The
first wave of free agency is in the books and we’ve already seen a lot of splash moves, like the
Indianapolis Colts trading for
DeForest Buckner, the
Arizona Cardinals trading for DeAndre Hopkins, the
Miami Dolphins signing
Byron Jones and the
Cleveland Browns signing tight end Austin Hooper.
We’ve
graded the Cardinals' and Dolphins' deals as “good” and the deals done by the Browns and Colts as “below average,” but all these deals have one thing in common: The respective teams expect their new players to have a huge impact in 2020, just as teams did this time last year.
However, reality often paints a different picture. Since 2006, players (excluding quarterbacks) who added value to their teams in a given season (i.e., they generated positive WAR) generated a total of 99 WAR less afterthey changed teams in free agency.
That’s roughly one-quarter of a win less per team and season than teams would have hoped for if they signed the new players hoping for just a replication of their previous year's success.
First of all, that’s not really a surprising result because of the presence of regression to the mean, as a player who added positive value is generally expected to produce less value the following year. We also find this for players who don’t change teams, so a natural question to ask is whether or not the change of teams strengthens the regression.
As a first rudimentary analysis, let us bin players into three groups based on the previous year WAR and find the average WAR difference for players who switched teams and players who didn’t.
WAR difference for players with positive WAR in the previous season, excluding quarterbacks, 2006 – 2019
| Previous Year WAR | WAR difference for players on the same team | WAR difference for players who switched teams |
| 0 – 0.1 | +0.01 | -0.01 |
| 0.1 – 0.3 | -0.03 | -0.06 |
| 0.3 + | -0.12 | -0.19 |
So, the better a player was in the previous season, the more he is expected to regress, which is a very intuitive result. Furthermore, we indeed obtain the first piece of evidence that switching teams increases the effect of regression to the mean.
To confirm this, we perform a more nuanced analysis, i.e., we control for experience (age). Older players see a stronger regression than younger players and are more likely to switch teams, as 79% of all players who have played no more than five years in the league stay with their teams. This number falls to 70% for players who have played in the NFL for six or more years.
To account for this, we build a model that regresses the magnitude of the regression to the mean (WAR difference) based on the combination of previous-year WAR and age. We use this model to normalize the WAR difference for each four-year player in our database (because of the length of the rookie contracts, the transition from Year to 4 to Year 5 observes the most team changes).
We find that the effect is still present across the full spectrum of player quality. Good players generally regress to the mean, and this effect is amplified when a player switches teams.
This suggests that a player's environment does matter to his performance — when a player performed well in a given season, it was likely because the environment was good. Therefore, the odds are that his environment worsens when he switches teams.
While NFL parity is largely working, in the sense that there are hardly any teams who can sustain their excellence for a long time, the saying “bad teams stay bad” still has some credibility. Our findings suggest there are multiple effects in play: The worst teams obviously need to improve the most, hence they sign the supposed best free agents and thus spend a lot of money on players who are expected to regress the most, especially on a new team.
INVESTIGATING INDIVIDUAL TEAMS
The following chart shows how much value each team has given up over the last 13 offseasons by losing players who gained positive WAR on other teams (whether this might be per trade or per free agency).
We then compare this value to the WAR those players gained for their new teams in the following season. Even though the
Buffalo Bills did their best in this regard, every single team lost players that, on average, performed worse for their new teams in the following season.
This fits our general discussion above. We’ve again excluded quarterbacks, since signing a quarterback from another team and having him perform well would break the scale here.
The
New England Patriots and
Baltimore Ravens have lost the best players in free agency. However, the teams that took the players on haven’t necessarily been happy with their shiny new players out of New England or Baltimore. Given our findings, it’s questionable to sign a player to a large contract after he had a good season or stretch in Foxborough. Just recently, examples like
Trey Flowers,
Nate Solder,
Dion Lewis and
Martellus Bennettcome to mind, none of which met expectations with their new teams (so far). The players who regressed the most after leaving the Patriots were veterans in the twilight of their careers.
Brandon Browner,
Darrelle Revis and Randy Moss all disappointed with their new teams after Bill Belichick decided they didn't help his team enough to keep them in New England.
It’s hard to find the direction of the causation, but it’s interesting that other teams that were blessed with stellar quarterback play for more than a decade but couldn't replicate the Patriots' success — the Steelers, Colts, Saints, Packers, Falcons and Chargers — are at the bottom of the chart. That is, they lost few of their good players to other teams during the offseason. It’s not clear whether they just couldn’t find enough good players or whether holding onto players for too long actually harmed them, potentially by missing out on valuable draft capital in the form of compensatory picks or trade compensation.
What about looking at it from the perspective of the new team? Which teams had their expectations disappointed the most through free agent signings? The following chart shows for each team how many WAR new players generated compared to how many WAR they generated for their old team during the previous season.
Of course, when losing a lot of value to other teams, compensation is required, and nobody was better at finding it than the Ravens and Patriots. The Ravens were even a unicorn in a sense — no other team managed to get more value out of their new players than they produced for their old teams the previous season. The Ravens added players who combined for 8.6 WAR on other teams the previous season and managed to get a production worth a total of 10.4 WAR out of them in the first season in Baltimore. It goes without saying that the Patriots are also at the top of the league in terms of creating an environment that helps their new players to succeed. Their new players from other teams regressed only by 10%, the lowest rate besides the Ravens.
We also observe the “bad teams stay bad”-effect in full play, as teams like the Redskins, Dolphins, Browns, Lions and Raiders all invested a ton in getting supposedly valuable players from other teams only to see them regress by more than 50% in Miami, Cleveland, Detroit and Washington.
CONCLUSION
It is, of course, hard to find actionable advice from our findings, as building an environment that allows new players to succeed is easier said than done. So is finding the sweet spot of when to part with veterans. However, it could be worthwhile to hesitate less when it’s time to part with merited veteran players, especially when the opportunity to get back some draft capital presents itself.
Our findings emphasize the value of building through the draft, as this is the only chance to observe a player in the environment of his own team. Consequently, realizing a player is good with the original team is more valuable than realizing a player is good with another team. This is related to a
study from our own Kevin Cole, who found that extensions of a team's own players produce more wins per salary dollars than unrestricted free agent signings.
From an analyst’s perspective, we should be much more cautious when assessing how an offseason affects team strengths. Most of us (
including PFF) were guilty of being too bullish on the
Cleveland Browns last year, mostly because nobody thought
Odell Beckham Jr. would produce less than 50% of the WAR he put up in his worst season before 2019 (other than the 2017 season he missed to injury).
As a more nuanced analysis of the shift of team strengths, Kevin Cole has created the
PFF Improvement Index, which projects the impact of each offseason arrival on his new team instead of looking back. It illustrates that the total shifts are lower than one might think, with half of the league hovering between -0.1 and 0.1 wins added. The largest improvement is currently projected for the
Indianapolis Colts with 0.36 wins added — and this is entirely explained by replacing
Jacoby Brissett, who “added” -0.85 wins below average in 2019, with a starting-caliber quarterback in
Philip Rivers.
Naturally, there are many teams that improve by considerably more than 0.4 wins from one season to the next. What we should keep in mind is that the reason for the improvement is never that the team made offseason moves that worked out
as expected. Apart from lucky bounces in close games, most teams improve because either offseason acquisitions or players already on the roster (and very often both) outperformed their expectation, something that is very hard to predict beforehand.
