a first down. A player’s typical receiving location in relation to the first down marker should be kept in mind, since receiving a dump off 7 yards short of the market versus a pinpoint pass 1 yard short represent two very different situations.

VIII) Role in Passing Game over Time (Y2 Passing Snaps% – Y1 Passing Snaps%), where Passing Snaps% (On-Field Passing Snaps/Total Team Passing Plays): A team that sees a player’s performance on an everyday basis has a clear information advantage over one that does not. However, a potential signal of player development does exist for the latter: an increase in a player’s participation in his team’s passing snaps relative to the prior season. Once again, key caveats exist: this measure could easily be jeopardized by a coaching change or an injury to a receiver higher-up the depth chart (leading to increased snaps for our receiver only out of necessity).

Which measures do you think are crucial when evaluating a wide receiver?

Situational Pass Blockers

Main Takeaway: A team could improve its pass protection in obvious passing situations by ensuring that its best pass blockers, whether they are starters or not, are on the field.

Situational pass rushing appears as common in today’s NFL as a Joe Thomas tirade against Roger Goodell. Since the Giants brought the concept back into vogue with their vaunted Nascar Package, teams have keyed in on the advantages of matching up rested pass rushers against fatigued offensive linemen in crucial passing situations. Although the former are typically of lesser overall ability than the latter, the strategy relies on defensive players’ freshness and high-level rushing ability to tip the competitive balance in their favor. More broadly, it allows teams to simultaneously maximize their on-field pass rushing ability and rest their exerted run stoppers. Pernell McPhee, the patron saint of situational pass rushers, is a testament to the role not only benefiting teams, but also players.

Thinking about this phenomenon recently, a question came to mind: why have teams not transferred this thinking to the offensive side of the ball and begun using situational pass blockers?

In obvious passing situations (where trotting out your best pass blockers would not tip your play selection to the opponent), offensive teams should logically field offensive lines that maximize their pass blocking to counter defensive teams that clearly maximize their pass rushing.

A few potential drivers of this substitution asymmetry include the following:

I) Teams have concerns about the abilities of “cold” offensive linemen who enter the game for a handful of spaced out plays. However, this “cold” phenomenon applies equally to situational pass rushers, who are frequently used by NFL teams.

II) With the prevalence of passing in the modern game, it is entirely possible that a team’s starting line could also feature its best pass blockers, thus leading to an absence of situational pass blockers. That being said, it is very hard to believe that this scenario holds true for all 32 teams in the NFL.

III) A much more collective action than rushing, pass blocking requires significant cohesion between linemen and teams are concerned that situational pass blockers may disrupt it. Assigning a single offensive line position to a situational pass blocker (e.g. RT), and devoting a certain amount of team practice snaps to the “optimal pass blocking line” so that it can develop cohesion, appear to offer remedies to this important issue.

IV) Teams believe that it is more efficient to protect their quarterback by supplementing their starting line with a blocking running back or tight end than by substituting in situational pass blockers. Despite this line of thinking initially appearing to hold significant weight, one must keep in mind the following: doesn’t sacrificing one to two potential receivers in a long-distance passing situation represent a very high cost to bear?

While this analysis clearly represents an elementary exploration of the issue, teams would definitely benefit from experimenting with situational pass blockers.