February 27, 2006
BASEBALL: Changing Pitcher Workloads 1920-2004
It's time to unveil one of my longer-running research projects. Offense has exploded over the past decade and a half in Major League Baseball, while at the same time the top pitchers of the age - Clemens, Maddux, Pedro, Randy Johnson - have stood further and further above the league. I've long suspected that these two phenomena were connected by a common fact: the percentage of each team's innings thrown by its top pitchers has been in steep decline for years, and as a result each team - even if it has just as many or more quality arms as in years past - is delegating an ever-growing percentage of its innings to be thrown by second-line pitchers. I should note that this phenomenon is mostly independent of the dilution of pitching inherent in expansion.* In other words, the problem isn't just that each team has fewer top pitchers because they are distributed more widely about the league - it's that each team makes less use of the best pitchers it does have, and reaches deeper into its own staff, than in years past. Three trends have driven this revolution: the five-man rotation, which takes innings from the top four starters and gives it to #5; the decline in complete games and related decline in innings per start, which shifts innings from the rotation to the bullpen; and the increasing specialization of relievers, which takes innings away from the bullpen aces (now, just "closers") and gives them to middle relievers.
We know all this, of course. But I wanted to quantify it, and if someone else has, I missed it.** So here's what I did. I took one season every five years from 1920 - the dawn of the modern, lively-ball era - through 2004 (I would have used 2005 but I started the study last July). I went through each team in each league and identified their top six pitchers. For most of the study, that meant top 4 starters and top 2 relievers. For some of the 1920-35 period, I used the fifth starter in the bullpen column because teams generally had a swing man with irregular starting duties do most of the relief work.
Picking a top 6 is more art than science, though I mostly followed the listings in baseball-reference.com's team pages. Wherever possible, I erred on the side of listing the better pitcher if there were two otherwise comparable workloads and usage patterns. I'll discuss a few other specific methodological issues below the fold.*** In general, I sought to look at a team's roster and figure out, from how it used its pitchers, who the team thought were its top 6. For example, I made sure to include Norm Charlton on the 1990 Reds even if it meant classifying him as the team's fourth starter. While I used an every-five year interval, I was off a year in 1996 to avoid using a strike-shortened season, and in 1946 to avoid a year of war depletion. Without further ado, here are the results, with more notes to follow the charts.
By Rotation/Bullpen Slot
First, a breakdown of the average workload and ERA for each slot:
Innings are rounded to the nearest tenth. The every-five-years pattern does miss a few things; since 1970 was a hitters' year we basically skip the whole 1966-74 period. And you can see that workloads were mostly off in 1946, as managers spread their work around with pitchers returning from the war; those numbers would be more dramatic except for Lou Boudreau's decision to throw Bob Feller and his 2.18 ERA for 371.1 innings in pursuit of a sixth place finish. Also, bear in mind the switch to the 162-game schedule in 1961, which added 72 innings a year to each staff's workload, as well as the (not counted here) dramatic expansions in the size of the postseason from 4-7 games through 1968, to 7-12 games in 1969, to 8-14 games in 1985, to 11-19 games in 1995. A World Championship team in 1955 could expect to throw 1449 innings if it extended the postseason to its logical limit; by 1996, that number was 1629.
For recent history, you can see vividly that front-line pitcher workloads have fallen off sharply in the past 20 years, much more sharply than in any prior period, and that it's affected both starters and relievers. Without compelling evidence that this has reduced pitcher injuries, I can't see how you justify this, although you could argue that the modern postseason makes it a necessity. Until about 1930, a team's #1 starter would pitch more innings than the whole second-line staff; now, the second-line pitchers throw nearly three times as many innings as the ace.
One of the more dramatic changes comes when you trace the use of the bullpen before and after 1950. From 1955 on, you can see clearly defined #1 and 2 relievers on most teams; from 1940 to 1950, there are guys who have those jobs on most teams, but their ERAs are usually around the league ERA rather than far superior; prior to 1940, I was largely counting in those slots guys who were doing most of their work as spot starters.
Rotation/Bullpen Shares of the Workload
Next, the breakdown by starters, relievers, their performance and share of the workload. "Adv" is the ERA of the rest divided by the ERA of the top 6 (i.e., the percentage advantage of the top 6 in quality vs. the rest).
There's a lot of different competitive factors at work here over time - expansion, war, integration. But if diminishing the workloads of frontline pitchers was really worth it, you would expect the "Adv" column to flatten out sharply as the work gets spread around more. It has flattened, buit only slightly - there's still no difference between 1925 and 2000.
The League Level
The third table looks at the league totals, and also at how the league ERA would change if you adjusted the best/rest mix to a steady 70% over time. Innings totals are rounded off to the nearest whole:
|Year||Teams||Top 6 IP||Rest IP||LgERA||Adj70|
There is probably no more dramatic set of numbers on this chart than the column here showing the explosive growth in the raw number of innings thrown eah year by pitchers who do not qualify as one of their team's top six pitchers. The total is up 50% just since 1985 (while the total thrown by front-line pitchers hasn't changed much, give or take a year-to-year variation, since 1970). It's up 80% since 1975, and has tripled since 1960, and the 1960 figure doubled the 1920 total. Yes, a lot of that is a function of the growth in the number of teams, and so spread out over more games. But there's just no way around the fact that an awful lot of the time of paying customers is spent watching lesser pitchers ply their craft.
* - I do, however, think that expansion favors offenses for two reasons. First, guys who just miss being major league pitchers tend to be noticeably worse, relative to the average pitcher, than guys who just miss being major league hitters; the dropoff is more serious. Second, many guys who rush in to soak up the extra position-player jobs in the majors have been trapped in the minors not because they can't hit but because they can't field, so their arrival helps scoring in both halves of the inning. Whereas a guy who can't make the majors as a pitcher can't make it because he can't pitch.
** - Baseball Prospectus has almost certainly done some research in this area for BP's pitcher translations (more on my own methods in that regard here), but if they published the backup, I missed it.
*** - When necessary, I looked at Games Finished and number of decisions, as well as number of saves, as a measure of which relievers were the main guys. One thing that slightly depresses innings is that, if a pitcher was traded in mid-season, I counted only his innings with that team. (In one case, Bert Blyleven in 1985, a pitcher threw enough innings to qualify with two different teams). Among other things, not only would the alternative approach have required me to do a lot of extra digging to check for trades, but it would have diluted the examination of how a team's innings were divided.
I think the interesting question is whether the trend of the best pitchers working less and less can be reversed. Closer usage is not likely to change, since the best ones are so effective, and become even more valuable in the post-season. But you would think that there would be some manager willing to do a lesser version of what Billy Martin did with the A's, and ask his starters to consistently pitch eight innings rather than hand the ball to his eighth best pitcher on a regular basis.
Actually, what happened in 1946 was Bill Veeck sending Feller out there for all those innings in pursuit of the season strikeout record.
There are two factors at work with team sports and owners: Free agency/agents and rapid depreciation. You see the depreciation effect with the NFL and the Knicks. The incentive therefore to wring as much as possible out of players in a 5 year span works well in the NFL, where the player is generally shot by then, the teams get to fully depreciate them, and the broken players gets screwed.
Anyway, back to pitching innings. It makes sense to protect you main asset, the pitcher. Not for the owner, who would rather get 5 great years then goodby as a rule, but for the agent. Considering the MLBPA is the most powerful of the sports unions, and they have no made their devil's pact with the large city teams, they come out ahead, the small fries get shafted, and that's why pitchers now pitch less. They get paid more, they don't have to pitch more, they think they last longer (they really don't in most cases--great atheletes last long, mediocre ones don't anyway), and a new class of pitcher, the closer, was invented in the 80's, which compbined with free agency to make a pitcher go out for 8-9 innings a rarity.
The landscape is certainly set for some clever team or teams to exploit a new competitive advantage by front-loading a greater proportion of innings to their best starters and best relievers.
It would be nice to revise the third table to show the ratio of Top 6 IP/ Rest IP.
I suspect that one reason the top pitchers stand above the league is because they're getting more rest. Can you imagine Pedro pitching in Feller's spot? He'd have been dead by July.
I'd like to see the splits for pitchers by month. Did the pitchers of the 1920s - 1970s wear down more over the season compared to today?
So my counter theory works like this:
1. Offense has exploded in the last 20 years because hitting benefits more from bigger, quicker and stronger players than pitching. There's a larger pool of athletes at hand and they train much better.
2. Top pitchers have not slipped as much as one would expect because they get more rest than their historic counterparts.
* * *
I think fewer innings pitched should lower the number of stress injuries even if its very hard to quantify. You'd have to compare pitchers with similar pitch speed and similar pitch varieties and you'd need enough in each category to get some statistical significance.
I read somewhere that batters today take more pitches, so that the reduction in innings pitched by starters isn't necessarily mirrored by a reduction in number of pitches thrown by starters. If so. it may not be possible for starters to increase their number of innings.
If you calculate ERA+ for the three groups -- SP, RP, and rest -- you get interesting results. For SP, it's declined slightly over the last century, from 108-109 to 105-106. But it should have gone the other way, since the top 4 starters contribute a smaller # of IP and thus have less impact on the lg mean. (So the Crank's perception that today's top starters 'tower' over the field more is actually the reverse of what's happened).
The other 2 groups have improved. The 'Rest' go from about .81 early in the century to about .89 today. And the top two relievers show huge improvement, from below 100 to about 138 today.
So, the strategy of shifting innings away from top starters (along with other changes) has resulted in substantial improvement by other types of pitchers relative to the top starters. And the net result is a virtual wash -- .1 R/G at most, probably nothing at all. Very interesting.
I lay the blame for a lot of this on managers like Tony LaRussa, and those who manage strictly by "the book". What I mean is now you have what can be described as an overspecialization among relievers, managers who bring a guy in to pitch to only one hitter. These managers are so busy focusing on lefty/righty match-ups that the tend to ignore how the individuals involved handle those situations.
In briefer words its my opinion that most pitchers today are overmanaged. I do understanf though that some managers err in the other direction by over-pitching their guys, like Don Baylor ruining the Cubs young staff.
The article begins by highlighting two factors: A) increased run scoring and B) top pitchers stand higher above their peers, and then suggest they are "connected by a common fact" C) the reduced % of innings thrown by top pitchers. But as best I can tell, the evidence you present tells us that C is, at most, a tiny contributor to A (10%), and B mainly isn't true in the first place.
Seems like this would have been worth a mention.
just a quick comment on the changes in a game i've been watching almost 70 years now; i saw a column in which it was asserted that in today's take-and-rake game, pitchers of years gone by would be taxed even more; i consider that patently misthought; i mean, if you didn't jump on the first decent offering you could get wood on from some of the more canny ones such as spahn, ford, etc., you'd be bait, pure and simple; and my other take is that in today's game with its disdain for the strikeout among batters (fans are supposed to like the longball above all else, another claim without foundation), pitchers would relish the free-swinging; i've had this argument before: a guy saying pedro has had it rougher than koufax because of todays' stronger guys and smaller parks; no, i say, because it's easier to miss bats now, moreso than ever...
I don't really buy the argument that high inning counts means shorter pitching careers. Everyone uses the Koufax analogy, but he was unique in various aspects:
1. He was deprived of minor league training, so he became essentially self taught (how does a self taught cardiac surgeon sound?)
2. His delivery, probably because of number 1, put undo stress on his joints--read his bio Lefty, where other pitchers can't duplicate what he did.
3. He developed arthiritis in his pitching elbow, not a good place for a pitcher to have it. He also could have lasted a couple of more years, but probably saw how Carl Hubbell had a twisted arm, and figured the rest of his life was worth m ore than some wins.
Seaver, Clemens, Carlton, Blyleven, Wood. All different types of pitchers, all lasted a long time.
I think a factor in shorter pitching innings is the length of each inning. Pitch fast, keep delivering, and the game lasted maybe 1 3/4 hours. I think there is a limit on how long you can last in pure time before you hit a wall.
Just a couple of comments.
Correlation is not causation. There's a reason that these pitchers don't pitch 8 innings every 4 days. It's a lot harder to do now than it used to be. Sliders and split fingers - which are a main part of probably 90% of your 'Top 6 pitchers todays' repetoire take more out of the arm than fastballs, curveballs and changeups. Hitters have progressed in skill due to various things (videotape, weightlifting, year round batting cages etc) and the pitchers add these two taxing pitches to comepensate for that. If Feller tried to throw 371 innings mixing in 20 splitters a game, his elbow would have exploded in conjungtion with the July 4th fireworks.
The other blindspot I see with this study is that you really need to do it by pitch counts and by 'intense situation' pitch counts.
But it's definetly an interesting study and I don't think it's reasonable to argue that it's better to go with the Kerry Wood pitching philosophy over the Livan Hernandez one.
This is an interesting study. Guy, you make some good points about who looks better now vs in the 1920s. I think the changing number of innings used is masking a couple of things though. Try looking at ERA+ using the scenario when all years have the same usage (top 6 get 70%). When I do that the things that pop out are:
1. RP1 and RP2 improve from league average or a little below to 130 or so.
2. All the starter categories get worse. SP1 goes from 120 to 110, SP2 from around 110 to 100 and SP3 and SP4 go from 100 to 92.
3. The others stay about the same 80-85%
As Bill James and others have pointed out, it may not make sense for pitchers to both make fewer starts and throw fewer innings per start. It's very difficult to get a handle on why starters' loads have tended to drop whether scoring was trending up or down. Adding sliders and splitters may've been tougher on pitchers' arms, but baseball's decision to grandfather out the spit ball in the 20's removed a pitch that caused a lot of arm strain from pitchers' arsenals.
Obviously, pitches per start are down for everyone - even guys like Greg Maddux and Tom Glavine who throw arm-friendly pitches. Randy Johnson and Roger Clemens, who are big slider/splitter guys, have always thrown more pitches than most people. They have only seen a dropoff as a result of being really, really old.
The last total is interesting, but the problem is you can't project the quality you get from 5 or so innings of a modern pitcher to any more innings. A lot of them are just flat out of gas by that time and cannot keep up pitching past that point - so the team ERAs might be significantly worse if these guys stayed in longer.
There's no physical reason many of these pitchers could not learn to pitch longer, after all men in far worse physical condition used to pitch longer. It would take better conditioning and discipline between starts like the Atlanta pitching program has, but it could be done.