Meet my new predictive metrics
Over the MLB All-Star break, I introduced expected pitching plus, a stat determined by expected swing percent plus, expected swing and miss percent plus, expected called strike percent plus, and expected barrels per expected contact swing plus. A necessary step in calculating xPitching+ was using generalized logistic regression to estimate a set of probabilities for each pitch (type):
- probability of a called strike (given batter takes pitch),
- probability of a swing,
- probability of a swing and miss (given batter swings), and
- probability of a barrel (given batter makes contact).
After I finished developing xPitching+, I wanted to see if the four probabilities above could strengthen my predictive metrics. It had been around two years since I had made changes to them, so an update was overdue.
One of the predictive stats is predictive walk percent. I tried 21 variables in a linear regression where the output variable was next-season’s walk percent (the inputs are from the previous season) .

The x_ stats are expected stats calculated from the xPitching+ logistic regressions, and heart; shadow; chase; and waste refer to Baseball Savant’s Attack Zones.

The final outline for pBB%…

Here are some graphs that show year-over-year correlations for the stat…




For batters and pitchers, predictive walk percent plus, which is predictive walk percent scaled so that 100 is league average, is more stable and predictive of future walk percent.
Another predictive metric is predictive strikeout percent. Here are the variables I tried…

Whiff percentages uses swings as the denominators.
The final outline for pK%…

Graphs




The third predictive stat is predictive weighted on-base average on contact, which communicates information on how much damage is done when the ball is put into play.

adjusted_ev, another Baseball Savant stat, is average exit velocity but exit velocities between zero and 88 mph are assigned a value of 88. weak, topped, under, solid, and barrel are Statcast classifications for batted balls (different launch angle and launch speed combinations).
The final outline for pwOBAcon…

Graphs




The gap between the lines is the greatest it has been, and that should not come as surprise! When it comes to strikeouts and walks, not as much randomness is at play as there is for average value on balls in play, especially for pitchers.
Two other predictive metrics are predictive bunt percent (percent of “plate appearances” ending in a bunt) and predictive bunt hit percent (bunt hits divided by bunt attempts).

Given the five predictive stats, one can calculate a predictive weighted on-base average.
Formula for 2025 batters
pwOBA = (pBB% * 0.692 + HBP% regressed * 0.723 + (100 – pBB% – pK% – HBP% regressed – pbunt%) * pwOBAcon + pbunt% * xbunt hit% / 100 * league bunt hit wobacon) / 100
Graphs for pwOBA




wOBA reveals how productive a batter has been at the plate (going off of box scores).
xwOBA is expected wOBA (it looks at BB%, K%, HBP%, and launch speed; launch angle; and, sometimes, sprint speed for batted balls).
pwOBA is predictive wOBA. It incorporates a multitude of measures to predict what the player’s wOBA will be the following season. How talented or skilled is the player?
Here are the top pwOBA+ batter-seasons (2020-25)…
- 2023 Ronald Acuña Jr.
- 2022 Yordan Alvarez
- 2024 Aaron Judge
- 2024 Juan Soto
- 2022 Aaron Judge
- 2024 Shohei Ohtani
- 2021 Ronald Acuña Jr.
- 2024 Bobby Witt Jr.
- 2021 Juan Soto
- 2021 Vladimir Guerrero Jr.
Bottom
- 2025 Jacob Stallings
- 2025 Martín Maldonado
- 2024 Martín Maldonado
- 2023 Tucker Barnhart
- 2024 Austin Hedges
- 2022 José Barrero
- 2021 Sandy León (as a lefty)
- 2023 Austin Wynns
- 2023 Martín Maldonado
- 2023 Austin Hedges
Basically all catchers.
The new pwOBA+ values are similar to the old ones, for the most part.

Graphs comparing the two in terms of correlations


Seems like the new pwOBA+ is more predictive but less consistent.
Here are the top pwOBA+ pitcher-seasons…
- 2021 Jacob deGrom
- 2022 Edwin Díaz
- 2021 Corbin Burnes
- 2024 Mason Miller
- 2022 Jacob deGrom
- 2021 Gerrit Cole
- 2022 Andrés Muñoz
- 2021 Liam Hendriks
- 2022 Emmanuel Clase
- 2023 Félix Bautista
Bottom
- 2023 Ty Blach
- 2024 Ty Blach
- 2023 Kyle Freeland
- 2023 Adam Wainwright
- 2021 Zach Davies
- 2024 Taijuan Walker
- 2024 Adam Cimber
- 2022 Madison Bumgarner
- 2023 Rich Hill
- 2023 Eric Lauer
Graphs



The new pwOBA impresses.
Here is how pwOBA+ stacks up against Steamer projections…




Here is how pwOBA+ compares to other pitching metrics…




Some combination of pwOBA+, Pitching+ (from FanGraphs), and projections would probably result in best predictions for pitchers.
In June, I posted a tweet where I looked at the players with the biggest differences between their wOBA+ and pwOBA+. The batter with the greatest negative difference was Andrew Vaughn. His wOBA with the White Sox was .233, his pwOBA .326. Interestingly, the Brewers traded for him.
Here was a graph for Vaughn at the time of the trade…
His graph through 8/21

With Milwaukee, he has posted an incredible .392 wOBA.
While my predictive metrics can’t tell the future, they are useful clues as to how a player might perform moving forward!
To view the Max’s Sporting Studio Leaderboards, click here.
Featured imaged from David Banks / USA TODAY Sports