Jahnke Value Model: Primer

Nathan Jahnke discusses his "Jahnke Value Model" study.

| 2 years ago

Nathan Jahnke discusses his "Jahnke Value Model" study.

Jahnke Value Model: Primer

In 2012 we introduced a new Pro Football Focus offseason preview called Performance Based Value. The premise is simple; we take how well a player played in the previous season and turn that into a dollar amount for how much that player should have made. We then compare that to how much of a cap hit that player had to determine which players were the most over or under paid in 2013. This year we decided to re-name it the Jahnke Value Model.

Here you can find a more detailed explanation on the process for coming up with the Jahnke Value Model as well as a link to this season’s various JVM articles all in one place for your convenience. We’ll be looking the most under and over-valued players from each roster based on their 2013 performance and will go through them in draft order, starting with Houston and working our way from there.

The by-team and by-position pieces linked here as they post:

GB | HOU | IND | JAX | KC | MIA | MIN | NE | NO | NYG | NYJ
OAK | PHI | PIT | STL | SD | SF | SEA | TB | TEN | WAS

QB | HB & FB | TEWR | OT | G & C
S | CB | LB | Edge Defenders | Interior D-Line

Here are some notes on the method used:

The Basic Idea

We set out to take the highest cap hit at a position and match that up with that position’s highest-rated player. Then take the second-highest cap hit and match it with the second-highest-graded player, and so on. All players who were either active or inactive for at least one regular season game were included, and players who spent the entire year on Injured Reserve or as free agents were ignored. The goal being to set salaries right, based solely on player performance during the 2013 season.

Making it Work

The goal was to matchup the top players with the top salaries. For example, LeSean McCoy was our highest-rated halfback in 2013, so he should make roughly the same amount as what the highest-paid halfback in 2013 made. Jamaal Charles was our second-highest rated, so he should make roughly the same amount as the second-highest paid halfback, etc.

One problem with just listing all of the players by grade and matching them up with salaries is that we needed to account for playing time. It wouldn’t make sense for a player with a +0.0 grade on 1 snap for the season to make more money than someone who had a -1.0 grade on 1000 snaps. Instead of using a normalization factor that compares everyone to an average, we’ve used a replacement level player normalization to come up with new overall grades for players.

Everyone who played at or below replacement level was given the league minimum dependent on their years in the league as well as games on the roster. Also everyone who had 100 or fewer snaps (outside of punters) also received the league minimum due to a small sample size. Everyone else above replacement level earned more than the minimum.

A second problem is that the distribution of players cap hits differs from the distribution of players overall grades. For example the best player at a position could have a grade of +30.0 while the second-best could have finished at +20.0; a sizable drop from the top. The top two cap hits, however, could be nearly identical. It wouldn’t make sense to have two players with vastly different grades marked as deserving the same pay. Each position was broken up into tiers, and best fit curves were created to find the right balance between the gaps in grades and gaps in salary.

From there we could plug in a player’s grade into the created formulas, and the output was their performance based value.

Exploring the Data

There are various ways we could use the Jahnke Value Model. While we could rank players by who should get paid the most at each position, that would basically end up the same as listing them by the highest grades. What is much more interesting is comparing how much money a player actually made in 2013 with how much he should have made based on his performance and that’s where we’ll head first.

Keep in Mind

While Jahnke Value Model is a fun exercise and does a decent job in showing to what extent players were under or over paid in the past, there are a number of limitations to this analysis:

1. These values for players apply to the 2013 season. Although it can have some indication for the future, it wouldn’t make sense to apply in that manner in all cases. This is especially true with injured players. Someone like Anthony Spencer played 38 snaps this season and will have a very low value according to the Jahnke Value Model. That certainly doesn’t mean he should be worth so little going forward. Creating a predictive Jahnke Value Model is something that we may look to do in the future, but currently it should only be used retrospectively.

2. We aren’t looking to say that one position group as a whole should get paid more or less than they currently are. If you add up all of the quarterback cap hits in 2013, it will roughly equal the sum of all of the quarterbacks in the Jahnke Value Model. Even though last offseason we did analysis to show that left tackles are likely overvalued in comparison to right tackles, this analysis isn’t trying to show anything like that — it’s just trying to show that certain left tackles should be paid more and others less based on how left tackles are currently paid.

3. For this we are using 2013 cap hits. In the majority of cases, the more years into a contract someone is, the higher their cap hit. Therefore a player at the end of their contract is more likely to be overvalued than someone who is at the beginning. We could have used something like average cap hit over the course of their contract, but chose not to. In too many cases players have a large cap hit in their last year that is never actually used due to the contract getting restructured or the player getting cut. While not perfect, simply using the 2013 cap hit made the most sense.




Follow Nathan on Twitter: @PFF_NateJahnke

  • Thomas James

    I’m currently a PhD candidate and this sounds complicated. Have fun Jahnke. 😉

  • Andrew

    This really isn’t a fantastic model for player value. Way too many assumptions made here in order to skip sound statistical methods of evaluation. For example, you don’t explain how you rank players in the first place. Secondly, you bypass attempting to determine true market value for a player by just placing arbitrary values on current players. For example, if one year all running backs perform horribly and none rush for more than 1000 yards, you can’t take the best from this group and say that he deserves $9 million dollars this year. That’s incorrect. Players’ values need to be based on a statistical model for determining contract value. For example check out my blog and model to be found at http://www.sportonomic.com
    An example is going live tonight in regards to this blog post. Yours is an interesting idea, it just doesn’t merit any academic or real statistical significance…