November 27, 2006
BASEBALL: Wins, Losses and ERA
One of the many fancy new features rolled out at Baseball-Reference.com lately is the ability to get career and year-by-year splits, in more detail than they have previously been available even at Retrosheet or through David Pinto, because they are derived from a newly completed database of all box scores dating back to 1957. It's a goldmine, the potential of which I have not yet fully absorbed.
One cool feature is the splits showing how pitchers over the course of their career pitched in their wins vs. their losses. I started examining some recent pitchers of interest and decided to do a comparative study.
I looked at 58 pitchers, including every 200-game winner to start his career since 1957, plus three Hall of Fame pitchers for whom all but their first two years are available (Koufax, Drysdale and Bunning), plus a bunch of other guys who cleared 150 wins and/or 2500 Ks or were otherwise notable: Dwight Gooden, David Cone, Dave Stewart, Fernando Valenzuela, John Smoltz, Andy Pettitte, Al Leiter, Kevin Appier, Bret Saberhagen, Ron Guidry, Rick Sutcliffe, Mark Langston, and Sam McDowell. The group is a cross-section of the top pitchers from the Sixties, Seventies, Eighties, Nineties and today. Some conclusions:
1. I was amazed at quite how dramatic the difference was between these guys' average ERAs in their wins and losses. Yes, you expect a big separation - but of the 58, all but 12 had ERAs below 2.00 in their victories, and all but three had ERAs above 5.00 in their losses. The lesson: yes, starting pitching matters quite a lot in determining wins and losses, in case you still had any doubt.
2. The 12 who posted ERAs above 2.00 are obviously weighted more towards the post-1994 period and almost exclusively towards good teams. In order: Kenny Rogers (2.39), Pettitte (2.33), Jamie Moyer (2.30), Jack Morris (2.29), Sutcliffe (2.26), David Wells (2.25), Mike Mussina (2.17), Langston (2.10), Appier (2.07), Charlie Hough (2.06), Dennis Martinez (2.06), and Stewart (2.05).
The guy who jumps off that list is Morris - he played in the 1980s, not the 1990s, but always had superior offensive firepower at his back. The relatively high ERAs in his wins suggests that his W-L record should be viewed with a bit of skepticism, which I have argued for years as being a reason to doubt his Hall of Fame credentials. The same doubts, to a lesser extent, may be raised about Mussina, depending how his career record ends up, and especially since Mussina lacks Morris' reputation as a big-game pitcher.
Appier was more surprising, until you remember that he got a disproportionate number of wins with the A's, Mets and Angels after his prime, as opposed to his best years in KC. Appier also posted a solid 3.50 ERA in his no-decisions, but more on that below.
(If you are wondering, Steve Trachsel had a 3.97 ERA in his wins in 2006, 2.23 career).
3. The three guys with the lowest ERAs in their losses should not be surprising, though I was mildly surprised. One was Bob Gibson (4.69), who pitched in the pitcher-heaven Sixties (in 1968 he had a 2.14 ERA in his losses and a .500 record when not throwing a shutout despite pitching for a first-place team). One from the same era who surprised me more, since he always had good Mays-and-McCovey offenses, was Juan Marichal (4.94), but then you have to figure that Marichal lost a lot of his games to Gibson and both lost a lot to Koufax and Drysdale. The third, unsurprising given his teams, was Tom Seaver (4.92).
4. The very best pitchers in their wins were the extreme power pitchers and guys who won a lot in Sixties-era pitcher's parks: Koufax (1.34), McDowell (1.43), Nolan Ryan (1.45), Drysdale (1.48), and Pedro Martinez (1.49). But the next rank, the guys in the 1.50s is a mixed bag: Jim Palmer (1.52), Gaylord Perry (1.54), Tommy John (1.57), Jerry Reuss (1.58), Gibson (1.59). Bert Blyleven was next at 1.60, if you were wondering.
5. The worst pitchers in their losses - Rogers (7.39), Wells (7.27), Pettitte (7.18), Leiter (7.04), Mussina (6.73), Sutcliffe (6.70), Stewart (6.62), Langston (6.57), Moyer (6.55), Gooden (6.54). Leiter (1.90-7.04) and Rogers (2.39-7.39) have the biggest differences between wins and losses, though Reuss (1.58-6.24) has perhaps the biggest percentage difference (I didn't run the numbers on that).
6. A fascinating field of study would be to look at these guys' no-decisions, but that proved labor-intensive and I may wait and see if further changes on the site make it easier to compute. One problem is that there is as yet no way, outside of reviewing individual games, to separate no-decision starts from no-decision relief appearances - for example, Koufax had a 6.90 ERA in 118 games where he appeared without a win, loss or save, but only 64 of those were starts, and much of the rest was probably early-career mopup work.
Anyway, while the results of this study can't be separated out from the various other influences on a pitcher's ERA and W-L record, I did find it interesting and illustrative.
In conclusion # 3 you state "Marichal lost a lot of his games to Gibson and both lost a lot to Koufax and Drysdale." Bill James did a study that showed Drysdale having a losing record face-to-face against Marichal and Gibson. Koufax and Marichal only started against each other three times. One was a Koufax no-hitter and another featured the Marichal-Roseboro bat fracas. Gibson losing to Koufax is probably the most correct part of that statement, but it would not be "lost a lot" as Gibson more often faced Drysdale than he did Koufax.
Excellent study, Crank. Those splits are amazing.
It's funny, because I've had debates with baseball fans for years, and W-L makes for strange bedfellows. Most of us who are sabermetrically inclined have friends who are not . . . but the degree of disrespect for "the stathead world" is key here: anyone who thinks W-L is more important than ERA, especially for one season, is unable to debate past point one. But many of what I call Tier Two fans have only gotten to the point where they salute ERA, and then move on.
Meaning, of course, that they're missing K/BB, K/9, IPs, BA independent of fielders . . . and W-L. Although it's more of a "feel" thing than pure statistical analysis, I look to career W-L before I look to career ERA. ERA, over 15 years, is way too influenced by ballpark, league, historical era, and fielders. But for about 3/4 of starters, the luck elements of W-L will balance out.
The Morrises, on one end, and the Bunnings, on the other, will always be there, but over a career most good pitchers will win, most bad pitchers will lose, and the vast majority of average pitchers will go about .500.
Again, great study; I await your ND data (hint, hint).
Fair enough about Marichal but if you look as a percentage of Marichal's losses - which no, I didn't - I'd suspect you'd see a fairly high percentage were to the very top pitchers of the era.
It was very well known then that Marichal would go against Drysdale a lot. Managers do tend to go with aces against aces. I wonder how much analysis they put into not just specific matchups, but how a pitcher does say in certain weather, or day vs. night, or something else. I'll bet that Stengle and Weaver, who were well known for some bizarre pitching "rotations" --especially Casey, factored more of that than other managers.
Right now the information and the answers are here. It's just what questions to ask.
Interesting study. Thanks.
This is not in opposition to your Morris *conclusion*, but I think we should consider a factor that some pitchers were known for, more in general pre-WWII, but more recently with some specific names attached.
Don Sutton & Catfish Hunter were both "known" (therefore, by logic, a whole bunch of other pitchers, not-known, too) for pitching to contact when they had a lead they considered significant. Rather than try to be too precise, they tried to "let" the batter put the ball in play. Both were known for giving up solo HRs in games w/leads considered safe. Bad for RA, good for getting CGs or at least deeper into games.
A look at the B-R splits, btw, doesn't show this as a big theme for them. There are a host of little reasons it might both be true and not show up in the splits.
But it may be why Morris' W-L ERA splits look different...that is, in certain spots when the game looked decided or when h didn't have his best stuff, he may have shifted gears to pitch to contact. Perhaps.
Someone should ask him, or Sutton
I've always thought it was a bit silly to apply that 'pitches to contact' reputation only to a small handful of pitchers, and thus elevating them above guys with better numbers, who presumably were also capable of concluding it was a good strategy. Is it really likely that Jack Morris knew that much more about winning games than Bert Blyleven, when the numbers clearly indicate that Blyleven knew more about other aspects of getting hitters out ?
I found this very interesting, thanks Crank for bringing it to light. I think the point about top pitchers such as Gibson, Koufax and Marichal facing each other is valid. Probably more than half the time an ace is facing an ace.
One note about Morris, maybe the fact that he pitched several years in the Metrodome influenced his numbers. If you do accept that...that Bert Blyleven must have been a hell of a pitcher.
About Gibson in '68, the Cardinals were not a powerhouse offensive team. It would ahve been interesting to see what his career nubmers wold have been if they ahd not lowered the mound. He was amazing.
Anyway, thanks for the info. I intend to share it often.
Interesting study. Your comment "but of the 58, all but 12 had ERAs below 2.00 in their victories, and all but three had ERAs above 5.00 in their losses. " got me to thinking about why this might be.
If a pitcher gives up alot of runs early; his ERA will soar unless the manager leaves him in. If the pitcher settles down, then his ERA can drop.
Years ago, the big time pitchers were left in even if they gave up runs early. Now the managers generally pulls the pitcher and bring in the long reliever. Hence the starter has no chance to settle down.
I wonder if this would cause higher ERA for losses?