New Baseball Win/Loss Metric: An Offense-Independent Approach

If a pitcher pitches nine innings and gives up just one run, but his team loses the game 1-0, he gets credited with a loss. Another pitcher pitches five innings and gives up five runs but leaves the game with his team ahead 7-5. When his team wins the game, he gets credited with a win. Another starting pitcher throws four innings of no-hitter ball but leaves after the fourth inning because of an injury. No matter what happens next, he cannot get credit for a win. If a relief pitcher starts the top of the ninth with his team ahead 4-3 and gives up an earned run, but his team scores a walk-off in the bottom of the ninth, he gets credit for a win.

Clearly, the current win/loss (W/L) procedure [1] is flawed. We propose a potential new way of calculating W/L records for pitchers given their performance independent of their team’s offensive performance.

Background

Traditionally, the W/L record was one of two key measures of pitcher quality, along with earned run average (ERA). The pitcher was the focal point of the team; thus, assigning credit to the pitcher for a win or loss was natural. Pitchers often pitched a complete nine-inning game and were therefore “responsible” for the outcome of the game. Cy Young, who played over 100 years ago, won 511 games, the all-time record for wins. Of the 815 games he started, he completed 749 of them.

Today, pitchers rarely complete games. Analytics from the Society for American Baseball Research (SABR) [2] created game strategies that limit the number of innings pitched to less than needed to complete a game.

The traditional W/L record is an important part of baseball lore. For the devoted analytics fan, there have been several improved pitcher metrics (e.g., WAR – wins above replacement, FIP – fielding-independent pitching, BABIP – batting average of balls in play) that measure performance far better [3]. But the fans, as well as general managers (GMs), should have a W/L procedure that is fan-friendly and logical. All too often, the “win” goes to the pitcher who was the last pitcher in the game when the winning team took the lead. So, pitchers can pitch most of the game, leave a close game with their team behind, be relieved by another pitcher and watch as their team then scores enough runs to win, the win going to the relief pitcher.

This article tries to provide a new way of assigning credit that compensates for four of the weaknesses of the current system.

Weaknesses of the Current System

First, pitchers today rarely complete their games. It is not unusual for a team to use four or more pitchers in a game, some pitching less than an inning, some throwing only one pitch, yet still being in contention for the win. How much does each pitcher contribute?

Second, the W/L assignment is not offense-independent. A pitcher might earn a win because his team had a great offensive performance that day or, vice versa, be charged with a loss if the team’s offense performed poorly.

Third, defense factors in how many runs a team gives up. Unearned runs are not credited to a pitcher’s ERA, but they absolutely factor in the current assignment of W/L.

Finally, many appearances by starting pitchers result in a “no-decision,” especially in situations in which they pitched well. Many superb starting performances go completely unrecognized by the traditional measure.

Data

To test the efficacy of alternative W/L methods, we compiled the records of every outing by every pitcher in Major League Baseball (MLB) for 2023. The dataset includes the records of 881 pitchers (213 starters and 668 relievers), with 21,013 lines of outings. The innings pitched and earned runs allowed are the two key factors used. The dataset includes 2023 playoff game appearances. As such, there are differences in some cases between the season records reported here and those that report only regular season games.

New Won/Loss

In this proposal, the designation of wins and losses is a function of the number of innings thrown and the pitcher’s ERA for that game. For 2023, the MLB average ERA was 4.31. If the game ERA for that game was below the league average, the pitcher is entitled to a portion of the “win.” If higher, the pitcher is charged with some portion of a “loss.” We are not considering whether the team won or lost the game, just how this pitcher performed.

Second, the portion of W/L is the number of innings pitched divided by nine. For example, a pitcher who had a game ERA of 3.0 (a “win” because it is less than the average) and completed six innings would get credit for two-thirds of a win. A pitcher who had a game ERA of 5.4 (a “loss”) and pitched five innings would get credit for 5/9 of a loss.

Note: This approach considers all pitchers to be the same whether they started the game or came in for relief. Also, all runs are considered the same. The first run is worth the same as the last. All innings are the same.

According to this system, all pitchers in a game, for both teams, could be credited with some portion of a win or some portion of a loss. Pitchers could get credit for some portion of a win while others on the same team, some portion of a loss. Because the assignment of W/L in this system is independent of the outcome of the game, a pitcher on the losing team could get credited with a portion of a win and a pitcher on the winning team could get credited with some portion of a loss. It depends on how they pitched, not how the team did.

With this approach applied to the pitchers in the opening paragraph, the pitcher who pitched nine innings of one-run ball (ERA equal to 1.0) would get credit for 1.0 win even though his team lost. The pitcher who threw five innings and gave up five runs (ERA equal to 9.0) would get credit for 5/9 of a loss even though his team won. The pitcher who pitched four innings of no-run ball (ERA equal to 0.0) would be credited with 4/9 of a win regardless of whether his team won or lost. The relief pitcher who gave up one run in one inning (ERA equal to 9.0) would get credit for 1/9 of a loss even though his team eventually won the game.

A perfect example of the power of this technique is to use it on Jacob deGrom’s 2018 performance with the Mets. He won the Cy Young Award that year with a 10/9 traditional W/L record for 32 appearances, an unusually weak W/L record for a Cy Young winner. In those 32 appearances, he gave up zero earned runs eight times, one earned run 10 times, two runs five times, three runs five times and four runs once – a spectacular season. The absurdity of the traditional method is that he had 14 outings of seven or more innings with zero or one earned run, yet in those games he “won” only five games and actually “lost” one, with eight “no-decisions.” His performance according to the traditional method looked anemic because the Mets offense gave him no run support.

Using this new method, his W/L record would have been 22.1/2.0. By this method, he was a 20-game winner and only “lost” two games of his 32 starts. Clearly, it was a Cy Young year, but you couldn’t see it from the traditional W/L method.

Results

All Pitchers

Table 1 shows the results for the top 30 pitchers based on total new wins according to this procedure. Starters get more win credit with this method. The increase in wins comes from the “no-decision” situations in which they pitched well but left the mound before the game was decided. Some also had a reduction in losses for games in which they got little run support. Relievers also get a boost in wins, most likely for good outings that resulted in “saves” or “holds” rather than wins.

Of the 881 pitchers, 616 saw their records improve (either more wins, fewer losses or a net positive combination). Of the 213 starters, 153 saw their records improve. Of the 668 relievers, 463 improved.

Table 1. Top Pitchers by Total New Wins

Gerrit Cole, the 2023 Cy Young winner, improves his record from 15/4 to 19.7/3.5. He clearly had a great year and, by the new method, came about as close as possible to winning 20 games. Logan Webb went from 11/13 (a mediocre record) to a new record of 18.8/5.2 (an excellent record). Similarly, Jordan Montgomery’s new record paints a very different picture than the traditional method: 13/12 to 17.7/6.7. Sonny Gray went from 9/9 to a very respectable 14.9/6.5. Framber Valdez goes from 12/14 to 16.3/7.0, a very different picture.

On the other hand, 20-game winner Spencer Strider went from a traditional 20/7 to a new 14.5/7.6, still respectable but not as good as Cole.

Another way of sorting the results is to list the pitchers by most improvement. Logan Webb had a total improvement of 15.6 games. Zack Greinke astoundingly went from a traditional record of 2/15 to a new 7.6/8.2, not a good year but not a wipeout as it first appeared.

Table 2. Top Total Improvement for Starters

Relievers

The traditional W/L method is really not designed for relievers. Most often, relievers are charged with either a save, blown save, hold or loss. Rarely, they get credit for a win, and when they do, it might be for the wrong reason, as previously mentioned . This method gives a reliever credit for some portion of a win for those outings when they shut out the other team.

Table 3 shows the top relievers as measured by their W/L improvement under this method.

Table 3. Top Total Improvement for Relievers

Michael King went from a W/L record of 4/8 to a new record of 9.3/2.3. He picked up a portion of a win each time he pitched a scoreless outing. In 31 of 49 appearances, he gave up no runs. So, in general, relievers get credit for “no-run” outings and get hurt when they give up any runs.

Note: On rare occasions, pitchers have an outing without getting an out. Although not included in this analysis, adjustments can be made for these rare outings.

Are All Runs the Same?

Who really “loses” the game? The reliever who gives up one run in a 10-9 loss or the starter who gave up the first nine runs? This analysis was based on the notion that each run, no matter when or under what circumstances it was scored, contributes the same toward a loss. Can it be proved that a run in the late innings of a close game is more important than an early run? Until science comes up with a method for measuring pressure (heart rate?), the idea that all runs are the same is not unreasonable.

But if you don’t buy that argument, alternate methods can be proposed. (Stay tuned.)

Conclusions

The W/L method proposed here has several significant advantages over traditional W/L, particularly for starting pitchers. There are no “no-decision” outings in this method. This is especially important for giving credit for good outings with no run support, which happen often.

This W/L method rewards good pitchers on teams with poor offensive showings. It “punishes” poor pitchers on teams who score plenty of runs. Of course, as mentioned, there are more sophisticated measures of pitching performance created by SABR analytics experts. However, those are a bit esoteric and not easily understood by the typical fan (although fans are getting more and more sophisticated). This method provides fans with a superior picture of a pitcher’s contribution throughout each season. In addition, although designed with the fan in mind, this method is sufficiently structured to give GMs a powerful way of judging pitcher quality.

Acknowledgment

Research assistance for this article was provided by Jacob Faber, Ariel Mansano and Jonathan Tordjman (Yeshiva University); and Austin Feit, Joseph Rubin and Udhav Goenka (NYU Stern).

References

1. https://en.wikipedia.org/wiki/Win%E2%80%93loss_record_(pitching)
2. The Society for American Baseball Research, https://sabr.org/ 
3. https://www.mlb.com/glossary/advanced-stats