Tuesday, March 5, 2024

An introduction to pwOBACON+ (Pitchers)

I’m proud to introduce a third predictive metric for pitchers: pwOBACON+ (predictive weighted on-base average on contact plus).

In case you missed them, the first two were pK%+ and pBB%+.

wOBA (weighted on-base average) is a statistic developed by Tom Tango, the Senior Database Architect of Stats for MLB Advanced Media.

The metric is calculated on a per plate appearance basis like OBP (on-base percentage). wOBA does exclude intentional walks (so do pK%+ and pBB%+) and is not adjusted for park or league.

Plate appearance outcomes are weighted by their typical run values (home runs with the highest, triples with the second highest, doubles with the third highest, and so on).

wOBACON is simply the weighted on-base average on the balls a hitter puts into play.

A well-accepted concept in the sabermetrics realm is the idea that pitchers, for the most part, have little control over the contact they allow. That’s why FIP (fielding independent pitching) was wrought into existence. FIP only accounts for strikeouts, walks, hit-by-pitches, and home runs. xFIP (expected fielding independent pitching) goes even further by estimating the number of home runs a pitcher should have allowed by multiplying a pitcher’s fly balls allowed by the league average HR/FB rate.

Here is a graph of wOBACON+ (weighted on-base average on contact divided by league average wOBACON multiplied by 100) for consecutive player-seasons of at least 250 total batters faced from 2015 to 2018…

As you can see, wOBACON+ is very inconsistent year-to-year (the correlation is about 0.1).

pwOBACON+ is my attempt to craft a more stable metric that better predicts wOBACON+ in season n+1 than wOBACON and xwOBACON (estimated weighted on-base average on contact).

Predictive weighted on-base average on contact plus looks at four Statcast batted ball classifications:

  • Barrels (must be hit with an EV of at least 98 mph | LA requirement is dependent on the exit velocity)
  • Solid Contact (borderline barrels)
  • Flares/Burners (flares are bloopers, and burners are hard-hit grounders)
  • Poorly/Under (medium EV fly balls and popups)

David Hess and Clay Buchholz surrendered barrels at the highest rate in 2019 (13.2% of batted ball events)

The pitcher with the highest solid contact percentage was Edwin Diaz at 11.6.

The pitcher with the highest flare/burner percentage was Tyler Chatwood at 33.2.

The pitcher with the highest poorly/under percentage was John Brebbia at 43.3.

Here are the four types of contact shown above sorted in descending order by wOBACON (these numbers might be off slightly due to Savant search acting up)…

  1. Barrels (1.446)
  2. Solid Contact (.662)
  3. Flares/Burners (.619)
  4. Poorly/Under (.091)

All four were converted into Z-scores, and coefficients/weights for each standardized score were determining by regressing wOBACON for consecutive player-seasons of 250+ TBF for years 2015 to 2018.

Combined Z-score is approximately equal to…

(Barrel Z-score * 0.18) + (Solid Contact Z-score * 0.22) + (Flares/Burners Z-score * 0.3) + (Poorly/Under Z-score * 0.3)

I would’ve guessed barrels would have received more weight than solid contact, as they are more stable, but the regression tells a different story.

I’m not shocked that flares/burners and poorly/under have a greater emphasis than barrels and solid contact because the former types of contact occur on a far more frequent basis.

% of total BBEs in 2019

  1. Poorly/Under (25.1)
  2. Flares/Burners (24.7)
  3. Barrels (7.4)
  4. Solid Contact (5.9)

Poorly/under batted ball events often results in outs; with that being said, the equation views them as a negative. While I obviously don’t know this for a fact, I would think that this is the case because they are balls hit in the air, and if they were hit harder and/or at a lower LA, the outcome could have been far worse for the pitcher.

Another thing I’d like to touch on is why poorly/topped and poorly/weak contact were excluded.

One reason why poorly/topped dropped out of the equation is because it has a strong relationship to poorly/under (the graph shown below is for 2019).

Poorly/weak is an infrequent occurrence (4.0% of all BBEs last year) and is not as stable YoY as barrels, flares/burners, and poorly/under are.


Once I had combined Z-scores for all single-seasons of at least 250 total batters faced during the Statcast era (2015-2019), I regressed the Combined Z-scores in season n against wOBACON+ marks for season n.

The line of best fit has the equation y = 9.0498x + 99.146. Plugging in a Combined Z-score into the equation will result in a pwOBACON+ value.

Considering how variable wOBACON+ can be for pitchers, it is no surprise that the equation produces values that are much closer to the mean wOBACON+ (100).

pwOBACON+ is a much more consistent stat than wOBACON and xwOBACON.

The correlation was about 0.55 for my out-of-sample testing (2018 to 2019 consecutive player-seasons).

For xwOBACON, it was about 0.25, and for wOBACON, it was about 0.1.

Predictive power (correlation to 2019 wOBACON)

  1. pwOBACON (0.22)
  2. xwOBACON (0.14)
  3. wOBACON (0.1)

Here’s a graph with 2018 pwOBACON+ as the x-variable and 2019 wOBACON+ as the y-variable.

Pitchers with the highest pwOBACON+ last year

  1. Josh Hader (109)
  2. Edwin Diaz (109)
  3. Chad Green (108)
  4. Nick Anderson (108)
  5. Raisel Iglesias (108)

Lowest pwOBACON+ last year

  1. Aaron Bummer (82)
  2. Luke Jackson (85)
  3. Alex Claudio (87)
  4. Jared Hughes (88)
  5. Hector Rondon (88)

Pitchers with the biggest difference between pwOBACON+ and wOBACON+


  1. Brandon Workman (99/57)
  2. Pedro Baez (100/70)
  3. Yusmeiro Petit (102/78)
  4. Jordan Yamamoto (105/83)
  5. Roberto Osuna (99/80)

Smallest difference


  1. Edwin Diaz (109/145)
  2. Luke Jackson (117/85)
  3. Edwin Jackson (105/134)
  4. Nick Kingham (102/128)
  5. Carlos Carrasco (105/128)

Lowest single-season pwOBACON+ since 2015

  1. 2016 Zack Britton (75)
  2. 2015 Zack Britton (76)
  3. 2019 Aaron Bummer (82)
  4. 2017 Scott Alexander (82)
  5. 2016 Jeurys Familia (82)


  1. 2018 Chad Green (110)
  2. 2016 A.J. Griffin (110)
  3. 2016 Michael Tonkin (109)
  4. 2015 Matt Cain (109)
  5. 2016 Phil Hughes (109)

You can access the 2019 leaderboard for pwOBACON+ here.

Note: regression amount same for all players (volume not considered); I want it to be a stat like wOBACON and xwOBACON