December 30, 2011
The BP Wayback Machine
Pitching to the Score
While looking toward the future with our comprehensive slate of current content, we'd also like to recognize our rich past by drawing upon our extensive (and mostly free) online archive of work dating back to 1997. In an effort to highlight the best of what's gone before, we'll be bringing you a weekly blast from BP's past, introducing or re-introducing you to some of the most informative and entertaining authors who have passed through our virtual halls. If you have fond recollections of a BP piece that you'd like to nominate for re-exposure to a wider audience, send us your suggestion.
The late Greg Spira tackled the notion that certain hurlers "pitch to the score" in the following piece, which was originally published in Baseball Prospectus 1997.
One of the game's great chestnuts is the theory that some pitchers "pitch to the score" and "know how to win" in a way other than always allowing the fewest number of runs possible. The reverse, that some pitchers tend to give up runs more often in close games, is also a popular theory, though it's invoked less often. This, some say, makes it worthwhile to look at a pitchers' won-loss record in addition to his ERA to judge his effectiveness, because the number of runs a pitcher allows would then not be completely indicative of his value to his team, if he gives up those runs in a way that contributes to more wins than the average pitcher with the same rate of runs allowed.
Over the years, I've seen virtually no evidence that would lead me to believe that this theory holds water for modern-day pitchers. However, just because it was clear that this theory was not applicable to the performances of most pitchers didn't mean that it might not be applicable to a few pitchers. I decided to look at the careers of individual pitchers and look at the relationship between their run support and runs allowed totals and their won-loss records, to see if that might tell us something about "pitching to the score."
It's pretty obvious to everyone that there's a relationship between wins on the one hand, and the difference between the number of runs a team allows and the number of runs a team scores. Baseball researcher Pete Palmer has found that, over the years, about 10 extra runs usually leads to an extra victory in the win column. However, a better tool exists with which to estimate how many wins and losses you expect from the runs a team scores and allows. This method, made popular by Bill James, involves squaring the number of runs scored by a team and dividing that result by the addition of the squaring of the team's runs scored to the squaring of the team's runs allowed. In other words...
Runs Scored Squared
Runs Scored Squared + Runs Allowed Squared
This results in a winning percentage, which you would then multiply by the number of games the team has played to get the number of wins, and then subtract the number of wins from the total games played figure to get the number of losses. Thus, you now have a good estimate of how an average team which scores and allows that number of runs would do in terms of wins and losses. In general, most teams finish fairly close to the result you would get from the formula. Teams that win an unusually high percentage of close games, however, will usually win more than the formula projects, and teams that lose an unusual percentage of close games (compared, in both cases, to how they do in non-close games) will do worse than the formula projects. The best recent example of this on a team level is the 1984 New York Mets. That team went 54-31 in games decided by a margin of 1 or 2 runs, and 36-41 in all other games. Because they did so much better in close games than in other games, thus putting their runs scored and runs allowed totals to more efficient use, they finished with a won-loss record of 90-72, while the formula predicted a record of 78-84.
Note that it has been found that a power of 1.83 works a bit better than a power of 2 for this formula. Since a button with a power of 1.83 is not found on most calculators, and since that kind of exactness isn't required for these kind of estimations, I'll stick with using 2.
In the past, this formula has been used mostly with teams. However, we can also estimate the expected winning percentage of a pitcher by using the number of runs he allows and his run support with the formula. If a pitcher really "pitches to the score" - thus allowing fewer runs in close games than in games in which he is granted a large lead - his won-loss record should be better than the record projected for the average pitcher with that pitcher's runs allowed and run support totals. That pitcher's career should show a pattern of him winning more games than we would expect. And, of course, the opposite should be true for a pitcher who allows more runs in close games than in blowouts.
Over the past few years, I've tested many pitchers to who have a "pitch to the score" reputation, including Jack Morris, Jack McDowell, Doc Gooden, and Dave Stewart. I have also followed the careers of two pitchers with the opposite reputation, Dave Stieb and Jose DeLeon. In no case have I found any pattern whatsoever of a pitcher winning or losing more games during his career than the formula projects. In all these cases I found pitchers who tended to win and lose the number of games that their runs allowed and run support totals project them to win and lose.
The pitchers who get a reputation of "pitching to the score" have one thing in common - they have all generally gotten good run support through most of their careers. It seems apparent to me that pitchers get this reputation because they get better run support than most other pitchers and thus have a W-L record that looks better than their ERA or runs allowed. In general, these pitchers get good run support because they are on good offensive teams for a number of years. Jack Morris is the best example of this; in his entire career, he was on only one below-average offensive team (1989), and he pitched for several great offensive teams. The "pitch to the score" theory seems to be an effort to imbue a pitchers' won-loss record with a value other than luck.
Another thing that pitchers with this reputation often have in common is that they pitch a lot of innings. This leads to more wins (as well as losses), and the number of wins often seems to impress people more than the won-loss record itself, if Cy Young voting is used as a measure. Of course, pitching more innings is generally a good thing, and makes a pitcher (especially an effective one) more valuable, but it has little to do with the issue at hand.
Note that in these examinations of careers, we estimated a player's projected won-loss record by multiplying the formula's estimation of expected winning percentage by the number of decisions a pitcher actually had. This solution isn't perfect - obviously a good or bad bullpen can greatly affect the number of decisions a pitcher has - but it is the best one available, and I don't think it represents a serious problem. Also note that no positive correlation has been found between a pitchers' run support compared to team runs scored from year to year in the AL, so there is no evidence that pitchers have an ability to inspire teams to score more runs in their starts. Of course, in the NL, pitchers' batting does affect run support, but that has no real bearing on the issue at hand.
Here is the data on the recent pitchers I've studied. First, on Jack Morris, thru 1993:
The first column is Morris' projected record based on the runs he allowed and the run support he recieved. The second column is Morris' assigned won-loss record. The third column indicates how many more wins his actual record credits him with as compared with the projection based on his run support. The fourth column is Morris' projected record based on the runs he allowed and the league average run support. The fifth column indicates how many more wins his actual record credits him with compared to the projection based on league average run support.
The conclusion is fairly obvious. Jack's records are clearly the result of how many runs are scored when he pitches and how many runs he allows. Thus, his ERA (or RA), along with his innings pitched, are a perfectly accurate measure of how valuable Jack has been to his teams.
Now, a look at the projected records (based on his actual run support and runs allowed) and actual records of another pitcher who reputedly "pitches to the score."
Jack McDowell (thru 1994)
The year McDowell won the Cy Young, 1993, he had a very high ERA early in the year but a very good won-loss record, and a number of commentators claimed that he had accomplished this by pitching to the score. Midway thru the season, I took a look at McDowell's performances and recorded when he allowed runs and when the White Sox scored for him. Here are the results:
In these 9 starts, McDowell allowed the go-ahead run to score *10* times. Six times McDowell had at least a 3-run lead, and 1 of 2 things always happened - either he blew the lead, or he shut the other team down by allowing only one run or none. Not once did he get a large lead and then allow the other team some runs but not enough to catch up - his ERA with large leads that he didn't blow is miniscule; 30 of the 33 runs he allowed enabled the opposition to either catch up or take a bigger lead over the White Sox. Not only is that not "pitching to the score," but it's the exact *opposite*. Only if you believe that McDowell is psychic and knew that it didn't matter if he put his team three or five runs down, or blew a lead, that his teammates would come back, can you argue that McDowell was "pitching to the score" - the score that would be, that is.
Two other pitchers who frequently were labeled with the "pitches to the score label" are Dave Stewart and Doc Gooden, both of whom often had won-loss records that didn't seem to go with their ERAs.
What both Stewart and Gooden had in common was not that they "pitched to the score" but that they pitched in pitchers' parks and received excellent run support year after year.
A pitcher with the opposite reputation is Dave Stieb. I looked at the bulk of his career in the same way:
Stieb didn't "win" quite as many games as the runs figures indicate he should have, but the difference of 6 victories is clearly within the range of normal random variation, and there is no visible pattern of Stieb consistently "losing" more than he should have.
We also should look at Jose DeLeon, the pitcher probably most identified with a "pitching well enough to lose" reputation recently. Note that DeLeon's 1985 and several other seasons are not included in this study because of either too few starts or too many relief appearances. (Years missing from the analyses of other pitchers may be absent because of the same reasons or because the analysis was in some cases done several years ago, thus not including those pitchers' most recent season or two.)
Once again, we see no real pattern. DeLeon's record is slightly worse than it is projected to be, but the amount is nowhere close to being statistically significant. In fact, the whole difference between his projected and actual records can be said to be the result of just one year, 1991, DeLeon's last year as a full-time starter. Out of curiosity, I went back and looked and DeLeon's box scores for that year, and found the reason for the large discrepancy: evidently, by 1991, DeLeon had lost a lot of his stamina. However, his manager, Joe Torre, took about three months to realize this. In that time, Torre consistently left DeLeon in the game just long enough to give up the go-ahead run in the 5th, 6th or 7th innings. Eventually, Torre stopped leaving DeLeon in that long, and by the end of the season, he had moved DeLeon to the bullpen.
None of this should imply that DeLeon was a good pitcher. He was a lot better than his "record" indicates, but that doesn't mean that he was particularly good. In DeLeon's case, his reputation of "knowing how to lose" got intertwined with an additional reputation of "having great stuff but not using it well." That latter reputation may very well be true, but that is a separate issue.
In addition to studying pitchers from 1980 to the present, I've been interested in looking at pitchers of the past to see if maybe a "pitch to the score" ability did exist in the past. I believe that it did exist in the far past, when baseball was a less competitive game, but I'm more interested in the recent past. Sandy Koufax is one pitcher who it has been speculated "pitched to the score," and thus was even better at contributing to winning than his great ERAs indicate. Thanks to David Smith and Retrosheet, I now have run support data as well as an ability to separate Koufax's starting performances from his relief performances, and I was able to examine the prime of his career to see if he "pitched to the score." The following is a chart of his run support, his average runs allowed per 9 innings in his starts, the w-l records that the formula projects for his starts, and his actual w-l record in his starts. The run support figures used are his run support per 9 innings when Koufax was in the game. Run support per start is also available (and in some cases, though not for Koufax, is the only information that's been published for pitchers), and it does have the advantage of including runs scored after Koufax was relieved, which certainly do affect his records, but overall I think the run support for the innings when Koufax was still in games is slightly more indicative of the quality of his support. Note that Koufax had one win in relief in 1961, so his record here differs from what's in baseball encyclopedias.
Obviously, it doesn't look like Koufax "pitched to the score" either, so his ERAs are indicative of his value.
Catfish Hunter is another pitcher of recent times I wanted to look at. Hunter's ERA relative to the league during his career (only 4% better, after park adjustment) makes him look like a very poor choice for the Hall of Fame despite the fact that he was easily voted in. So the question of whether Hunter won more games than his ERA and run support suggests he should have is an important one. Thanks again to David Smith and Retrosheet, I was able to obtain Hunter's run support figures and other data necessary to look at Hunter's career through the pythagorean prism.
Once again, we find no evidence that W-L record shows anything that adjusted ERA doesn't. Hunter in fact finishes 6 games under his projected record, but that is unlikely to have any significance other than randomness.
None of this proves that there are no pitchers who "pitch to the score," or win or lose more than their ERAs suggest they would. This is obviously not a systematic study of all pitchers, present or past. But I think the evidence presented does suggest that anytime a pitcher is tagged with a reputation that indicates that his (adjusted) ERA is not indicative of his pitcher contributions, we should take it with an extremely large grain of salt, and not believe it unless there is clear evidence that it is true.