Remember me
Forgot your username/password?
 
Print
Home>Articles

Reliever Won-Lost Records

By Bill James
May 17, 2008

 

Do Reliever Won-Lost Records Mean Anything in Modern Baseball?

 

            A few days ago I posted here a poll question:  Do the won-lost records of relievers mean anything in modern baseball?   Do relievers who pitch well tend to have ANY better won-lost records than relievers who don’t pitch well?   Or is the relationship even inverse, that relievers who pitch poorly are more likely to pick up wins?

            This is a very simple study of a very simple question. . .took me about twenty minutes to do.   I took the records of all pitchers:

            a)  since 1990,

            b)  who did not make a start,

            c)  and pitched at least 40 times in relief,

            d)  and had at least four decisions.

 

            There are 1,835 such pitchers.   These I sorted into three groups:

            A)  Winning Percentage over .600,

            B)  Winning Percentage of .400 to .600,

            C)  Winning Percentage under .400.

 

            And I figued the performance averages for each group. 

            There were 670 relievers with winning percentages over .600, 570 with winning percentages of .400 to .600, and 595 with winning percentages under .400.   These were the aggregate won-lost records of the groups:

            Group A          3690-1514       .718

            Group B          2409-2406       .500

            Group C          1269-3071       .282

 

            The answer is:  the pitchers who had better won-lost records did pitch better, and the pitchers who pitched better did have slightly better won-lost records.  

            The games and innings pitched of the three groups were near the same—62 games, 68.0 innings for Group A, 63 games, 68.2 innings for Group B, and 61 games, 64.1 innings for Group C.

            The strikeout/walk data was just a hair better for Group A—K/W of 57/26 for Group A, 57/27 for Group B, and 54/26 for Group C. 

            The pitchers in Group A got out 71% of the hitters they faced, in Group B 70%, and in Group C 69%.   Each group faced about 300 batters per season.

            The average ERA of Group A was 3.40, of Group B, 3.71, of Group C, 4.03.  I looked at the league ERAs and ERAs relative to league just to make sure there wasn’t some distortion there, and there wasn’t.

           

            I sorted the data by the other axis, so that Group A was the most effective pitchers (2.61 ERA), Group B the pitchers of middling effectiveness (3.68), and Group C the least effective (4.95).   The winning percentages, by group, went .563, .496, and .458.

           

            It is worth noting that the pitchers with the best winning percentages had an average of 22 Games Finished and 7 Saves, whereas the pitchers with .500 records had averages of 27 Games Finished and 9 Saves, while those with losing records had averages of 29 Games Finished and 12 Saves.   In other words, use as a closer does increase a reliever’s liability for losses, and makes it more likely that he will post a losing record.   But there are two things that cause losing records in relievers—being used as a closer, and ineffectiveness.

            In modern baseball we don’t evaluate relievers by their won-lost records, and obviously we shouldn’t; I’m not suggesting that we move in that direction.   But that’s not the question I was asking.   What I was asking was this:  has the nature of the modern bullpen rendered won-lost records for relievers wholly irrelevant, or are they still indicative of performance at some low level?   And the answer is:  they are still indicative of performance at a low but not terribly low level.

 

 

 
 

COMMENTS (6 Comments, most recent shown first)

PeteRidges
Here's a difference (or a paradoxical similarity?) between a relief win and a relief loss. A relief loss could always have been prevented, if the pitcher had pitched better. But then you've got a vultured win, which would also have been prevented if the pitcher had pitched better: many times, a reliever CANNOT get a win unless he finds a way to give up a run (or three).
http://www.baseball-reference.com/boxes/BOS/BOS200505110.shtml

Here's a suggestion for a study: choose a measure for individual game performance by a reliever, either the change in expected runs or the change in win probability. Then look at the range of values for the winners and losers. My guess would be that the losers would generally tend to have "bad games" but the winners would have about as many "good games" as "bad games". So you end up with the slightly odd conclusion that losses are a meaningful stat, but wins aren't...
8:27 AM May 28th
 
Trailbzr
Responding to Bill's last comment, isn't the average win about 9 innings and the average loss 8.5 because half of winning teams don't bat in the bottom of the ninth?
6:03 PM May 19th
 
bjames
Following up on what I posted an hour ago, I thought I would study the records of starting pitchers to document the effect that I expected. (The effect I expected and the reasons for expecting it can be summarized as follows: 1. When pitchers have more decisions, they are more likely to have records near .500. 2. When pitchers have more decisions per inning, they are likely to have more decisions total. 3. Therefore, when pitchers are near .500 they are likely to have slightly more decisions per inning.)

I looked to see whether this was true in fact. To do this, I took the records of all pitchers with 30 or more STARTS in a season since 1980--a total of 1,689 pitchers. These I sorted into "high winning percentage" (.600 or better), "low winning percentage" (.475 or less) and "middle group" (.476 to .599).

What happened was that the expected effect was swamped by a different effect, which I should have anticipated but didn't. The "high winning percentage" group averaged one decision for each 26.77 outs. The "middle winning percentage" group averaged one decision for each 25.87 outs, or 8.62 innings. But the "low winning percentage" group averaged one decision for each 25.54 outs, or 8.51 innings.

For an obvious reason. You can get a loss quicker than you can get a win. You have to get 15 outs, as a starter, to get a win. You can get a loss without getting anybody out. Thus, the more often you get knocked out early, the more decisions you will have per inning pitched, as a starting pitcher.
5:28 PM May 19th
 
bjames
The innings per decision are lower for mid-range pitchers than for "end group" relievers because, with more decisions, it becomes easier to fit in the mid-range group. With only four decisions there is only one data point in the .400 to .600 range (2 wins, 2 losses), and with five decisions there are actually NO data points in the .400 to .600 range (since the higher range includes .600 and the lower one includes .400.) But with seven decisions there are two data points in the .400 to .600 range (4-3 and 3-4).

When you have more decisions you are more likely to finish near .500. This same effect would cause the decisions per inning to be higher near .500 in a study of starting pitchers, but in a study of starting pitchers, because the number of decisions is much larger, the efect would be negligible. It's a more significant effect here because the decision numbers are so small that a high percentage of the pitchers with four or five decisions are in the end point groups.
4:20 PM May 19th
 
Trailbzr
I found most interesting the paragraph:
2.61 ERA = .563
3.68 ERA = .496
4.95 ERA = .458

If park effects were negligible (which they probably aren't) the difference between 2.61 and 4.95 for a starter would be about .300 points of W/L. That it's 1/3 that significant for a reliever suggests that a reliefs records are 2/3 driven by use context.

BTW, the .400-600 pitchers had 1 decision per 8.13 innings, suggesting there could be some kind of distortion created by the group definitions.
9:13 PM May 17th
 
wovenstrap
Does quality of team muddy the data in any way? The good teams might have a tendency to (a) get better relief pitchers and (b) give them an environment in which they can prosper more. So let's say group A has mostly Yankees pitchers, group B is dominated by the Phillies or somebody, and group C is chock full of Royals. Wouldn't just the "team" identities of those three groups -- in and of themselves -- lead to better pitchers being in Group A etc.?
6:58 PM May 17th
 
 
©2024 Be Jolly, Inc. All Rights Reserved.|Powered by Sports Info Solutions|Terms & Conditions|Privacy Policy