Better Than the Team

              This is not a very significant study, but. . .it’s a study, so take it for what it is worth.   Suppose that we take the method outlined on Monday (A Day at the Beach), but ask this question:   By how much (how many wins) did the team exceed expectations with each starting pitcher on the mound?

              On April 23, 2003, the Detroit Tigers played the Oakland A’s in the Oakland Coliseum.   The Tigers were 43-119 that year; the A’s were 96-66—plus the A’s were at home, the Tigers on the road.     The Tigers were 20-61 on the road; the A’s were 57-24 at home.    Batting third for the A’s that day was Miguel Tejada, a shortstop who hit 30 homers and drove in 115 runs; batting third for the Tigers was Craig Paquette, who hit .152 that year and did not drive in a run.   The Tigers young pitcher Jeremy Bonderman entered the game, the fourth start of his rookie season, with a 10.22 ERA.   The A’s starter, Barry Zito, had won the Cy Young Award the previous season, 23-5 with a 2.75 ERA, and entered this game 3-1 with a 3.00 ERA, having pitched a shutout in his previous start.  You’d have to figure that the A’s are going to win that game. 

              But they didn’t, of course; why else would I have brought it up?    Bonderman was great; he pitched eight innings, three hits, one run, five strikeouts, no walks.   If he had done that all season he would have been Steve Carlton in ’72.  Bonderman finished the season 6-19, but. . .hey, heck of a game there.   

              Or, on the other hand, Jack Sanford in 1962.   Sanford was 24-7.   Not that I diminish that, but the team behind him was pretty good.    Willie Mays drove in 141 runs, the team won 103 games, led the majors in runs scored and scored easily more runs than the ’61 Yankees.   The offense was literally so good that Willie McCovey, with a .957 OPS, couldn’t crack the starting lineup; no exaggeration.   Matty Alou hit .292 for the Giants (.310 the previous year), but couldn’t get off the bench, either, as all three regular outfielders (Mays, Felipe Alou and Harvey Kuenn) hit .300, averaging 30 doubles and 28 homers apiece. They had two left-handed hitting catchers who, between them, had 526 at bats, hit 35 homers and drove in 100 runs, and also drew 93 walks.   Their catchers could hit cleanup for most teams, but they usually hit seventh for the Giants.   Their third catcher, John Orsino, got traded to the Orioles that winter, and hit .272 with 19 homers for the Orioles the next year, .824 OPS.  

              Anyway, on May 27, 1962, the Giants played the Mets in San Francisco.   The Giants were 103-62 on the season.   The Mets were 40-120.   The Giants were at home.   Probably we should figure that the Giants will win this game, and they did; Jack Sanford pitched seven strong innings, one of the catchers hit a homer, Felipe Alou drove in three runs and the Giants won, 7-1.  

              Sanford started against the Mets again in San Francisco on July 3, and beat them again, and started against them again in San Francisco on August 9, and beat them for a third time.   In the three starts he was 3-0 with a 0.72 ERA.    My point is, though, that by traditional methods Jeremy Bonderman receives no more credit for beating the A’s with a horribly overmatched team, on the road, than Jack Sanford does for beating the hapless Mets at home with a great team behind him.    But now, given the method I introduced on Monday, we can fix that!!

              Sorry.  Sometimes I wish I could hype things.  The 1962 Giants at home against the Mets had an 83% chance of winning a game, so when they did win that game, the May 27 game, that exceeded expectations by .17 wins.    The 2003 Tigers in a road game against the A’s, had a 19.6% chance of winning a game, so when Bonderman beat the A’s, that exceeded expectations by .804 wins.   

              A simpler version of this line of analysis was actually popular in the 1970s.   The Sports Encyclopedia:  Baseball had just two or three pages of analytical stats, but one of their main features was a comparison of a pitcher’s won-lost record to the winning percentage of his teams.    What I am doing here is essentially the same as that—comparing a pitcher to his teammates—but it advances the ball three little steps.   First, my method considers all starts, rather than just credited wins and losses.   Second, this method is aware of whether the pitcher is at home or on the road, and third, this method is aware of (and takes account of) the quality of teams that the pitcher pitches against.  

              My data here starts in 1952, is incomplete for the early years and ends in 2013.   I apologize for that; I don’t have meaningful data for active pitchers, but I can’t get that data file updated.   I have partial-career data for guys like Warren Spahn and Early Wynn and Bob Lemon, and don’t have anything for Walter Johnson and Lefty Grove.   But here’s what I did with the data I have, if I haven’t already explained that.   When Jeremy Bonderman beats the A’s on the road with a badly overmatched team, I credit Bonderman with +.804 wins.   When Jack Sanford beats the Mets at home, he gets only +.170 wins.  Bonderman beating the A’s once is a bigger deal than Sanford beating the Mets three times.  

              Of course, there are inherent inequalities in the process.   Greg Maddux is implicitly being compared to John Smoltz and Tom Glavine; I am guessing you all know this.    Anyway, suppose we do this for every game of a pitcher’s career, counting positives and negatives.

1.  Hall of Famers

              I didn’t track the data season by season.   Carlton in ’72 was +13.5 wins, which I would assume is the record for a season.    Partial-career data there for Spahn, Roberts and Lemon, also missing 13 starts for Whitey Ford and 3 for Bunning, 2 for Drysdale.   Basically, if you improve your team by 20 games over the course of a career, that’s a Hall of Fame standard, and better than many of the Hall of Famers.

2.   Hall of Fame Candidates

              By that heading I mean plausible Hall of Fame candidates; I am not referring to where they are in the voting structure right now.   I just mean that these are guys that you could reasonably argue should be in the Hall of Fame, maybe.   

              Is Bob Buhl really a plausible Hall of Fame candidate?  Probably not.    Hey, he won 166 games; it is more than Koufax or Dizzy Dean.  

              I love Catfish Hunter and don’t want to kick him out of the Hall of Fame or anything, but I have to note Catfish and Luis Tiant.    Catfish made 476 starts in his career, Tiant 484.    Catfish was 224-166; Tiant was 229-172.   Both American League pitchers, same era, 1965-1979 for Catfish, 1964-1982 for Luis.   3.26 ERA for Catfish, 3.30 for Tiant; Tiant’s ERA is actually much better if you adjust for the parks.   

              But Catfish, pitching for much better teams for most of his career, improved his teams by 11 games; Tiant, by a little more than 30 games.   Just noted.   Buhl should have been in the next group, which is "Probably Not a Hall of Fame Candidate":

3.   Probably Not a Hall of Fame Candidate

              Kenny Rogers is an interesting name here—219 career wins with almost 300 career relief appearances, and, as we see here, it wasn’t his teams that were carrying him; he was carrying his teams.    Ramon Martinez, Pedro’s older brother, is another interesting name way up on the list; people forget how good he really was.    If he had stayed healthy, he’d be in the Hall of Fame, too.    Chuck Finley is comparable in this data to Rogers; he had 200 career wins as well.     You can make a Hall of Fame case for Buehrle or Saberhagen.  You CAN make a Hall of Fame case for some of these guys, not the ones below:

4.   Definitely not a Hall of Fame Candidate

              Not Hall of Fame candidates, but these were good pitchers.    The guys below were good pitchers or REALLY good pitchers, but in short careers.   We’ve talked a lot about Fidrych, but who remembers Tom Filer and Brian Tollberg?

Good Short Careers

              There are some good short careers on the other charts, of course. . . J. R. Richard, Master Johan, Jim Maloney, Sal Maglie and Mike Maroth.    Herb Score was actually under water, -2.1, in part because he pitched for teams with other Hall of Famers in the rotation.   The Cleveland Indians of the 1950s were the Atlanta Braves of the 1990s, with a Hall of Fame caliber pitcher on the mound more days than not.  

Not as Good by this Method as I would assume they would be

              Some guys, this method doesn’t seem to like.    I always think of Jose Rijo as being a Hall of Fame caliber pitcher who just couldn’t stay healthy, but this method doesn’t think so:

              I always think of Chris Short as an outstanding pitcher, but the method doesn’t like him.   Dean Chance was a Cy Young winner in 1964, a 20-game winner in 1967, a 128-game winner in his career, but the method sees him as being not much better than his teams.   Clay Kirby was an outstanding starter for the expansion Padres for a couple of years; I expected him to do well, but he doesn’t.   Joe Horlen had a 1.88 ERA in 1964, was 19-7 with a 2.06 ERA in 1967, but for his career his teams did better without him on the mound than with him.    People love Dock Ellis and love to talk about him, but he wasn’t really helping his teams a whole lot.   Some guys had long careers with relatively low numbers:

              And then, some guys just had REALLY low numbers:

              Jeremy Bonderman in his career had 200 starts and was +2.7; just tying up a loose end.  Again, don’t make more of it than it is; it’s just one more way of looking at the general question.  

              While I was doing this I got interested in the question of how to treat no-decisions, how to think about them.    Each start by a starting pitcher is one of four things:

              1)  A win for the pitcher,

              2)  A loss for the starting pitcher,

              3)  A no-decision for the starting pitcher, but a win for the team,

              4)  A no-decision for the starting pitcher, and a loss for the team. 

              In my data there are 86,638 wins for pitchers, 89,131 losses for starting pitchers, 33,999 no-decision/wins and 31,506 no-decision/losses. 

              In Wins, starting pitchers pitched an average of 7 and a half innings, and had an ERA of 2.00 (actually 1.998).   In losses, starting pitchers pitched an average of 5.4 innings and had an ERA of 6.61.  But what about the no-decisions?

              In No-Decision/Wins, starting pitchers had an ERA of 4.71; in No-Decision/Losses, an ERA of 4.16, pitching about five and a half innings a start in either group.    So that means that pitchers who had a no-decision but their team WON the game actually pitched worse than pitchers who had a no-decision but the team LOST.   I did not know that; until I did that study I didn’t know which way that would go.  That kind of makes sense when you think about it, but anyway, what interested me is that this data can be used either as an argument for or against this method. 

              Against:   This method gives a pitcher credit when he has a no-decision but his team wins, but gives him a negative number when he has a no-decision but his team loses.    But research shows that pitchers who have no-decisions in losses actually pitch BETTER than pitchers who have no-decisions in losses.   What kind of sense does it make, then, to give positive credit for appearances with a 4.71 ERA, but no credit for appearances with a 4.16 ERA?  

       For:   The data shows that the ERAs of pitchers who have no decision in a game are about half-way between the ERAs of pitchers who win, and those who lose.   Doesn’t it make sense, then, to give the pitcher half credit for those games?    This is what we do, when we give him credit but for only half of the games.  

       Yes, the ERA is a little bit better when the no-decision is a loss, but really, that’s a very small difference compared to the 4.61-run difference between wins and losses—or even compared to the difference between wins and no-decisions.  

              This second argument somewhat miss-states the evidence, when it says that we are giving "half credit" to the pitchers for those games.   If half of those are wins and half are losses, then it is a wash, and on balance there is no credit being given.    It is what it is; it is only what it is.   Thanks for reading.