The mystery of innings pitched per start before and after the introduction of the DH

Here’s something I’ve never understood:

IP_per_start

This is the average innings pitched by the starting pitcher, by year, broken down by leagues.

There are at least two things about this plot that make no sense to me. Click through for more.

Mystery #1:

Forgetting the issue of the DH, I am perplexed why there is such a strong correlation between the two leagues. As you can see, with the exception of what happened in 1974 and 1975, the two leagues are strongly linked in terms of year-to-year change. Even without a statistics degree, I know this is not random chance. I would expect, however, that for what is a largely random grouping of teams, there would be noise in the data–for example based on specific pitchers and what kind of year they had on a team-by-team basis. Instead, we are seeing MLB-wide variations that are clearly affecting both leagues similarly every year. Given the up-and-down fluctuations of these systematic changes, the only solid theory I have is that it was the weather. We know that average yearly temperatures can vary by several degrees, and I can imagine this having a systematic effect on the tiring of pitchers across the country. (I have not been able to locate historical weather data–if you can, I’d love to get access to it to see how it correlates with the plot above.)

Mystery #2:

Logic suggests that the introduction of the DH would allow starting pitchers to go deeper into games. As we still see today in the NL, managers occasionally lift their starter a bit early in favor of a pinch-hitter at a key offensive moment of the game. In theory, the introduction of the DH in the AL in 1973 would have eliminated such choice, allowing managers to leave their starter in based only on his pitching performance. However, what we see is not a rise in the AL, but rather a drop in the NL.

At first, this seems really weird. The NL had no rules change in 1973, and so why would it be dramatically affected by a rules change in the other league? And why didn’t starting pitching duration change in the AL in 1973 with the introduction of such a radical rules change?

The lack of change in the American League can be explained by a few possibilities. Notably, scoring went up dramatically in 1973 (as one would expect by replacing pitchers in the lineup with a bona fide hitter.) Therefore, this may have tended to drive down starting pitcher duration based only on pitcher performance. The notion that starters were no longer removed for pinch-hitters may not have been the dominant factor.

It’s also true, though, that scoring in 1973 went up in the NL as well. This wasn’t due to the DH but was a “regular” fluctuation, perhaps related to the weather. It may be that the graph above shows us just what our intuition expects, but that it’s overlaid on an overall downward trend. In other words, in the mid 1970s, offense increased, starting pitcher duration shortened, but not as much in the AL because of the elimination of starting pitchers being removed for pinch hitters. So instead of seeing the NL stay the same and the AL go up, we see the NL drop and the AL stay the same.

I can imagine some other less likely possibilities having to do with the manager philosophy over starting and relief pitcher usage, but I won’t spell those out in detail here.

 

0 0 votes
Article Rating
Subscribe
Notify of
guest

25 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Ed
Ed
10 years ago

Andy – I suspect that part of the problem is that your graph doesn’t have a zero point and therefore small changes from year to year appear to be larger than they really are. So for example, the NL decrease from ’73 to ’74 is only about 1.5-2% which I would hardly characterize as dramatic.

Ed
Ed
10 years ago
Reply to  Andy

Andy: Perhaps I should have been clearer. You raised several different points in your post one of which was the following:

“The NL had no rules change in 1973, and so why would it be dramatically affected by a rules change in the other league?”

Again, the way you drew the graph distorts the picture. There really was no dramatic drop in the NL until the 76-77 change. Before that, what you have going on is normal year to year change of a few percentage points.

Ed
Ed
10 years ago
Reply to  Andy

Just to be clear Andy, I wasn’t trying to be difficult. I used to work in a research organization and charts like yours used to drive my boss crazy! 🙂 So it’s been drilled into me about the missing zero issue.

mosc
mosc
10 years ago

More relief pitcher usage as time went on during that period is particularly dramatic

Mike L
Mike L
10 years ago

Neat charts. I’ve never really bought the idea that NL managers think that much differently than AL managers do, at least on the really big asset allocation issues such as pitch counts, five man rotations, and bullpen usage. There are outliers (Sparky, for example) but I would suggest that there’s a group think, or conventional wisdom, which permeates almost all organizations. So I think the leagues would be correlated. It seems to me that you have two large groups of athletes that, in the aggregate, have a fairly similar median level of talent. You can get new stars coming on… Read more »

John Autin
Editor
10 years ago

All I know … That graph makes me nostalgic for a video game I played in the ’80s. What was it called? … Tempest.

Ed
Ed
10 years ago
Reply to  John Autin

Oh I loved Tempest!!!

John Autin
Editor
10 years ago

Andy, interesting questions. But there are so many other things going on at once in that period, it’s hard to tease out the specific effect of the DH. – Large yearly fluctuations in scoring. In 2 years from 1968-70, NL scoring went up 32%. Strategies would be reacting to that for years. – The rise of the union, more job security, players asserting themselves more, less willing to pitch through an injury. – Expansions at the beginning and end of the graph period. – Rise of Astroturf meant fewer rainouts, a more consistent schedules, fewer doubleheaders. (I can’t explain the… Read more »

Evan
Evan
10 years ago

I often suspect differences in the way baseballs are manufactured from year-to-year (be they accidental or by secret directive of MLB) when I see otherwise unexplained changes in runs/game.

I was thinking the large drop in 1977 might be expansion related, but I have no idea why the effect wouldn’t be more diluted in the NL because the expansion draft was limited to AL teams.

Doug
Doug
10 years ago

I look at the chart and see correlation before and after the DH. What changed was the longer starts switched from being in the NL to being in the AL. Which makes sense because, as you say in the piece, the AL would no longer have to lift an effective starter for a pinch-hitter. But, that difference is not big – half-an-out longer starts. Ergo, if it’s time to pinch-hit for the pitcher, it’s probably also time to take him out of the game anyway. With such a small difference, it’s not surprising that stronger forces in the game (as… Read more »

RonG
RonG
10 years ago

Why don’t you extend this chart another twenty years to get a better sense of the proposed correlation?

Hartvig
Hartvig
10 years ago

I think that a large part of the decline in the National League starting in 1972 can be attributed to the diminished effectiveness and/or retirement of complete game workhorses Bob Gibson & Juan Marichal plus Fergie Jenkins and Nolan Ryan being traded to the American League. Add having the most successful National League franchise of the period being managed by a guy with the nickname Captain Hook & I think you’ve got your answer.

Hartvig
Hartvig
10 years ago
Reply to  Hartvig

Missed Gaylord Perry moving from San Francisco to Cleveland the year prior as well. Add Bill Stoneman blowing his arm out and by 1973 the only guys you’ve got left in the NL from the CG and IP leaderboards in the early 70’s are Carlton and Seaver.

Brent
Brent
10 years ago
Reply to  Hartvig

And a Niekro or two? (actually it was probably only Phil at that point, but my comment was too short when I just said and a Niekro)

Mike L
Mike L
10 years ago
Reply to  Hartvig

Hartvig @15, the average team throws 1450 innings a season. There were 24 teams in 1973. Roughly 17500 innings per league, and if the average starter was at 6.5 IP per start, that would imply about 10500 innings came from the starters. The average of the top 10 starters in MLB (ex of Wood) was about 300 innings and 40 starts (mind-boggling, but whatever). Take out four horses and replace them with league average and you get a reduction of about 160 IP of starting pitcher innings. That would move the needle even more than the four tenths of a… Read more »

Richard Chester
Richard Chester
10 years ago
Reply to  Mike L

By using the Pitching Split Finder on the PI and a spreadsheet I found that the number of IP by starters in 1973 for both leagues combined was slightly in excess of 25452.

Mike L
Mike L
10 years ago

Richard, my bad. Forgot we were in base 9. If you took four great 300 IP starters in the NL and replaced them with four league average “fifth” starters, and lost 1 inning per start, you would see an overall reduction in NL averages of about one tenth of an inning pitched.

Brent
Brent
10 years ago

Well, can we check out individual managers and see if that explains the blip? For instance, the 1973 Texas Rangers had 35 complete games, while the 1974 Texas Rangers had 62. The difference? Billy Martin or at least that would be my guess.

oneblankspace
oneblankspace
10 years ago

I want to change pitchers. If he can just get one more out, I can use a pinch hitter instead of having to make a double switch. I think I’ll leave him in.

What? I have a DH and the pitcher does not have to bat? He gone.

John Autin
Editor
10 years ago
Reply to  oneblankspace

The point made by o.b.s. shows how the DH cuts both ways in impacting start length. When you don’t have to consider pinch-hitting for the SP, sometimes he stays longer than he otherwise would, and sometimes he leaves sooner.

In 2012, both AL and NL starters averaged 5.9 IP and 25.0 batters. The only difference, probably, is that NL starters are more likely to finish an inning.

John Autin
Editor
10 years ago
Reply to  John Autin

Confirming that no-DH starting pitchers are more likely to complete an inning: I ran the data for 2011-12 combined, for NL-only and AL-only games, counting the number of times the SP lasted a round number of innings from 3 to 9. (I figured less than 3 IP would be mainly injury or disastrous results, thus pinch-hitting would play little role in the decision to remove him.) Then I calculated the percentage of total starts. (1) Overall, NL starters were significantly more likely to complete a round number of innings from 3 to 9, by 70% to 63%. The difference was… Read more »

Doug
Doug
10 years ago
Reply to  John Autin

With the tiny, tiny benches now, the double-switch is probably seldom used, in comparison to past years, as a device for NL managers to make a mid-inning pitching change and keep the new pitcher in the game beyond the end of that inning.

Just a hunch, but I suspect the differences between AL and NL even-inning starts were likely smaller back in the 1980s.