Tuesday, July 23, 2024
AnalysisMLB

Introducing a new and improved pBB%+ (Pitchers)

On Saturday, I wrote up an article introducing a new and improved pitcher pwOBAcon+. In this piece, I am going to detail a revised predictive walk percent plus (pBB%+) metric for pitchers. It is an improvement upon the old version.

There are four variables in the formula for pBB%+: strike percent plus, swing percent with pitcher ahead in count plus, batted ball events divided by pitches plus, and walk percent plus.

I did my best to exclude pitches that resulted in catcher interference, a sacrifice bunt, a missed bunt, a bunt foul-tip, a foul bunt, a swinging pitchout, or a foul pitchout when it was convenient to do so. Additionally, I removed all pitches thrown in plate appearances that ended in a bunt. I excluded pitchouts and intentional balls as well.

Here are the lowest single-season strike percent plus marks since 2015

  1. Tyler Chatwood 2018 (91.0)
  2. Wade Miley 2017 (92.9)
  3. Chris Tillman 2017 (93.2)
  4. Jace Fry 2019 (93.3)
  5. Adam Ottavino 2017 (93.7)
  6. Francisco Liriano 2018 (93.8)
  7. Tyrell Jenkins 2016 (93.8)
  8. Tyler Chatwood 2017 (93.8)
  9. Tyson Ross 2017 (93.9)
  10. Justin Anderson 2018 (93.9)

Highest

  1. Kenley Jansen 2017 (108.3)
  2. Pat Neshek 2017 (108.3)
  3. Max Scherzer 2015 (108.1)
  4. Sean Doolittle 2017 (107.4)
  5. Kenley Jansen 2016 (107.3)
  6. Addison Reed 2017 (107.0)
  7. Bartolo Colon 2015 (106.9)
  8. Chris Paddack 2019 (106.6)
  9. Justin Verlander 2018 (106.6)
  10. Miles Mikolas 2018 (106.6)

Here are the lowest single-season swing percent with pitcher ahead in count plus marks since 2015

  1. Yovanni Gallardo 2015 (80.9)
  2. José Quintana 2017 (83.4)
  3. José Quintana 2018 (84.1)
  4. Yovanni Gallardo 2018 (84.2)
  5. Mike Pelfrey 2017 (85.1)
  6. Wade Miley 2015 (85.3)
  7. Zack Greinke 2020 (85.4)
  8. John Means 2019 (85.9)
  9. José Quintana 2016 (86.1)
  10. Doug Fister 2016 (86.9)

Highest

  1. Blake Treinen 2018 (118.4)
  2. Javy Guerra 2019 (114.9)
  3. R.A. Dickey 2015 (114.5)
  4. Jonathan Papelbon 2015 (114.4)
  5. Pat Neshek 2017 (114.0)
  6. Kenley Jansen 2016 (113.7)
  7. Sean Doolittle 2017 (113.2)
  8. Josh Hader 2019 (113.0)
  9. Kevin Gausman 2019 (112.8)
  10. R.A. Dickey 2017 (112.4)

Here are the lowest single-season batted ball events divided by pitches plus marks since 2015

  1. Craig Kimbrel 2017 (70.7)
  2. Josh Hader 2018 (71.3)
  3. Dellin Betances 2017 (71.4)
  4. Aroldis Chapman 2018 (72.5)
  5. Corey Knebel 2017 (72.6)
  6. Aroldis Chapman 2015 (73.0)
  7. Matt Barnes 2019 (73.0)
  8. Kyle Barraclough 2016 (73.7)
  9. Dellin Betances 2015 (74.3)
  10. Craig Kimbrel 2018 (75.4)

Highest

  1. Zach Neal 2016 (127.0)
  2. Bartolo Colon 2018 (126.2)
  3. Mike Leake 2019 (125.9)
  4. Richard Bleier 2019 (125.8)
  5. Mark Buehrle 2015 (125.4)
  6. Brett Anderson 2019 (125.4)
  7. Bartolo Colon 2017 (123.7)
  8. Mike Leake 2018 (122.8)
  9. Iván Nova 2017 (122.6)
  10. Bartolo Colon 2017 (122.3)

Here are the lowest single-season walk percent plus marks since 2015

  1. Bartolo Colon 2015 (55.2)
  2. Clayton Kershaw 2016 (56.3)
  3. Josh Tomlin 2016 (56.5)
  4. Mike Leake 2019 (57.3)
  5. Josh Tomlin 2017 (58.1)
  6. Phil Hughes 2015 (59.2)
  7. Miles Mikolas 2018 (59.4)
  8. Hyun Jin Ryu 2019 (60.1)
  9. Zack Greinke 2019 (61.7)
  10. Jeff Samardzija 2017 (63.0)

Highest

  1. Tyler Chatwood 2019 (178.2)
  2. Trevor Rosenthal 2019 (148.3)
  3. José Leclerc 2017 (147.1)
  4. Dellin Betances 2017 (142.9)
  5. Tommy Kahnle 2015 (139.9)
  6. Carolos Rodón 2015 (138.0)
  7. Tanner Rainey 2019 (137.7)
  8. Robbie Ray 2020 (137.6)
  9. Jace Fry 2019 (137.6)
  10. Tayron Guerrero 2019 (136.4)

pBB%+ comes from a linear regression where strike percent plus, swing percent with pitcher ahead in count plus, batted ball events divided by pitches plus, and walk percent plus in season n were the x-variables and non-adjusted (not regressed) walk percent plus in season n+1 was the y-variable. A pitcher’s pBB%+ value is what one would predict that pitcher’s non-adjusted walk percent plus to be the following season given that pitcher’s strike percent plus, swing percent with pitcher ahead in count plus, batted ball events divided by pitches plus, and walk percent plus.

Increases in strike percent plus and batted ball events divided by pitches plus in season n are associated with a decrease in non-adjusted walk percent plus in season n+1. This makes sense intuitively, as a pitcher will never walk anybody if he only throws strikes, and a batted ball event ends a plate appearance.

In contrast, increases in swing percent with pitcher ahead in count plus and walk percent plus in season n are associated with an increase in non-adjusted walk percent plus in season n+1. More walks in the present corresponding to more walks moving forward is to be expected. I am not sure why the coefficient for swing percent with pitcher ahead in count plus is positive. One guess I have is that the pitchers who induce the lowest percentage of swings when hitters are behind in the count have above average command and are primarily getting called strikes on such pitches.

Here are the lowest single-season predictive walk percent plus marks since 2015

  1. Bartolo Colon 2015 (41.1)
  2. Phil Hughes 2015 (44.8)
  3. Miles Mikolas 2018 (47.4)
  4. Josh Tomlin 2016 (49.1)
  5. Mike Leake 2019 (50.0)
  6. Clayton Kershaw 2016 (50.8)
  7. Josh Tomlin 2017 (51.2)
  8. Bartolo Colon 2016 (52.5)
  9. Clayton Kershaw 2017 (54.0)
  10. Matt Strahm 2019 (54.5)

Highest

  1. Tyler Chatwood 2018 (171.3)
  2. Trevor Rosenthal 2019 (151.0)
  3. Jace Fry 2019 (149.7)
  4. José Leclerc 2017 (146.9)
  5. Dellin Betances 2017 (143.8)
  6. Justin Anderson 2018 (143.5)
  7. Adam Ottavino 2017 (143.1)
  8. Carlos Rodón 2015 (142.1)
  9. Tanner Rainey 2019 (141.5)
  10. Tyson Ross 2017 (140.7)

That list should look very familiar to the walk percent plus list considering walk percent plus correlates very strongly to predictive walk percent plus.

My model was built off of consecutive player-seasons of at least 250 batters faced from 2015-2018 (n = 541).

In looking at how predictive 2018 pBB% was of 2019 BB% compared to 2018 BB%, one can see that pBB% is sizably more year-to-year than BB% is. When it comes to predicting future walk rate, predictive walk percent is slightly more predictive of future walk rate than walk rate is.

The fact that predictive walk percent plus is more stable than non-adjusted walk percent plus makes it a great stat to look at early on in the season, and pBB%+ is more predictive of future walk rate than walk rate is… just not by much.

Featured image- Creator: Rich Schultz | Credit: Getty Images