Part 2 of Using Contact% & Pull% to Predict a Batter’s Decline

Last week, I examined how a decline in pull rate (Pull%) and contact rate (Contact%) may be a precursor to overall decline. The idea was that if the hitter was worse at doing either or both, they could be in for a larger than normal decline. The research was an initial stab at the data and I got some great comments for future areas of stuady on it. With a few tweaks, I was able to take the ideas and refine the research in order to have a better understanding of what can be a good sign of a major decline.

Idea #1 – Use O-contact% instead of Contact%

Reader Brendan stated the following:

It would be interesting to take this concept and run an ANOVA amongst the 4 groups + control for o-zone contact rate AND z-contact rate separately, being that o-zone contact % is the one peripheral that drops off the cliff dramatically for older players. We’d get an idea about the extent to which overall wRC+ is affected by o-zone contact decline. Hardy’s 2014 and 2015 o-zone contact rate was much lower compared to recent years.

So I analyzed the data with various combinations of Contact%, O-Contact% (outside the zone contact rate), and Z-Contact% (inside the zone contact rate). I looked at several different combinations and none were really any better than Contact%. I decided to move forward with just Contact% for the rest of the analysis.

Idea #2 – Weight of Contact% and Pull%

This again came from a reader (AJS):

Interesting — but what’s the logic for multiplying the decline in contact rate by 2

And my response:

The normal range of change in Contact% is about half that of Pull%. Sorry for not adding it.

That was the reasoning I used, but it was a terrible reason. I regressed the data to get some better multipliers. I was close on my initial guess, but the better estimate for the Contact% multiplier was 2.5.

With Contact% being such a major factor, I went and removed Pull% from the equation and found it was still needed. Without Pull% added to the equation, the results merely mirrored the league average decline.

Finally on this front, I made one small change which helped get better results, I made sure both of the two values declined before the player was included.

Idea #3 – Effect of injuries

This idea of using injuries came from Peter:

Question with regards to McCutchen. Was there any noticeable difference in his numbers for the month of April versus May-October? He seemed to be bothered by a knee injury for about a month and a half which was preventing him from turning on any pitches, so I’m curious to see if April is responsible for him being on this list.

Besides McCutchen, Hunter Pence, who missed most of the season with an arm and oblique injury, topped our original list of declining players. I ran the values on players who went on the disabled list, and they have huge rebounds the post injury season. If you notice a player who declined, but had a major injury, expect a rebound.

Pulling the ideas together

Putting all the ideas above together (new weighting of Contact%, requiring both Pull% and Contact% negative, and using only non-injured players), the following declines in wRC+ were noted for the players with different levels of Decline (2.5*change in Contact% + change in Pull%). Also, the average amount of decline of wRC+ for the average age of player (29-years-old) examined was ~4 wRC+, for context.

wRC+ Change for Decline in Pull% and Contact%
Decline wRC+ (Y1 to Y2) wRC+ (Y2 to Y3)
-20% -18.4 -7.9
-18% -11.9 -7.1
-16% -12.9 -4.7
-14% -8.8 -3.2
-12% -6.0 -4.1
-10% -4.7 -4.0

The big wRC+ drops continue in “Declines” over 16 points. Additionally, none of the hitters in this range rebound/regress.
Looking onto 2016, here are some players who meet the above thresholds because of declines from the 2014 to 2015 season.

Players on Steep Decline from Drop in Contact% and Pull% (Revised)
NAME wRC+ wRC+ wrc_drop12 Contact% Change Pull% Change Decline Age
Seth Smith 132 113 -19 -7.00% -8.00% -24.0% 32
Sean Rodriguez 98 79 -19 -2.00% -17.00% -22.0% 30
Travis Snider 122 81 -41 -6.00% -4.00% -19.0% 27
Brett Gardner 111 105 -6 -5.00% -5.00% -18.0% 31
Adam LaRoche 127 75 -52 -6.00% -3.00% -17.0% 35
Rene Rivera 113 33 -80 -4.00% -7.00% -17.0% 31
Kole Calhoun 125 105 -20 -6.00% -1.00% -16.0% 27
Casey McGehee 102 52 -50 -6.00% -1.00% -15.0% 32
Edwin Encarnacion 151 150 -2 -4.00% -5.00% -15.0% 32
Hank Conger 82 107 25 -5.00% -2.00% -14.0% 27
Brad Miller 87 105 18 -3.00% -7.00% -13.0% 25
Ruben Tejada 89 95 6 -3.00% -6.00% -13.0% 25
Nelson Cruz 137 158 21 -3.00% -4.00% -12.0% 34
Andrew McCutchen 169 146 -23 -3.00% -5.00% -12.0% 28
David Peralta 110 138 28 -1.00% -8.00% -11.0% 27
David Ortiz 134 138 3 -4.00% -2.00% -11.0% 39
Chris Carter 122 101 -21 -1.00% -9.00% -11.0% 28
Eugenio Suarez 86 105 19 -1.00% -8.00% -10.0% 23
Derek Norris 122 98 -24 -2.00% -6.00% -10.0% 26

I would like to thank those who commented in the original thread which helped with these adjustments. Finally, let me know if there are any other suggestions.

We hoped you liked reading Part 2 of Using Contact% & Pull% to Predict a Batter’s Decline by Jeff Zimmerman!

Please support FanGraphs by becoming a member. We publish thousands of articles a year, host multiple podcasts, and have an ever growing database of baseball stats.

FanGraphs does not have a paywall. With your membership, we can continue to offer the content you've come to rely on and add to our unique baseball coverage.

Support FanGraphs




Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won three FSWA Awards including on for his MASH series. In his first two seasons in Tout Wars, he's won the H2H league and mixed auction league. Follow him on Twitter @jeffwzimmerman.

newest oldest most voted
AA1989
Member
AA1989

How much variance in wRC+ is predicted by each component?