clock menu more-arrow no yes mobile

Filed under:

An Alternative to Pitcher-of-Record Wins & Losses

The Diamondback starters’ performances were very much improved this season through 15 September compared to last season.

Zac Gallen
Zac Gallen.
Photo by Justin Edmonds/Getty Images

I tip my hat to Jim McLennan because the last paragraph of his AZ Snake Pit Game Preview inspired me to look for a measure that could replace the traditional pitcher-of-record win-loss decisions for starting pitchers. As usual, I will focus on the Diamondbacks.

This season’s starters performed better than last season.

As a team, starting pitching improved compared to last season. Three of the following measures showed significant improvement per Baseball Savant (data through 15 September):

  • Allowed home runs per PA improved from .040 to .030.
  • Allowed hits per PA improved from 232 to .219.
  • Allowed walks per PA improved from .081 to .075.
  • Strikeouts per PA changed slightly from .203 to .200.

This season, defense was important in all games. All but two games had more balls in play (when defense is important) than three true outcomes (strikeouts plus walks plus homers). The two exceptional games were Merrill Kelly on 8 April and Zac Gallen on 21 June. Data from Baseball Savant. Caleb Smith’s 2 inning start was not considered.

This season, Zac Gallen and Merrill Kelly were the best starters, with Davies close behind

  • Their ERAs (2.5 Gallen, 3.01 Kelly, and 4.06 Davies) were better than the 4.43 average ERA (starters and relievers) in the Majors. Data from Baseball Reference through 15 September.
  • Excluding two newly promoted starters (Ryne Nelson and Drey Jameson), only Zac Gallen and Merrill Kelly had RA9 better than their RA9average. Gallen’s RA9 was 2.72 vs 4.30 RA9ave and Kelly’s RA9 was 3.11 vs 4.37 RA9ave.

RA9ave is total runs a theoretical average pitcher would allow against the opponents faced, in the parks played in, and with the Diamondbacks defense per Baseball Reference.

Which pitchers had better average game scores this season compared to last season?

Eight pitchers started at least one game for the Diamondbacks in both this season and last season.

The following pitchers’ average game scores improved:

  • Zac Gallen improved from 51.7 to 61.5
  • Merrill Kelly improved from 50.2 to 57.8
  • Humberto Castellanos improved from 43.8 to 45.2
  • Corbin Martin improved from 28.7 to 38.3

The following pitchers’ average game scores were less:

  • Madison Bumgarner went from 50.3 to 45.5
  • Tyler Gilbert went from 54.7 to 45.3
  • Caleb Smith went from 43.5 to 20.0
  • Luke Weaver went from 50.4 to 11.6

Game Score is a better performance measure than pitcher-of-record wins and losses.

Pitcher-of Record: An over-simplified description. A starting pitcher is charged with a win when he pitches at least 5 innings and his team maintains a lead that it never relinquishes after his last half-inning. A starting pitcher is charged with a loss when he gives up the go-ahead run and the other team never relinquishes the lead.

Looking at starting pitchers, several weaknesses of pitchers-of-record wins and losses follow.

  • Looking at all pitchers not named Gallen, Kelly, Davies, or Bumgarner, only 40% of Diamondback starts qualified that pitcher for a pitcher-of-record win due to the 5 inning requirement.
  • When a starter leaves the game with the lead, the bullpen often relinquished the lead. When that happened the bullpen took away a starter’s well-earned pitcher-of-record win.
  • As much as, or more than the starter’s pitching performance, a very large win-loss decider is how many runs the Diamondbacks score, and when they score those runs. That obscures and reduces the value of pitcher-of-record.
  • Only one measure of pitcher performance is incorporated : runs scored. It is a top-level measure.

Looking at starting pitchers, the corresponding strengths of game scores follow:

  • Every time a starter pitches, he receives a game score, potentially earning a game score win or loss.
  • Except in the case that the starter leave the game with runners on base, the bullpen cannot take away the starters game score win.
  • Game score wins or losses look at the performance of the pitcher and the batters he faces.
  • Game Score calculates how well a starting pitcher performed. The calculation includes most of the factors we looked at when deciding that the Diamondback starters performed better this season than last season. Examples are homers, hits, walks, and strikeouts. Instead of plate appearances, it looks at innings pitched.

For starters, how could game scores be used to decide game score wins and losses?

For starting pitchers, I propose a metric to replace pitcher-of-record wins and losses:

  • When a game score is 60 or above, it is a game score win (GS-Win).
  • When a game score is 40 or below, it is a game score loss (GS-Loss).
  • Game score will be the average of three methods: Baseball Reference, FanGraphs, and 538.com.

The following table shows wins and losses for the game score method and for the pitcher-of-record method.

Diamondbacks’ starters, 2022 through 15 September. Data from Baseball Savant.

This season’s game score wins and losses show Diamondbacks starters are very much improved compared to last season.

Looking at the totals for this season through 15 September to last season shows:

  • Game score wins increased from 38 to 45. More wins could be added in the last few weeks of the season.
  • Game score losses decreased from 59 to 31. Although more losses could happen, the losses will be much lower than last season.

Unintended consequences of using game score to determine wins and losses.

Relief pitchers are no longer pitchers of record. That is not a problem two reasons.

  • Relievers are credited with holds and saves, which a starter cannot earn. Holds and saves are important for relievers. Perhaps additional measures, such as Nate Silver’s goose eggs could become more prominent.
  • Who looks at a relief pitcher’s wins and losses? I have not. Is the pitcher-of-record win or loss meaningful for relievers?

Both team’s starters could be winners or both team’s starters could be losers. That is not a problem for two reasons.

  • Both starters winning could aptly describe a pitching duel, in which case both pitchers are deserving.
  • Both starters losing could aptly describe a situation where both pitchers would have lost against an average performance, but instead faced a less challenging situation.

Summary.

This season the Diamondbacks’ starters allowed less homers per PA, hits per PA, and walks per PA. Season average game score improved for Zac Gallen, Merrill Kelly Humberto Castellanos, and Corbin Martin.

Game score is a better measure than pitcher-of-record wins and losses for the following reasons:

  • In every game the starter gets a game score, the bullpen cannot take away a starters win unless the starter leaves runners on base.
  • Game score looks at the performance of the pitchers and the batters they face without being impacted by runs scored by the Diamondbacks.
  • The calculation of game score includes factors that we looked at in determining the Diamondbacks did better this season than last season.

The proposed criteria for wins and losses follows:

  • When a game score is 60 or above, it is a game score win (GS-Win).
  • When a game score is 40 or below, it is a game score loss (GS-Loss).

The Diamondback starters’ performances with the proposed alternative were very much improved this season through 15 September compared to last season.

  • Game score wins increased from 38 to 45.
  • Game score losses decreased from 59 to 31.

Two consequences of the proposed method follow:

  • Relief pitchers would no longer be pitchers of record. For relief pitchers, perhaps an additional measure could be added to holds and saves.
  • In a single game, both teams’ starting pitchers could earn game score wins, for example in pitching duel games.
  • In a single game, both team’s starting pitchers could earn game score losses.