The Recency Effect

For those of you that haven't read Jonah Keri's two part series on Fangraphs titled "Who is the Most Valuable Player in Baseball (Part 1 and Part 2), stop what you're doing, click on the links, and read.  In part two, Keri makes an outstanding point with regards the dangers of putting too much stock into a player's most recent season of performance.

"The Recency Effect is a phenomenon that can overpower even the savviest thinkers. We tend to place too much emphasis on an event that just recently occurred, and lose sight of the much longer string of events that immediately preceded it. If you eat a banana and you’re blowing chunks 10 minutes later, you might shy away from bananas in the future, even if you’ve eaten hundreds of them in the past without incident.

Ramirez’s 4.4 WAR season in 2010 is our rancid banana (assuming it was the banana that even made you sick in the first place, and not something totally unrelated). A player posts a wOBA over .400 three seasons in a row, does it while playing the most demanding position on the diamond, and yet some people might downgrade him severely because of one very good but not quite great season."

Typically, when I project a player's future performance, I use the last three seasons my guide.  Why do I use three seasons?  Well, a three season sample size allows me to minimize not only the effect a player's age and experience can have, but also the potential effect for outliers in the given sample set.  Let's say Player A has been a consistent .365 wOBA hitter throughout the first six years of his career.  In year seven, he unexpectedly produces a .410 wOBA.  Which type of hitter do you expect him to be in year eight?  Many fans, writers, and (some) analysts will make the assumption that Player A's improved performance is because he's "figured something out" or "matured as a hitter"; therefore, his next year's performance will more closely resemble the .410 wOBA line rather than match his prior track record.  While this might be true in some cases, it's highly atypical for player to suddenly and drastically change his true performance baseline.  There are far more plausible reasons for Player A's sudden performance improvement, most of which are outside of his realm of control.  For instance, unusual variances in "luck" factors such as a player's batting average on balls in play (BABIP), home run per fly ball ratio (HR/FB), and fly ball/line drive ratio often plays a significant role.  In nearly all circumstances, unusual variances in one season do not carry over to the next season.  They are, after all, unusual for a reason.  Instead, those variances regress to the mean.*  As a result, Player A's performance regresses along with it.  Rather than putting together an encore .410 wOBA season, he typically will produce a season that's closer inline with his true .365 wOBA performance baseline. 

*  In statistics, there's a concept called regression toward the mean.  For those of you who are not familiar with the concept, regression toward the mean is defined as "a statistical phenomenon that occurs whenever you have a nonrandom sample from a population and two measures that are imperfectly correlated."  In layman's terms, if you flip a coin 50 times, and it comes up "heads" 35 times, chances are those results were due to an unusual statistical variance.  If you were to flip it another 250 times, it's very unlikely the coin would continue to come up "heads" 70% of the time.  In all likelihood, the random variance in luck that occurred during the first 50 flips would regress toward the mean, and the coin would come up "heads" approximately 150 times (out of 300), as opposed to 210 times.  The larger the sample size you use, the greater the likelihood random variances and luck become non-factors.

Joe Mauer is a great example of the recency effect.  In four-plus seasons prior to his 2009 MVP season, he consistently produced a wOBA within twenty points of his .375 average.   While power had always been projected for Mauer, it never developed in the manner in which most scouts had hoped.  Mauer had developed doubles power along with his ability to hit for average and get on base, but his penchant for hitting the ball on the ground (49.5% GB rate versus 27.8% FB rate) precluded him from ever becoming the consistent 20-25 home run hitter many expected him to become.

In 2009, seemingly out of nowhere, Mauer began hitting a lot more home runs--28 of them in fact.  While it came as quite a surprise to most in the analyst community, many fans, writers, and coaches proclaimed that Mauer had finally discovered his power stroke.  (Or as Duane Winn of the Bleacher Report ridiculously opined, started using steroids.)  Surely, if he's suddenly hitting more home runs, then he must be hitting more fly balls as a result.  Right?  Wrong.  In 2009, Mauer's fly ball rate was nearly identical to his career track record.  The only difference was this time, 20.4% of his fly balls went over the wall; whereas, between 2004-2008, they'd only breached the wall 8.8% of the time.  Why the sudden jump in HR/FB rate?  Well, I can think of two obvious, logical reasons.  One, Mauer made a few mechanical adjustments to his swing that allowed him to get some additional loft and distance on his fly balls.  Two, he got lucky.  While it's certainly possible that a swing adjustment created some of his additional power, the analyst in me looks at his batted ball rates and his HR/FB ratio, and surmises that Mauer's increased power numbers were largely due to luck.  In all likelihood, his HR/FB ratio, and consequently his home run total, would regress back to the mean in 2010.

In 2010, Mauer proved many analysts right--his HR/FB rate regressed back to a rate more inline with that of his career.  Rather than posting a gaudy 20.4% rate like he did in 2009, he produced a 6.7% rate in 2010.  Without a corresponding surge in the number of fly balls to counteract his regressed HR/FB rate, Mauer was unable to sustain the power numbers he'd achieved just one year prior.  He hadn't discovered his power stroke.  Instead, his 2010 season showed that his 2009 power surge was encouraged by the HR/FB fairy that carried several of his warning track fly balls a couple of extra feet over the fence.  If Mauer had made a meaningful adjustment in his swing that allowed for increased power numbers, we would've seen not only this effect in terms of an elevated fly ball rate, but also a repeat performance in terms of his power abilities.  That didn't happen.  Considering the benefits of such a change, it's unlikely Mauer would choose to revert back to his old swing as a result.  Therefore, the "swing change" theory doesn't hold water.  Instead, luck seems like a more plausible conclusion.

So what does this all mean?  It means that we need to exercise caution when dealing with outlier seasons, whether good or bad.  If a player has been performing at a certain level for a long period of time, and then produces a career season, we should be skeptical of that career season.  We should ask questions.  Why did it occur?  What uncontrollable factors may have been at play?  What is the true talent level of the player in question?  If a player performs outside of his typical norms, there's usually a reason.