Win, Lose or Save?
Published June 10th, 2007The Problem
Wins, Losses and Saves are possibly the most arbitrary statistics in all of baseball. At times, wins are given to pitchers who had poor games, but were saved by their offense driving in high amounts of runs. Other times, pitchers throw gems, only to be penalized by their offense’s poor output. And saves are given almost by simply showing up at the end of the game. None of these stats are tightly based on any sort of statistical analysis – they are given according to a set of rules that don’t take the quality of the pitcher’s performance into account.
The Solution
Win Probability Added and Leverage Index, I believe, are possibly strong tools for awarding wins, losses and saves. While there are strong criticisms of the use of WPA and LI in aggregate, they are excellent at telling us how well a player did in a given situation or game, and how much that performance was worth. By analyzing how well a pitcher performed in a game we can properly dole out wins, losses and saves – these are single game awards.
The Reasoning
Because WPA tells us how well a player improved his team’s chance of winning in given game, we can look to see which pitcher increased his team’s chances to win the most. We can also look at which pitcher hurt the team’s chances the most. That is the way to award wins and losses. It is the job of the pitchers to give up fewer runs than their team scores. It is not the pitcher’s fault if his team happens to score runs while he’s off the mound. As long as he turned in the best performance, he deserves the win. A pitcher should also be rewarded for performing above average for as many innings as possible.
A relief pitcher looking for a save shouldn’t be handed one just for coming in in the 9th up by three runs. The Save has been ridiculously watered down. And many times the most critical point in the game doesn’t come in the 9th. Sometimes a relief pitcher gets the team out of a jam in the 8th that was more threatening than anything that happened later. Yet the closer gets the save and the earlier reliever gets a pat on the back. Using WPA and LI, we can change that. We can look and see if any relievers came into the game in tight, critical situations and excelled. Those are the individuals that deserve the save – not just the guy who was last man standing on the mound.
So Wins for the best overall performance, Losses for the worst, and Saves for getting the team out of a nasty situation. One more idea I’m adding here – a reliever can get a save and a win. If he’s the best pitcher in the game and came in and got out of a nasty jam, he deserves the accolades.
The Method
To test out this theory, I downloaded 2006 event data from baseball-reference.com. I generated Win Probability Added and Leverage Index using tables from baseballgraphs.com. Everything was adjusted for Run Environment per park. I computed Leverage Index using the quick-and-dirty method outlined in the WPA spreadsheet – simply found the effect a strikeout would have on the given situation and divided the effect by 0.27. I spot-checked my generated data with fangraphs.com and most players/events lined up very closely. I know they use slightly more sophisticated ways to determine WPA and LI.
I then scored every player’s performance in a game. I used a very simple formula: WPA x pLI x IP. WPA gives the measure of the player’s performance. I multiplied that by pLI because that tells us how important the situations were when the player achieved his WPA. I then multiplied that by IP – pitchers should be rewarded for doing well over more innings. That is important.
I then took the winning team and plucked out the best pitcher. He got the win. I then plucked out the worst pitcher on the losing team. He got the loss. Saves was slightly more complicated. I looked, of course, only at relievers. I took the best reliever who had a pLI of 2.0 or more. Thus, I defined pLI’s of 2.0 or more as “critical performances†worthy of save consideration. Whoever had the best performance with a pLI of 2.0 or greater got the save.
Note: Unlike current, official baseball rules, I allowed a reliever to be credited with both a win and a save. If he was the best pitcher in the game, and he handled the most critical situations with aplomb, he deserves both accolades. It is also of note that I found 6 games where no pitcher deserved a win, and 81 games where no pitcher deserved a loss. I didn’t expect this, but it certainly raises some interesting thoughts, especially considering how lopsided that is.
The Results
What did I discover? That the current arbitrary method of awarding wins and losses – at least in 2006 – wasn’t all that bad after all, for top starting pitchers at least. Among starters, John Koronka looks to have had the worst luck – he had 7 wins but deserved 11. Joe Blanton was the biggest thief, he had 16 wins last year but only deserved 10. On the other side of things, Jason Bergmann only had 2 losses on his record last year, but deserved 6. He also only started 6 games. By far Ted Lilly had the worst rap, credited with 13 losses but only deserving of 4.
The real interesting part of this little work has been saves and reliever wins. Closers saw a large decrease in save numbers, but a large increase in wins. I think I’m OK with this – the current save rule really waters down the stat, almost to the point of making it meaningless. WPA/LI analysis also rewards middle relievers who did a really good job – moreso than the “hold†statistic does. A lot of setup men saw large bumps in their win and save totals. Regardless of whether this method does a better job than the official method or not, I think it’s time to look at the save category and rethink it. I do know that some people have looked at awarding saves by WPA and gLI (leverage index upon entering the game, so there’s no rewarding self-made drama) and there may be some validity to that method as well.
It should be no surprise that the players who lost the most saves by this method were players on good teams. They simply came in for a lot of those sorta cheap one inning saves, because they were on good teams. Francisco Rodriguez went from 47 saves to 15, Trevor Hoffman from 46 to 16, Todd Jones from 37 to 8, and Billy Wagner from 40 saves to 11. That isn’t to say these players were not valuable, however. K-Rod gained 6 wins (2 to 8), Hoffman 7 wins (0 to 7), Jones 5 (2 to 7) and Wagner 8 wins (3 to 11).
At the other end were players who got silent saves. Scot Shields went from 2 official saves to 12 deserved saves. Duaner Sanchez received 8 saves by the method, as opposed to 0 official saves. Chad Qualls deserved 7 saves, but received none, and Todd Coffey earned 15 saves instead of the 8 he was credited for.
Top 10 in Wins
Top 10 in Saves
Discussion & Conclusion
My goal here has been to try and look at a better way to award Wins, Saves and Losses. They should be given to the individuals who deserved them, based on merit – not based on who was on the mound when a go-ahead run was scored. I plan to look at more seasons and see if the trends in 2006 hold up. The goal of the Win, Loss and Save are individual awards for individual games. That is why I did not aggregate WPA and LI over the season, and then make an extrapolation of record. Since WPA and LI are highly situational, I felt it made sense only to evaluate them in that context. There are many statistics that have been developed to measure overall performance of pitchers – but I think this WPA*LI*IP is a good way to measure individual performance of a pitcher in a single game.
The only part I feel iffy on in all this is my arbitrary decision to call pLI’s of 2.0 or greater as critical relief appearances. I think my next step is to work on maybe finding a more concrete, data-supported definition of a critical relief appearance.
You can download player records - both their official and records generated with my formula, here.
(No Ratings Yet)
Loading ...
