Jairus Byrd’s Contract Situation, Part 3

**Ed note: This is part three of a five part series analyzing the Jairus Byrd contract situation. For part one, click here. Part two can be found here.

Now that we have a method to determine salaries for safeties based on their expected production as well as an aging curve to estimate that production, we now must determine the term of the agreement the Bills should offer Jairus Byrd.

The NFL’s Collective Bargaining Agreement, as negotiated before the 2011 season, doesn’t explicitly put much restriction on the length of a free agent’s contract. However, portions of contracts extending beyond the end of the “Final League Year” (2020, as definited by the CBA) may count towards the 2019 and 2020 team salary. No player is currently contracted beyond 2019 (according to Spotrac.com), probably because the salary cap figures for 2019 and 2020 are not yet known. (Annual salary caps are determined based on a formula determined by the CBA, it can be found here: NFL CBA.)

Ten safeties signed new contracts for more than one year this offseason (Goldson, Quin, Moore, Reed, L. Landry, Chung, Huff, Wilson, D. Landry, and Sanford). Those players will be just over 29 years old next season, on average, and signed for an average length of 3.5 years. However, the average age of the players at the time when the contracts expire is more telling. Six of the ten players will be either 32 or 33 once these deals expire.

Jairus Byrd won’t be 32 until 2018. Keeping him for that long would require a six year deal. Only Kyle Williams and Mario Williams have contracts that long. Kyle signed before the 2011 season and Mario signed before the 2012 season. Could Byrd be the third six-year deal for the Bills defense in consecutive seasons?

If we compare Byrd’s expected production (as measured by EPA+ again) to the top safeties over the past six years, we can get an idea of the type of player the Bills might get every season. The graph below compares Byrd’s expected production level to the average production needed to be a top-five or top-ten safety in the NFL.

Based on the graph above, Byrd should be a top-ten safety through the 2017 season. That would equate to a five-year contract, not unlike Goldson, Quin, and Moore. If Byrd is playing better in 2017 than the regression currently estimates, he could earn himself another large contract and become the next Ed Reed, who was a top-fifteen safety at 34 years old.

A five year deal for Byrd should be worth somewhere around $32.3 million to Byrd (based on the table in the last post). That’s a total cap hit amount, not accounting for signing bonuses, workout bonuses, and other incentives. Would it be enough?

That might be a hard sell, since Dashon Goldson signed a five-year agreement worth $41.25 million earlier this spring. It seems that the Buccaneers overpaid quite a bit for a player with similar production as Byrd but is two years older. Buffalo’s safety is undoubtedly looking for more.

Next time, we’ll analyze how much the team will need Jairus Byrd going forward.

Jairus Byrd’s Contract Situation, Part 2

**Ed note: This is part two in a five part series analyzing the Jairus Byrd contract situation. For part one, click here.

Last time, we examined about how much safeties, namely Jairus Byrd, should make based on their historical production. In this post, I’d like to forecast Byrd’s future play based on his age so we can then apply the polynomial regression determined last time to his expected production in order to find annual salaries Byrd should receive.

To get an overview of how safeties age, I first compiled a list of the top fifty safeties for the past six seasons, based on their EPA+. 78% of the 300 players were under 30, and only 3% of the players were over 35. The distribution of the players by age is in the graph below.

The drop off after age 27 is pretty staggering, but it makes sense. The top fifty safeties each season make up 78% of the 64 starting safeties in the NFL and any player below that fiftieth ranking could be considered replacement level. Once a player slips below that level, they can be replaced by a younger player.

In order to determine an aging curve for safeties, I plotted the maximum performance for each age over the past six seasons. I limited the age range to be from 22 to 33 because those age groups had at least five data points each. From that data, I only need to correct one data point: Ed Reed’s 30 year old season in 2008. He was ridiculously good that season and I determined his 2008 performance was an outlier for any aged safety.

Interestingly enough, the distribution resembled a parabolic shape, which is logical. As a player gets more experience and playing time, he gets better until his body starts to break down or he loses a step. That loss in ability happens quickly, as depicted by the age distribution above.

The safety aging curve, as measured by maximum performance is in the graph below with a polynomial trend line (black line) and the Ed Reed outlier (in red) is below.

The graph and trend line above compared maximum performances for each age group, which is independent of any one player. Because of that, it resembles the greater population with relative accuracy. The trend line predicts a player to have his peak season when he is 26, which matches our age distribution for the top 300 players.

Byrd was 26 last season. In fact, his 26 year old performance was the best of any other 26 year old safety in the last six seasons. What can we expect him to do next season? How quickly will his play diminish?

Because the trend line estimates maximum value for each age group, I converted the expected maximum EPA+ values into percentages of the largest value (the regression’s total when age is 26). The table below breaks down the expected EPA+ percentages of a player’s 26 year old season, by age.

Age	% of Age 26 Season
23	97%
24	99%
25	100%
26	100%
27	99%
28	97%
29	95%
30	91%
31	86%
32	81%
33	74%

Once I applied the percentages above to Byrd’s 2012 season, I was able to estimate his production until he is 33 (2019 season). The graph below shows his historical production (red) and his expected future production (blue), again measured by EPA+.

Based on the curve, we can expect Byrd to be better than his 2011 self until he is 32. He wouldn’t slip below a top-25 safety (assuming the average for top-25 safeties average above 30 EPA+ per season) until after his 33 year old season. Byrd could be a productive contributor to the Bills for at least seven more seasons.

Based on the average EPA+ salary formula determined in the last post, I calculated the salaries based on Byrd’s projected performance for the next seven seasons. The results are in the table below.

Season	Age	Byrd Exp EPA+	Annual Salary
2013	27	46	$7,142,363
2014	28	45	$6,914,957
2015	29	44	$6,569,978
2016	30	42	$6,118,001
2017	31	40	$5,573,360
2018	32	37	$4,954,164
2019	33	34	$4,282,276

Based on those annual salaries, a five year contract should cost the Bills somewhere around $32 million. His value for the next two seasons is around a total of $14 million. Notice how his expected annual salary for next season is greater than the franchise tag’s $6.916 million salary.

Should they sign Byrd for five years? Seven? How long should Byrd want to be under contract himself? We’ll investigate contract terms in the next post in this series.

Jairus Byrd’s Contract Situation, Part 1

 

Jairus Byrd is one of Buffalo’s most important defensive players and is currently not under contract. The Bills placed a franchise tag on Byrd, but the Pro Bowler has yet to sign the tender or attend optional team workouts. The Bills and Byrd are in contract negotiations regarding a long term deal, but how much is the safety worth? How much should the Bills pay him over how many more years?

This is the first of a five-part determination of how much the Bills should pay Byrd, what they can expect from him in the coming years, how long they should sign him for, how much the team values him, and whether the Bills really need Byrd to become a perennial playoff contender.

In this post, we are going to examine the market rate for safeties. I used AdvancedNFLStats.com’s EPA+ metric over the past three seasons to compare 30 safeties that will be playing the 2013 season outside of their rookie deal. I then compared their EPA+ over the past three seasons to their cap hits over the past two seasons and next season (assuming those seasons “earned” the trailing year’s deal).

I used EPA+ in this study because it measures a player’s positive plays by adding the expected points the opposing offenses failed to gain on plays in which each player was involved. For example, if an offense was facing third and ten fifty yards from the end zone, their expected points (the average points scored by an offense in this situation or later in the series) would be a certain value. A player’s EPA+ would be the offense’s expected points after they forced an incompletion, interception, sack, or whatever minus that prior play’s expected points value.

Basically better players have higher EPA+ totals (in theory). Any metric can be used for this study, but I preferred EPA+ in this situation because it just looks at the results of plays and isn’t subjected to individual stats or an inaccurate weightings of them.

I first compared the maximum EPA+ for the thirty players to their average cap hit from the 2011 to 2013 seasons. The results, along with a polynomial trend line, are in the graph below.

The trend line in this case has a slowly reducing slope as the player’s performance increases, showing a decreasing cap hit for each incremental increase in EPA+. Furthermore, the deviation from the trend line increases as the EPA+ increases. This might be a result of players having breakout seasons or having a great season, getting paid, and then losing productivity.

Based on this data and trend line, Jairus Byrd should be paid $4,639,651, more than $2 million less than the franchise tag salary he is set to make in 2013.

While maximum performances are interesting, they don’t always show how consistent or reliable a player may be. For that reason, I also compared average EPA+ over the past three seasons and the average cap hits from 2011 to 2013. The results are in the graph below.

This trend line is a better fit (R = 0.4544 compared to 0.4217), meaning it explains more of the data points. This trend line also differs from the maximum EPA+ regression because it has an increasing upward slope as EPA+ increases. That makes more sense, since consistently elite safeties are hard to come by and are paid very well for their services.

Based on this regression, Jairus Byrd would be worth roughly $4,800,396 per season. That total is again well below Byrd’s franchise tag salary.

Byrd, however, has been an elite player over the past three seasons. His average EPA+ over the past three seasons would place him as the eleventh best safety in this study. He was the fourth best safety in 2012, according to EPA+. Players with an average EPA+ greater than 35 (twelve total) were paid an average of $136,253 above the salary estimated by the trend line, so there’s hope for Byrd to fetch more than $5 million per season.

Next time, we’ll investigate the aging curve for safeties and determine whether Byrd should be paid more than the trend lines due to the fact that he is coming into his prime rather than declining.

Tyler Myers and Ron Rolston

Ron Rolston was promoted from Buffalo’s “interim head coach” to “head coach” yesterday. Apparently he was given the job as a result of the praise he received from the players in their end of the season interviews. While the team didn’t really improve overall (the difference between their average goal differentials with Ruff and Rolston were not statistically significant), Tyler Myers’ play did improve. (*Disclaimer:* The sample sizes from this season are small. I was able to perform some hypothesis testing on the data, but I’m stretching things a bit due to Myers playing in just 39 games last season. I really need at least 30 games with and without Ruff to claim the tests I ran are statistically significant, but we didn’t have that luxury this season.)

Under Ruff, Myers averaged a plus/minus of -0.4667 and had a -0.04167 with Rolston as coach. While the plus-minus stat is flawed and doesn’t truly take an individual’s play into account, the improvement is noteworthy. The graph below shows Myers’ five-game plus/minus average over the course of the season. The red portion of the line is when Ruff was coach, and the blue is when Rolston was coach.

The increase was statistically significant, meaning there is little chance that the improvement was a result of normal variation. To test the significance, I used Myers’ time under Rolston as the sample population (larger sample size) and applied the central limit theorem to determine the average plus/minus difference under Ruff from the population (time under Rolston). The resulting probability that the difference would be due to normal game to game variation was 0.4%.

Myers’ time on the ice was another notable improvement under Rolston. The graph below shows the game by game playing time for the big defenseman.

Myers played 22 minutes or more in his final seven games in 2013. Before then, he played more than 22 minutes only eight times in the previous 32 games. When Ruff was coach, Myers averaged just under 19.5 minutes per game. That average increased to 21.75 minutes when Rolston was coach.

The increase (or decrease under Ruff, as I calculated it due to sample sizes) was even more significant than Myers’ plus/minus improvement with Rolston. The probability that Myers’ playing time increase was just a result of normal variation was 0.00001%.

While Buffalo’s overall goal differentials per game did not significantly increase, the Sabres’ best player (or at least potentially best player), Myers, did improve under Rolston. Clearly Rolston’s message gets through to Myers better than Ruff’s did, and it might translate into more wins and a playoff berth next season.

Looking at Big Plays

 

Buffalo’s defense last season wasn’t good, ranking second to last in rushing yards allowed. But they gave up fewer big plays (plays of twenty or more yards) than the average NFL team. The Patriots defense (the eighth worst overall defense last season in terms of yards allowed, but tenth best by points allowed) gave up fifteen more big plays last season than the Bills. Does that mean the Bills had a better defense than the Patriots in 2012?

On the surface, it would seem that Buffalo’s defense was better because they didn’t allow as many soul-crushing plays. However, New England was winning for the majority of the time last season and opponents were throwing against a soft zone or prevent defense in the hopes of catching up in the game. The graph below compares the number of total plays (plays from scrimmage, kickoffs, punts, PATs, etc.) the Bills and Patriots spent ahead, behind, or tied.

The timing of the big plays is also important. Because New England was ahead so often, they mostly allowed big plays when ahead. The Bills had huge issues holding their opponents down when tied or behind.

70% of Buffalo’s big plays allowed came when the Bills were either trailing or tied. 60% of New England’s big plays allowed occurred when they were ahead. The graph below compares when the Bills and Patriots allowed big plays. The distribution isn’t much different than the graph above.

The types of big plays allowed are also telling. While the Patriots allowed a bunch of big plays, they were almost all passing plays. That furthers the notion that they allowed big plays when ahead and teams were pass-happy attempting to get back into the game. Despite allowing fifteen fewer big plays, Buffalo allowed fourteen more big rushing plays. The AFC East big play distribution is in the table below.

Even though the Bills didn’t allow as many big plays as the Patriots, they occurred at worse times. This proves that season statistics and totals don’t accurately measure teams against one another. The context, the when and where, is important.

Marquise Goodwin and NFL Comparables

 

NFL.com compared third round draft pick Marquise Goodwin to Mike Wallace. That’d be a great result, if he becomes that level, but is that comparison really meaningful? Don’t these “NFL comparisons” kind of suck?? And what about TJ Graham, wasn’t he supposed to be our burner wide receiver?

Mike Wallace was a very productive collegiate receiver, gaining almost 2,000 yards in three seasons. He led the Ole Miss Rebels in receiving yards in both his sophomore and junior (final two) seasons. Goodwin was third in receiving yards on his junior and senior Texas teams. Graham only emerged as the top NC State receiver in his final season. The three player’s production throughout their college careers is in the graph below.

Goodwin and Graham had similar production throughout their college careers, while Wallace was clearly the more productive player. Goodwin was a consistent producer at Texas, but only had two 100+ receiving yards games. Graham didn’t have a 100+ yard receiving game until his senior year!

Mike Wallace averaged two 100+ yard receiving games per season. And that was on an Ole Miss team that attempted passes less than 45% of the time.

Those NFL comparisons are also based on player potential. Wallace performed well at the 2009 NFL Combine, but his performance on the agility drills (short shuttle and three cone drills) were just average. He’s a straight line burner with average size. The graph below compares the combine results of Goodwin, Graham, and Wallace as measured by standard deviations above and below the average wide receiver.

TJ Graham’s combine results are those of a one-dimensional receiver. He’s very fast (although not as fast as Wallace or Goodwin), but small and not very strong, agile, or explosive.

Wallace possessed the explosive metrics (broad jump and vertical leap) as well as speed that only 1.5% of receivers have, but he doesn’t match Goodwin’s all-around athletic ability. Goodwin, although small, has the speed, agility, and explosiveness you would expect out of an Olympic long jumper.

Goodwin’s potential could compare to Mike Wallace, a deep threat burner receiver, but that’s a lazy comparison. Wallace proved to be that deep threat in his college career, while Goodwin was used more in open space and a running threat. I expect Goodwin to be used in a similar way in this Bills offense.

On that note, I don’t think Goodwin means the end of TJ Graham. With last year’s lack of depth at receiver, Graham gained a lot of NFL experience. While he had some mental errors last season (the most memorable was the miscommunication between Graham and Fitzpatrick at the end of the second Patriots game last season), Graham still hauled in 31 receptions for 322 yards.

Goodwin is an interesting addition to an offense that just got a whole lot faster. This offense will thrive if its players will get the ball in open space.

On Matt Barkley and Ryan Nassib

Just as I was starting to accept Ryan Nassib as the Bills quarterback option in the draft I see this:

 

Three sources I trust very much are saying the #Bills prefer Matt Barkley over Ryan Nassib.

— Matt Miller (@nfldraftscout) April 24, 2013

 

So that provoked me to look into the two players’ total yards per attempt (TY/Att) over the course of their careers. For comparison’s sake, I calculated the five game trailing average for the past four seasons. The result is the graph below.

As you can see, Barkley consistently outperformed Nassib. The USC quarterback had a wider rage of performance, but his upside was much better than Nassib’s. Furthermore, Barkley hadn’t performed below Nassib’s career average since the very beginning of his junior season.

Tomorrow should be fun. I can’t wait.