The value of free agents in the NFL - Part 3

Check out Part 1 and Part 2 of this ongoing and overly long series on the value of free agents in the NFL

It doesn’t matter if your players perform slightly worse if you can get a good deal on them. We’ve been looking at performance but value is the real target here.

Before we get all the way to value we’ll start with a look at cost.

The average player who changes teams is not cashing in on a big payday in the way we might think. While players like Paul Kruger (in 2013 to the Browns) or Matt Cassel (in 2009 to the Chiefs) come to mind when looking at free agency, NFL contracts are basically a series of one year options held by the team. 

For good players, the ability to hold out can give them some leverage for a raise as long as they continue to be good. If a player is merely average they can only hope that the team would have to take a big cap hit to cut them, most won’t see a dollar more from the team if they are let go[1].

Figure 10 - Change in SC% (Prior Starters)

Figure 12 - Changed Team

Figure 11 - Same Team

In the season before changing teams, the average salary cap hit for a player who will change is 2.20% (1.24% for non-starters, 2.87% for starters). Once they change their salary cap hit is 1.56% (0.89% for prior non-starters, 2.03% for prior starters). Most of these players are taking a pay cut to stay in the league.


For players remaining with the same team the prior season cap hit is 2.68% (1.08%/3.28%) and the subsequent season is 3.00% (1.18%/3.69%).

This is promising news for teams signing players from other teams. Even though the performance data showed a decline, it’s possible that there’s enough decline in salary to make the player an attractive value.

Prior Starters	Coefficient				P-Value
	Skilled		Line		Skilled		Line
	Offense	Defense	Offense	Defense	Offense	Defense	Offense	Defense
Intercept	0.38%	1.63%	1.06%	1.31%	0.05	0.00	0.00	0.00
Prior Salary Cap %	57.61%	48.68%	58.87%	54.76%	0.00	0.00	0.00	0.00
Prior AV (All)	0.21%	0.17%	0.19%	0.14%	0.00	0.00	0.00	0.00
Prior AV (Changed)	-0.17%	-0.14%	-0.11%	-0.17%	0.00	0.00	0.00	0.00
Career Season	-0.02%	-0.12%	-0.10%	-0.06%	0.29	0.00	0.00	0.01

Prior Non-Starters	Coefficient				P-Value
	Skilled		Line		Skilled		Line
	Offense	Defense	Offense	Defense	Offense	Defense	Offense	Defense
Intercept	0.52%	0.78%	0.58%	0.68%	0.00	0.00	0.02	0.00
Prior Salary Cap %	33.17%	30.37%	48.77%	49.48%	0.00	0.00	0.00	0.00
Prior AV (All)	0.13%	0.25%	0.26%	0.19%	0.00	0.00	0.00	0.00
Prior AV (Changed)	-0.06%	-0.19%	-0.37%	-0.16%	0.01	0.00	0.00	0.00
Career Season	-0.01%	-0.06%	-0.04%	-0.05%	0.57	0.01	0.09	0.00
	Prior Starters				Prior Non-Starters
	Skilled		Line		Skilled		Line
	Offense	Defense	Offense	Defense	Offense	Defense	Offense	Defense
R^2	0.63	0.36	0.52	0.47	0.29	0.25	0.19	0.35

 
Note: Those values that are bold/italicized are where p>0.05 for that coefficient 

The above figure is a regression for salary (as a percentage of that year’s salary cap) against prior salary (percentage of prior year salary cap), prior salary for players who change teams, prior AV and career season. While career seasons and age combined to be helpful in performance, they both point the same way in looking at value as salaries trend upwards over time. I removed age because it seems more logical that tenure in the league is explanatory given that the NFL calculates minimum salaries based on tenure.

It is apparent that the relationship is strong even when controlling for different points in career and performance levels of players who change teams. Prior starters who change teams have most, or all in the case of non-skill players on defense, of their bump in salary from performance cancelled out.

Interestingly, the coefficient for career season is negative for all combinations. Because this analysis includes mainly players making above the minimum salary, the intercept and prior SC% terms put most players well above the minimum and the increasing salary floor does not impact the regression enough to make additional seasons predict higher salaries.

Value

We’ll address value the same way we looked at player performance and salary. Prior AV will still be included, as will a player’s number of seasons in the league.

To these two variables we will add a measure of the player’s prior value. In order to make this easier to understand this value will be expressed as AV per 1% of the salary cap (SC) – the value a team gets for their money. This variable will be paired with a dummy variable for team changes the same way Prior AV did in the performance regression: 0 if a player stayed on the same team and the player’s AV per 1% of SC if the player changed teams.

AV per 1% of SC is not as intuitive as I would like, so allow me to offer some context. The figure below is a good way to frame AV per 1% of SC.

Figure 13 - Total Team AV and Pythagorean Winning %

Some teams were abnormally lucky or unlucky while others were in a very easy or very difficult division, but a team needs to pull in roughly 2.3 AV per 1% of SC to reach the level of an average playoff team. The average team has 2.0 AV per 1% of SC and a team that misses the playoffs averages 1.8 AV per 1% of SC.

The good news is that players on their rookie contract outperform this benchmark on average. Players in seasons 1 through 4 deliver 3.8 AV per 1% of SC (4.4 for playoff teams, 3.4 for non-playoff) while delivering 46% of the league-wide AV[2]. Besides drafting well (something I have covered before) the average playoff team over the 2003-2009 period brings in 2.1 AV per 1% of SC spent on veterans[3].

The Data

Now that we have a bit of context it’s time to move on to the data. As before we start with the regression for all prior starters:

Figure 14 - Value Regression

Rather than the significant negative impact changing teams has on performance and salary, changing teams appears to have an insignificant impact on the value of prior starters. Value reverts strongly as Prior AV per 1% SC has a coefficient of 0.20 compared to 0.54 for Prior AV and 0.56 for Prior SC%. Even if the change term were significant, the impact is much smaller than for either of the two other factors. 

Prior Starters	Coefficient				P-Value
	Skilled		Line		Skilled		Line
	Offense	Defense	Offense	Defense	Offense	Defense	Offense	Defense
Intercept	2.14	1.76	1.88	2.28	0.00	0.00	0.00	0.00
Prior AV / 1% SC (All)	0.16	0.17	0.27	0.22	0.00	0.00	0.00	0.00
Prior AV / 1% SC (Changed)	-0.04	0.06	-0.04	0.08	0.22	0.14	0.35	0.01
Prior AV (All)	-0.02	-0.07	-0.10	-0.06	0.21	0.02	0.00	0.00
Career Season	-0.04	0.07	0.04	-0.03	0.08	0.07	0.16	0.28
Prior Non-Starters	Coefficient				P-Value
	Skilled		Line		Skilled		Line
	Offense	Defense	Offense	Defense	Offense	Defense	Offense	Defense
Intercept	2.10	2.47	3.41	2.72	0.00	0.00	0.00	0.00
Prior AV / 1% SC (All)	0.33	0.31	0.56	0.08	0.00	0.04	0.07	0.58
Prior AV / 1% SC (Changed)	-0.10	-0.03	0.51	0.21	0.17	0.84	0.03	0.09
Prior AV (All)	-0.02	-0.71	-0.61	0.05	0.82	0.00	0.12	0.78
Career Season	-0.05	0.20	-0.10	0.00	0.28	0.02	0.28	0.97
	Prior Starters				Prior Non-Starters
	Skilled		Line		Skilled		Line
	Offense	Defense	Offense	Defense	Offense	Defense	Offense	Defense
R^2	0.07	0.07	0.12	0.13	0.07	0.05	0.07	0.01

Note: Those values that are bold/italicized are where p>0.05 for that coefficient 

Bringing out the individual combinations of offense/defense and skill/line it is striking how many coefficients fail to meet the threshold for significance after the performance and salary results.

Maybe those performance and salary results are themselves the explanation. The implications on value when players change teams, suffer lower performance and earn lower salary are unclear because the factors cancel each other out. It is clear that the model is struggling to explain value. The R-squared for all prior starters is 0.09 and rises to 0.13 for the strongest combination (defensive line prior starters).

For the prior non-starters there is even less to see. The regressions do not do a good job here.

At an aggregate level the prior starters go from delivering 2.2 to 1.8 AV per 1% SC. Those who changed teams increase from 2.0 to 2.2 while those who stay on the same team drop from 2.3 to 1.7.

In the next post, the final post in this series, I will spend some time reviewing what we can conclude from this analysis and what the implications are for roster construction.

---

[1] One-time payments such as signing bonuses are amortized over the life of a contract (with some limitations) to reflect the player’s cost to the team in a given year for salary cap purposes. If a player is released their remaining payments to be amortized become “dead money” and hit the cap in the current or immediately following season depending on the timing of the release. This reduces the amount teams have available to spend under the cap.

[2] Both playoff teams and non-playoff teams spend roughly the same amount on players in their first four seasons (33.5% for playoff teams and 35.0% for non-playoff teams)

[3] This all ignores the rather large problem that not all AV is created equal. I’ll get into this in a later post where I adjust AV to make it more representative of the positional differences in value. The allocation built into the statistic is based on draft value spent on each position.