What Pitch Counts Hath Wrought:  Part Deux

Last week’s article generated, by far, the most e-mail response of any I’ve yet had the pleasure of presenting at The Hardball Times. First of all — before anything else — may I express my very sincere thanks to every single one of the many of you who took the time to send me an e-mail last week. Many of your responses were complimentary, but quite a few were critical, and please understand that I welcome and value all of them equally: your feedback is stimulating and extremely helpful, even when — perhaps, especially when — it’s challenging. Thank you so very much.

Because of the high volume of discussion, and specifically in response to several of the key points raised in many of your e-mails, I decided to postpone my planned article on bullpen usage patterns until next week, and instead devote this week’s space to a follow-up on the issues raised last week. So let’s explore in more detail the topic of the workload handled by top starting pitchers in the current era, and over the past half-century.

Last Week’s Issues

The criticism several of you offered last week centered on three primary concerns:

1) The methodology of focusing only on each season’s single top workhorse didn’t provide the means to generalize about pitcher workloads more generally. We’ll call this “The Single Top Workhorse Question.”

2) The articulation of the workload issue focused too much on the question of pitch counts, and severely underemphasized the question of days of rest. We’ll call this “The 5-Man Rotation Question.”

3) The lack of reduction in pitcher injury rates in the current era was presented as a given, without supporting evidence. We’ll call this “The Actual Injury Rate Question.”

Each is a very valid observation, so let’s address each in turn.

The Single Top Workhorse Question

Several folks commented that although the data last week clearly showed that the single most-used pitcher in previous seasons threw many more pitches than the single most-used pitcher in current seasons, it’s not valid to assume that this relationship holds among pitchers beyond that one top ace. The single pitcher throwing the most pitches in a season, several of you suggested, could very well be an outlier, an exceptional individual from whom little should be generalized.

In answering this question, first let’s review the historical pattern of single top workhorses we saw last week. Here’s the highest estimated pitch total (EPT) for each season 1946-2002 which we presented last time, and in addition let’s express each as a percentage of the average highest total (PAHT) for the entire period:

 AVG    4696    100.0%
 
Year     EPT      PAHT 
2002    4116     87.6%
2001    4018     85.6%
2000    4067     86.6%
1999    4304     91.6%
1998    4224     89.9%
1997    4099     87.3%
1996    4198     89.4%
1993    4145     88.3%
1992    4172     88.8%
1991    4128     87.9%
1990    4095     87.2%
1989    4074     86.7%
1988    4414     94.0%
1987    4627     98.5%
1986    4290     91.3%
1985    4260     90.7%
1984    4254     90.6%
1983    4597     97.9%
1982    4664     99.3%
1980    4736    100.8%
1979    5346    113.8%
1978    5216    111.1%
1977    5605    119.3%
1976    4949    105.4%
1975    4837    103.0%
1974    5684    121.0%
1973    5614    119.5%
1972    5614    119.5%
1971    5799    123.5%
1970    5000    106.5%
1969    5028    107.1%
1968    4888    104.1%
1967    4569     97.3%
1966    4892    104.2%
1965    4999    106.4%
1964    4709    100.3%
1963    4759    101.3%
1962    4786    101.9%
1961    4420     94.1%
1960    4347     92.6%
1959    4421     94.1%
1958    4362     92.9%
1957    4343     92.5%
1956    4851    103.3%
1955    4515     96.1%
1954    4872    103.7%
1953    5134    109.3%
1952    4700    100.1%
1951    4838    103.0%
1950    4912    104.6%
1949    4695    100.0%
1948    4757    101.3%
1947    4712    100.3%
1946    5954    126.8%

Here’s how I summarized this data last week:

– The workloads handled by top pitchers in the 1970s (well over 5,000 pitches) were not typical of the second half of the 20th century.

– The workloads handed by top pitchers since 1989 (practically never exceeding 4,200 pitches) is also not typical.

– The norm for the entire 1950-2000 era is somewhere around 4,300-4,700 pitches, or about 10% above the limit that modern aces are held to.

(Hey, I guessed pretty well; the actual average for the era turns out to be 4,696 pitches.)

All this makes sense so far, yes? Okay, then let’s delve a little deeper into the estimated pitch count database to test whether this pattern holds for top starting pitchers beyond the single top workhorse each year.

I took a look at it in two passes. First, I pulled the estimated pitch count data for the top number of pitchers equal to the number of major league teams each season: for example, for 1998-2002, when there were 30 MLB teams, I calculated the average number of estimated pitches thrown by the 30 pitchers throwing the most pitches; for 1993-97, with 28 MLB teams, the average pitches of the top 28 heaviest-worked pitchers, and so on. Let’s call this the First Tier of pitchers.

Then I went one cut deeper: for the Second Tier of pitchers, I calculated the average estimated pitch count for the next heaviest-worked pitchers equal to the number of MLB teams. For example, for 1998-2002, the 30 heaviest-worked pitchers beyond the top 30; for 1993-97, the 28 heaviest-worked pitchers beyond the top 28, and so on. Here’s that data, alongside the Single Top Workhorse data we just reviewed:

        Single Highest      First Tier       Second Tier
                        
 AVG    4696    100.0%    4009    100.0%    3411    100.0%
 
Year     EPT      PAHT     EPT      PAHT     EPT      PAHT 
2002    4116     87.6%    3531     88.1%    3125     91.6%
2001    4018     85.6%    3581     89.4%    3179     93.2%
2000    4067     86.6%    3603     89.9%    3186     93.4%
1999    4304     91.6%    3594     89.7%    3248     95.2%
1998    4224     89.9%    3678     91.8%    3290     96.5%
1997    4099     87.3%    3650     91.1%    3184     93.3%
1996    4198     89.4%    3760     93.8%    3324     97.4%
1993    4145     88.3%    3753     93.6%    3281     96.2%
1992    4172     88.8%    3745     93.4%    3316     97.2%
1991    4128     87.9%    3687     92.0%    3276     96.0%
1990    4095     87.2%    3636     90.7%    3199     93.8%
1989    4074     86.7%    3694     92.2%    3282     96.2%
1988    4414     94.0%    3859     96.3%    3320     97.3%
1987    4627     98.5%    4031    100.6%    3287     96.4%
1986    4290     91.3%    3923     97.9%    3390     99.4%
1985    4260     90.7%    3944     98.4%    3457    101.3%
1984    4254     90.6%    3816     95.2%    3425    100.4%
1983    4597     97.9%    3924     97.9%    3331     97.7%
1982    4664     99.3%    3938     98.2%    3413    100.1%
1980    4736    100.8%    4001     99.8%    3396     99.6%
1979    5346    113.8%    4014    100.1%    3418    100.2%
1978    5216    111.1%    4142    103.3%    3466    101.6%
1977    5605    119.3%    4127    103.0%    3440    100.8%
1976    4949    105.4%    4191    104.6%    3465    101.6%
1975    4837    103.0%    4269    106.5%    3704    108.6%
1974    5684    121.0%    4565    113.9%    3883    113.8%
1973    5614    119.5%    4614    115.1%    3808    111.6%
1972    5614    119.5%    4306    107.4%    3563    104.5%
1971    5799    123.5%    4438    110.7%    3816    111.9%
1970    5000    106.5%    4382    109.3%    3618    106.1%
1969    5028    107.1%    4421    110.3%    3715    108.9%
1968    4888    104.1%    4156    103.7%    3611    105.9%
1967    4569     97.3%    4090    102.0%    3403     99.8%
1966    4892    104.2%    4117    102.7%    3378     99.0%
1965    4999    106.4%    4285    106.9%    3420    100.3%
1964    4709    100.3%    4070    101.5%    3431    100.6%
1963    4759    101.3%    4095    102.2%    3533    103.6%
1962    4786    101.9%    4108    102.5%    3500    102.6%
1961    4420     94.1%    4033    100.6%    3494    102.4%
1960    4347     92.6%    4036    100.7%    3412    100.0%
1959    4421     94.1%    3988     99.5%    3437    100.8%
1958    4362     92.9%    3924     97.9%    3324     97.4%
1957    4343     92.5%    3885     96.9%    3358     98.4%
1956    4851    103.3%    4195    104.7%    3440    100.8%
1955    4515     96.1%    3856     96.2%    3246     95.2%
1954    4872    103.7%    4054    101.1%    3374     98.9%
1953    5134    109.3%    3971     99.1%    3132     91.8%
1952    4700    100.1%    4056    101.2%    3408     99.9%
1951    4838    103.0%    4158    103.7%    3398     99.6%
1950    4912    104.6%    4267    106.4%    3666    107.5%
1949    4695    100.0%    4190    104.5%    3537    103.7%
1948    4757    101.3%    4023    100.4%    3404     99.8%
1947    4712    100.3%    4041    100.8%    3301     96.8%
1946    5954    126.8%    4045    100.9%    3185     93.4%

Do you see a significant difference between these three patterns? I sure don’t. I don’t see any reason not to describe the data demonstrated by both the First Tier and the Second Tier of heaviest-worked pitchers just about exactly as we described that of the Single Top Workhorse:

– The workloads handled by top pitchers in the 1970s were not typical of the second half of the 20th century; they were significantly higher.

– The workloads handed by top pitchers since the late 1980s/early 1990s is also not typical; they’ve been significantly lower.

In short, this data provides no reason to dispute the basic assertion made last week: the top starting pitchers from mid-century until the late 1980s/early 1990s routinely threw significantly more pitches per season than they have since.

The 5-Man Rotation Question

Several of you pointed out that last week’s article presented the workload issue focused almost entirely as a function of pitch counts, and only mentioned in passing the question of days of rest. Regarding this criticism, please allow me to be very clear: I stand guilty as charged. I wasn’t nearly as careful and precise in my choice of language as I could have and should have been.

Let’s see if I can do better this time. Let’s make explicit a significant point that was merely presumed last week:

– The nearly universal imposition of the strict 5-man starting rotation in MLB, beginning in the late 1970s/early 1980s and becoming rigorously orthodox in the 1990s, not outing-by-outing pitch count limits per se, has been the immediate factor in limiting the seasonal workloads of top starters.

Only one pitcher since 1989 has started as many as 37 games in a season (Greg Maddux in 1991). Only two pitchers since 1979 have started as many as 40 (Charlie Hough in 1987, and Jim Clancy in 1982). Contrast this with the fact that there was never a full season from 1946 through 1989 in which at least one pitcher didn’t make at least 37 starts, and in the majority of those years one or more aces started 40 or more.

Obviously the practice through which seasonal pitch counts have been limited is the requirement of at least four days’ rest between nearly every start, not rigorous adherence to the tenets of Baseball Prospectus’ Pitcher Abuse Points. Indeed Rany Jazayerli and Keith Woolner themselves have advocated the resumption of a three days’ rest pattern for top aces, so long as individual game pitch count limits are followed, and it was sloppy of me not to make that clear. PAP per se doesn’t impose the 5-man rotation, and I shouldn’t have left open the implication that it does.

But here’s why I was so quick to toss pitch counts and the 5-man rotation together into the same big bucket: one way or another both start from a similar premise, and yield a similar result. When the minimum of four days’ rest between starts became the norm, it was as a means of reducing total seasonal pitch counts (in the expectation that effectiveness would be maximized and/or injury rates reduced, which we’ll get to). While the 5-man rotation didn’t and doesn’t require an outing-by-outing pitch count limit (and many four-days’-rest starters have been “abused” according to PAP), it is nonetheless a mechanism whose primary effect is to limit seasonal workloads.

My point is that while there may well be different pros & cons between (a) a strict 5-man rotation without single-outing pitch count limits, and (b) a usage pattern that would allow more frequent starts but strictly limit single-outing pitch counts, the practical effect of either (a) or (b) is to disallow the level of seasonal pitch counts that were routinely achieved by top starters until the mid-to-late 1980s.

Moreover, while PAP per se certainly hasn’t been driving MLB decision-making, it’s abundantly clear that an ever more cautious approach to single-game pitch count limits has been building in MLB for the past 15 or 20 years. This is vividly evident in the remarkable decline in managers’ willingness to allow complete games. In the mid-1960s, about one quarter of all MLB starts were completed; after climbing slightly in the early 1970s, by 1980 the rate had been reduced to 20%. Look at what’s happened to it since:

Year    MLB Complete Game %
1980            20%
1981            18%
1982            17%
1983            18%
1984            15%
1985            15%
1986            14%
1987            13%
1988            15%
1989            11%
1990            10%
1991             9%
1992            10%
1993             8%
1994             8%
1995             7%
1996             6%
1997             6%
1998             6%
1999             5%
2000             5%
2001             4%
2002             4%
2003             4%

That pattern has nothing to do with the 5-man rotation. That pattern is a demonstration of MLB managers’ ever-increasing decision to limit starters’ single-game pitch counts, whether in accordance with the precise recommendations of PAP or not.

It’s an empirical fact that in the past 15 or so years, all MLB teams have chosen to restrict the workloads of their top starters to the lowest levels yet seen in history. This really hasn’t been about PAP, but it has been all about a set of assumptions and expectations very much in line with those of PAP: namely, the belief that limiting the workloads of top starters to unprecedented levels is a necessary and appropriate step in order to optimize their long-term value. I trust that this much is beyond dispute.

The Actual Injury Rate Question

Finally, several folks commented that in last week’s article the lack of reduction in pitcher injury rates in the current era, whether factual or not, was presented as a given, without supporting evidence. Again I say mea culpa.

I assumed that the lack of significant or noticeable reduction in injury rates was as obvious to most everyone else as it has been to me, and I shouldn’t have. Let’s explore the injury rate issue and see if I can make apparent the reasons for my sloppy assumption.

First, let’s clearly acknowledge the inherent difficulty involved in accurately comparing injury rates between the current era and those of the past. There are two significant factors that combine to make this a tricky task: (a) the dramatic increase in the willingness of teams to place injured players on the Disabled List, and (b) the dramatic improvement in the capacity of sports medicine to diagnose and treat injuries.

It’s simply a fact that teams in the modern era make far more liberal use of the DL than they did a generation or two ago; this phenomenon has been noted by many observers, including this USA Today article. There can be only two reasons in explanation for this development: either today’s player’s get hurt far more frequently than their predecessors did, or teams choose to use the DL more readily than they used to for other reasons (likely financial, based on the vast increase in the investments tied up in player contracts).

I find the latter explanation far more believable than the former, but whatever the case, there is absolutely nothing in DL usage rate data that provides any suggestion that current players (especially pitchers, the most frequent inhabitants of the modern DL) are being injured any less frequently than they used to.

Huge advances in the science and technology of sports medicine have rendered the injuries that do occur today less devastating in terms of recovery time and quality. Particularly for pitchers, many injuries that were career-destroyers in the past (e.g. rotator cuff, elbow ligament) are frequently overcome via surgery today. Thus it may well be the case that injuries are occurring no more or less frequently to pitchers today, but they’re able to recover more quickly and fully from the injuries than ever before. I strongly suspect that this is exactly what’s happening.

But do I have data to prove it? No, I’m not aware of any. But here’s the issue: neither am I aware of any data that indicates that injury occurrence to pitchers has decreased. I’ve never seen such data based on DL visits (which of course on their face suggest just the opposite), nor on career length (and of course the career pitch count data we’ve reviewed suggests just the opposite as well).

I could be missing something. There are certainly others who’ve studied the pitcher injury question much more thoroughly than I have, namely Jazayerli and Woolner, as well as Will Carroll. But if these analysts have discovered data that indicates any decrease in pitcher injury rates in the current era compared with the past, I haven’t seen it.

First let’s consider the work of Carroll, who has undertaken the extremely admirable endeavor of placing player injuries front and center in his analysis. If anyone has detected a meaningful decrease in injury rates to pitchers over the past 15 years, one would think it would be Carroll. But I haven’t found such a finding in his body of work, including his very impressive and important piece (co-written with Nate Silver) entitled “The Injury Nexus.”

But then Carroll does write this, in his “Kicking the Crutches Out” piece published in June on MLB.com:

If the ‘old school’ wants to look back at the days they consider better than the game we see on the field today, they’ll need to accept the results those days gave them. We remember the ones who survived — Tom Seaver, Whitey Ford, Jim Palmer, Robin Roberts, Lefty Gomez, Bert Blyleven, Ted Lyons. And there are a few flameouts that never matched the success of their early 20’s — Dwight Gooden, Vida Blue, Denny McLain.

But there are other guys who didn’t survive, lost to the game and conveniently forgotten. Pete Donahue, Ralph Branca, Gary Nolan, Dan Petry, John Rigney, Dave Rozema, Dean Chance, Russ Bauers, Bill Monbouquette, Mel Harder, Steve Hargan, Mike McCormick, Van Mungo among them. For the most part, we remember the survivors, and think they were representative of all players of the past.

With all due respect to the great potential value that Carroll’s line of inquiry may yet yield, this is just completely silly. In the first place, reeling off a random list of pitchers who encountered arm problems over a period of sixty years is hardly an illuminating exercise. In the second place, just what in the world is up with this particular list of pitchers, anyway?

“Didn’t survive?” Mel Harder lasted in the major leagues until he was 37, Van Mungo and Steve Hargan each until they were 34. “Lost to the game?” Gary Nolan rebounded twice from major arm trouble to become a 15+ game winner both times; Mike McCormick came back from a major sore arm to win a Cy Young Award. “Conveniently forgotten?” Granted, I hang around with a rather geeky crowd, but most fans I know are pretty familiar with the names Ralph Branca, Dean Chance, and Bill Monbouquette. David Frishberg even wrote a song about Van Lingle Mungo.

Yes, they all suffered from arm trouble, and without the benefit of modern sports medicine, it severely undermined all their careers. But it doesn’t require any great effort at all to come up with a random list of pitchers who have suffered serious arm trouble within the past five years, too: how about Matt Mantei, Ryan Dempster, Dustin Hermanson, Chan Ho Park, Darren Dreifort, Tony Armas Jr., Scott Elarton, Jesse Foppert, Joe Mays, and Kurt Ainsworth? That’s ten; I suspect it would be easy to come up with another list of ten more. The notion that the occurrence of arm trouble is some dirty secret of the past, or that modern pitchers have been liberated from this scourge, is utterly baseless.

And the excellent, impressive body of work of Jazayerli and Woolner, including the central article “Analyzing PAP (Part Two): The Long-Term Injury Risk of High Pitch Counts,” provides nothing, not a word, indicating that pitcher injury occurrence rates have declined in the era in which seasonal workloads have been substantially reduced to historical lows.

If there had been ever a moment in which Jazayerli and Woolner would have been likely to present such data, one would think it would have been in The Neyer/James Guide to Pitchers, in which they were provided space to rebut Bill James’ contention that “abuse” of specific pitchers hadn’t resulted in any noticeable increase in injury rates. What better opportunity for them to roll out the finding that overall pitcher injury occurrence rates are down in the modern reduced-workload era — yet “A Response in Defense of PAP” contains nothing of the sort.

The reason it isn’t there, I’m almost certain, is that it doesn’t exist. If it did, one can be confident that Jazayerli and Woolner, and Carroll, would have discovered it and provided it to us; there’s every reason for them to do so. If they have, or someone else has and I’ve somehow missed it, I welcome the chance to be corrected; if anyone is aware of this information, please direct me to it.

All Righty Then

Last week I asserted that on the issue of pitch counts I firmly agree with the position of Bill James, as presented in his article “Abuse and Durability.” To be very clear, allow me to present the specific conclusions that James draws with which I completely concur:

“I do not question that there is a risk of working a young pitcher too hard and destroying his arm, robbing him of his future” (p. 461)

Questioning the wisdom of how pitchers are being deployed in the modern era should absolutely not be construed as an assertion that pitchers, most particularly young pitchers, aren’t very susceptible to injury, and shouldn’t be handled with care.

“I do not question that the pitch counts are a potentially useful tool in steering clear of this disastrous result.” (p. 461)

If I failed to make this clear last week, let me make it very clear here. Pitch counts are an essential tool; to ignore them would be completely foolish.

“I do not question that Rany and Keith’s studies were done in good faith and with the best of intentions.” (p. 461)

And their work deserves the highest respect.

“I think there is a natural balancing of risks, in almost any physical activity, and that this balancing of risks, with respect to the use of pitchers, has gotten out of whack.” (p. 462)

In the ardent attempt to avoid the worst outcomes in pitching careers, the possibility of the very best outcomes has been precluded.

“Most injuries to pitchers are not the result of chronic overuse; some are, particularly to young pitchers, but most are not.” (p. 463)

This is simply what the historical data tells us.

“Backing away from a pitcher’s limits too far doesn’t make a pitcher less vulnerable, it makes him more vulnerable. And pushing the envelope, while it may lead to a catastrophic event, is more likely to enhance the pitcher’s durability than to destroy it.” (p. 463)

A positive reinforcement dynamic between durability and heavy workload is evident in many realms of athletic endeavor, and the historical data suggests it’s true in pitching as well.

References & Resources
In addition to the articles mentioned and linked above, we again made significant use of the database of historical estimated pitch counts provided by Tangotiger.