clock menu more-arrow no yes mobile

Filed under:

Revisiting a new stat to evaluate hitters

How PHAMwOBA is holding up after another year of data

The year is 8560 and I am still using Tommy Pham in VEB article feature images.
Photo by Dilip Vishwanat/Getty Images

Last year, I tinkered around with several metrics to arrive at PHAMwOBA, a catch-all figure designed to evaluate hitters’ underlying level of success at the plate. To understand the rationale behind the stat I highly recommend that you read over the original article before proceeding (I really do mean that as my summary of it in this post is incredibly brief), but PHAMwOBA at its core takes three factors into account: quality of contact, a player’s speed, and the home ballpark he plays in. The former two are calculated using Statcast metrics–expected weighted on-base average (xwOBA, which looks at a batter’s launch angle and exit velocity numbers) and sprint speed–while I used FanGraphs’ park factors for the latter.

PHAMwOBA had 11.4% more predictive power than the unaltered version of xwOBA from 2015 to 2017, but I wanted to check in on its efficacy with a fourth full season of publicly-available Statcast data now at our disposal. A couple technical notes about the nitty-gritty formulas themselves:

  • For the sake of user-friendliness the new overall formula for PHAMwOBA=xwOBA+Park Adjustment+Speed Adjustment
  • The updated formula for Speed Adjustment=0.00414*SprintSpeed-0.11178, where sprint speed is measured in feet per second
  • The updated formula for Park Adjustment=8.43E-4*ParkFactor-0.0843. Note that the park factor value to plug in is now expressed like it would be on FanGraphs. (So insert 96 for Busch Stadium in 2018 instead of .96 like the previous version of PHAMwOBA called for.)
  • Speaking of the park adjustment, PHAMwOBA now uses FanGraphs’ three-year park factor numbers as opposed to the single-season ones.

Of the 115 players with at least 400 plate appearances in 2017 and 2018, this new PHAMwOBA had 9.7% greater correlation in predicting 2018 wOBA–the hitter’s actual level of production–relative to using their 2017 xwOBA instead.

To see if I could improve upon PHAMwOBA’s ability to project future wOBA, I added the following age adjustment to account for the fact that sprint speed regresses as a player ages: Age Adjustment=(-0.487+0.107*NewAge-0.00408*NewAge^2)-(-0.487+0.107*PreviousAge-0.00408*PreviousAge^2)

Whenever using PHAMwOBA for year-to-year purposes, add that age adjustment onto their sprint speed from the previous season to find a revised PHAMwOBA that had 16.1% stronger correlation in 2018 than raw xwOBA.

As far as this season goes, which Cardinals stand to gain the most by looking at PHAMwOBA instead?

2019 Cardinals: PHAMwOBA vs. wOBA

Player PHAMwOBA wOBA Difference
Player PHAMwOBA wOBA Difference
Jose Martinez 0.394 0.342 0.052
Marcell Ozuna 0.395 0.348 0.047
Harrison Bader 0.374 0.355 0.019
Jedd Gyorko 0.232 0.218 0.014
Matt Carpenter 0.328 0.320 0.008
Dexter Fowler 0.366 0.362 0.004
Yadier Molina 0.301 0.298 0.003
Paul Goldschmidt 0.348 0.348 0.000
Paul DeJong 0.389 0.393 -0.004
Kolten Wong 0.298 0.309 -0.011
Matt Wieters 0.313 0.343 -0.030
Yairo Munoz 0.255 0.330 -0.075
Tyler O'Neill 0.203 0.292 -0.089
Lane Thomas 0.299 0.487 -0.188

This should mostly be taken a good sign, as the only regular starter for St. Louis who has been noticeably “lucky” offensively thus far is Kolten Wong.

I’ll likely continue toying with various ways to refine PHAMwOBA; one idea I already have is to add some sort of adjustment for batted ball type, as speed is more valuable on a ground ball than a fly ball, for example. As always, I’m all ears if you have any suggestions or ideas for what to test.