Tuesday, March 5, 2024
AnalysisMLB

Introducing a new and improved pwOBAcon+ (Pitchers)

About ten months ago, I published an article introducing a pitching stat called predictive weighted on-base average on contact plus (pwOBAcon+). In the time since then, my capacity to code effectively in R(Studio) has increased significantly. When I first created this stat, I had to manually download countless Baseball Savant queries; it was a process that was extremely time-consuming. Luckily, I am now able to calculate pwOBAcon+ in an efficient manner due to the fact that I can compute it directly from Statcast pitch-by-pitch data. As soon as I realized there were minor differences in the values of the original variables, I decided to rerun the regression and try out some new inputs.

The three prevailing explanatory variables in the pwOBAcon+ formula are hard hit percent plus, flare/burner percent plus, and poorly/under percent plus.

In order for a batted ball to constitute as one that is hard hit, it must leave the hitter’s bat at an exit velocity of at least 95 mph.

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

  1. Jake Arrieta 2015 (86.8)
  2. Kenta Maeda 2016 (87.4)
  3. Clayton Kershaw 2015 (87.6)
  4. Ryan Yarbrough 2019 (87.9)
  5. CC Sabathia 2016 (88.2)
  6. Brandon McCarthy 2017 (88.6)
  7. Ryan Yarbrough 2018 (88.8)
  8. Noah Syndergaard 2018 (89.0)
  9. Aníbal Sánchez 2019 (89.0)
  10. Zack Wheeler 2018 (89.2)

Here are the highest

  1. Adrian Sampson 2019 (112.0)
  2. David Hess 2019 (111.5)
  3. Matt Moore 2017 (110.6)
  4. Robbie Ray 2017 (110.5)
  5. Marcus Stroman 2016 (110.5)
  6. Sean Manaea 2017 (110.5)
  7. C.J. Wilson 2015 (110.2)
  8. Tyler Duffey 2016 (110.2)
  9. Iván Nova 2016 (110.2)
  10. Ricky Nolasco 2017 (110.2)
Clip from Baseball Savant

Hard hit percent plus is simply hard hit percent (tracked hard hits divided by tracked batted ball events) divided by the league average hard hit percent, multiplied by 100, and regressed based on the number of tracked batted ball events. League average for all plus stats is 100.

The other two variables in the equation are flare/burner percent plus and poorly/under percent plus, both of which are types of quality of contact from Statcast.

Flares and burners are distinct from each other. One can think of a flare as a line drive of sorts. In fact, over 60% of the line drives hit have fallen under the category flare/burner in the 2021 regular season. A burner, on the other hand, is simply a ground ball that is well-struck.

Here are the lowest single-season flare/burner percent plus marks since 2015

  1. Brett Anderson 2015 (97.5)
  2. Marco Estrada 2015 (97.7)
  3. Garrett Richards 2015 (97.8)
  4. Ariel Miranda 2018 (97.9)
  5. Drew Pomeranz 2016 (98.1)
  6. Caleb Smith 2019 (98.2)
  7. CC Sabathia 2016 (98.2)
  8. Matthew Boyd 2018 (98.2)
  9. Marco Estrada 2018 (98.3)
  10. Masahiro Tanaka 2015 (98.4)

Here are the highest

  1. Clayton Richard 2017 (102.6)
  2. Miles Mikolas 2019 (102.2)
  3. José Fernández 2016 (rest in peace) (102.1)
  4. Mike Wright 2018 (102.1)
  5. Johnny Cueto 2017 (102.0)
  6. Homer Bailey 2018 (102.0)
  7. Phil Hughes 2015 (102.0)
  8. Mike Leake 2018 (101.9)
  9. Jeff Samardzija 2015 (101.8)
  10. Kyle Gibson 2019 (101.8)
Clip from Baseball Savant

The third and final component of pwOBAcon+ is poorly/under percent plus. These are high launch angle batted ball events. Only 3.7% of these batted ball events have gone for extra-base hits this year.

Here are the lowest single-season poorly/under percent plus marks since 2015

  1. Brett Anderson 2015 (49.3)
  2. Scott Alexander 2017 (49.6)
  3. Brad Ziegler 2018 (51.0)
  4. Zack Britton 2019 (51.1)
  5. Zack Britton 2015 (52.8)
  6. Scott Alexander 2018 (53.0)
  7. Zack Britton 2016 (54.0)
  8. Dan Otero 2017 (55.5)
  9. Clayton Richard 2017 (56.1)
  10. Marcus Stroman 2016 (57.0)

Here are the highest

  1. Chris Young 2015 (162.5)
  2. Tyler Clippard 2015 (161.0)
  3. Marco Estrada 2015 (159.9)
  4. Emilio Pagán 2017 (157.5)
  5. Ryan Buchter 2016 (153.6)
  6. Marco Estrada 2018 (151.4)
  7. Marco Estrada 2017 (150.0)
  8. Héctor Santiago 2015 (149.5)
  9. A.J. Griffin 2017 (149.5)
  10. Justin Verlander 2015 (149.2)
Clip from Baseball Savant

pwOBAcon+ comes from a linear regression where hard hit percent plus, flare/burner percent plus, and poorly/under percent plus in season n were the x-variables and non-bunt wOBAcon+ in season n+1 was the y-variable. A pitcher’s pwOBAcon+ value is what one would predict that pitcher’s non-bunt wOBAcon to be the following season given that pitcher’s hard hit percent plus, flare/burner percent plus, and poorly/under percent plus.

In the equation, all three independent variables have a positive coefficient. In other words, an increase in say hard hit percent plus from season n is associated with an increased in non-bunt wOBAcon+ in season n+1. For your information, season n refers to the current season, and season n+1 would be the subsequent season. Hard hit balls are obviously not ideal, and flares/burners go for hits more than half the time. I would’ve thought poorly/under events would be great for the pitcher. I think they are, but they may signal worse results on balls in play moving forward (maybe hitters are just missing extra-base hits).

Here are the lowest single-season pwOBAcon plus marks since 2015

  1. Brett Anderson 2015 (89.2)
  2. Zack Britton 2016 (90.0)
  3. Jake Arrieta 2016 (90.2)
  4. Dallas Keuchel 2015 (90.5)
  5. Matt Bowman 2016 (91.5)
  6. Zack Britton 2019 (91.9)
  7. Zack Britton 2015 (92.3)
  8. Alex Claudio 2017 (92.5)
  9. CC Sabathia 2016 (92.5)
  10. Tyson Ross 2015 (92.6)

Here are the highest

  1. Andrew Moore 2017 (107.4)
  2. Phil Hughes 2015 (107.1)
  3. Ricky Nolasco 2017 (106.9)
  4. Madison Bumgarner 2019 (106.5)
  5. David Hess 2019 (106.1)
  6. Emilio Pagán 2017 (106.0)
  7. Chad Green 2017 (105.9)
  8. Dylan Bundy 2016 (105.8)
  9. A.J. Griffin 2016 (105.8)
  10. Ryan Buchter 2016 (105.7)

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 pwOBAcon was of 2019 non-bunt wOBAcon compared to 2018 non-bunt wOBAcon and 2018 non-bunt xwOBAcon, one can clearly see that pwOBAcon is far more predictive of future non-bunt wOBAcon than non-bunt wOBAcon and non-bunt xwOBAcon are. The left barplot conveys that pwOBAcon far exceeds non-bunt wOBAcon and non-bunt xwOBAcon when it comes to stability.

The new pwOBAcon+ is more stable and predictive than the old one despite including fewer variables. Sometimes simpler is better!

If one ever wants to know if a certain pitcher might get better or worse results when it comes to hitters putting the ball into play, one’s best bet is to look at his pwOBAcon.

Featured image- Creator: Elsa | Credit: Getty Images