Tango on Baseball Archives

Sheehan: Pitcher Workloads (June 19, 2003)

Joe has an excellent article on pitcher workloads.

He made reference to hitters hacking in the 70s based on anectodal evidence. I have pretty good evidence that hitters were not hacking away in the 60s (see my Koufax article comparing his actual pitch counts versus my estimate, an estimate that assumes that hitters hit the same way now as they did then). If Joe meant that the hitters only hacked in the 70s, then that's something that we can try to verify as well.

The important takeaway in terms of workloads are
- walks and K add more pitches / PA
- higher OBA means more batters / game meaning tougher to get a complete game
- higher level of talent means that pitchers have to work closer to 100% more often
--posted by TangoTiger at 10:23 AM EDT

Posted 10:28 a.m., June 19, 2003 (#1) - Koufax pitch counts (homepage)
.

Posted 12:56 p.m., June 19, 2003 (#2) - Dactyl
Thanks for posting this article. It makes a lot of sense.

Posted 1:24 p.m., June 19, 2003 (#3) - Robert Dudek
The last element, higher level of talent making the picthers work harder is pure speculation. The fact that more hitters hit for power now is manifested in higher walk and higher OBP rates already. Pitchers have learned new pitches and are able to gather information about hitters' weaknesses in more systematic and precise ways than ever before. So... as the hitter's approach has become more scientific, so has the pitcher's.

We are seeing pitchers who seek to reduce the number of pitches they throw by constantly challenging the hitters (i.e. not giving up walks). Curt Schilling, David Wells and Roy Halladay are good examples of this. Greg Maddux has not thrown many pitches per inning and I suspect he has a pitches/IP rate about as low as any post-WW II ace pitcher.

This article would have been much more interestiung if Joe had gotten a hold of a pitch count estimator like Tango's and actually checked how many pitches were being thrown per start by today's aces versus yesteryear's.

Posted 1:52 p.m., June 19, 2003 (#4) - Vardibidian(e-mail) (homepage)
I was also interested in his point that power-hitting middle infielders used to be shifted to outfield spots, but now are encouraged to remain middle infielders. Is there some quantitative backup for this (I suppose by comparing minor league fielding position to major league)? And is the DIPS thing likely to increase the perceived value of slick-fielding, light-hitting shortstops and second basemen?

I recently read Hidden Game of Baseball for the first time, and was struck by the point that there are no really bad fielders anymore (compared to early in the game), and so the value of a good fielder is less, compared to the average. But if that difference was hidden in BABIP, and was attributed to the pitcher, will we now see more emphasis on good fielding?

I haven't done any analysis (and am not capable of doing much), but it seems to me that even in a high-scoring era with a lot of homeruns, it may be efficient to have good fielding up the middle even if it leaves a hole in the lineup, in part because of getting starting pitchers deeper into the game (and, in the longer term, boosting the perceived value of the pitcher for trade purposes).

Thank you,
-Vardibidian.

Posted 2:53 p.m., June 19, 2003 (#5) - RossCW
I don't see any evidence for the claim that more BB and K's means more pitches are being used. It seems to make sense on the surface, but are there any studies that show what the current average number of pitches is before a ball is put in play, compared to K's and walks?

Even if you look at current data that does not mean the same percentages were in play 20 or 40 years ago when striking out was considered a much worse outcome. Batters would choke up and focus on just making contact to avoid getting a third strike. That could easily mean many more foul balls and weak hits in play after lengthy at bats than presently.

- higher OBA means more batters / game meaning tougher to get a complete game

Again - is there solid evidence for this - or is it speculation?

A rough OBA of (H+BB)/(AB+BB) shows MLB OBA in the 50's to be consistently higher than last year's. It was high in 1999 and 2000, but last year it was .327, slightly below the average of .328 since 1920 and in 2001 it was right at the average.

I also am not convinced that OBA is a placeholder for the number of pitches a pitcher throws - especially across eras. There are a lot of other variables that effect pitches thrown.

Posted 3:21 p.m., June 19, 2003 (#6) - eric (homepage)
Ross,

I believe it was Tango who presented some study that showed there are more pitches during an average AB that results in a K or a BB than one that results in a BIP.

This is intuitive. You need a minimum of three pitches to get a K, so the average has to be higher than three for a K, and a minimum of four pitches to get a BB, so the average has to be higher than four fora BB. You only need to throw one pitch to get a BIP. So all those one- or two-pitch ABs should pull down the average number of pitches for an AB that results in a BIP.

In regards to the article, it's funny that Sutcliffe would talk about how many pitches he threw, and cite specific games. Any pitcher has high pitch-count games. In the homepage link is a game log of Clemens' 1996 season...three games over 150 pitches with a high of 161 (which was accomplished in 7.2 innings!). Randy Johnson likewise has high pitch counts, and I'm sure most pitchers that have had a decently long career have that one game where he got left in for a very long time (at least pitchers who came up before about five years ago). Oh well, I guess people on here don't need to be told that Sutcliffe says stupid stuff sometimes.

Posted 3:37 p.m., June 19, 2003 (#7) - tangotiger
Varb: it would be interesting to see if the spread of hitting talent is tighter across years, and not concentrate so much on the position.

Rob: yes, pure speculation, but a variable to consider.

Ross: there are about 5.4 pitches thrown per BB, 4.8 per K, and 3.3 per BIP. So, if you have a pitcher with lots of (BB+K) / PA, you can bet that that pitcher will have more pitches / PA than league average. Dan Quisenberry I will wager had the lowest pitch / PA count of any pitcher with at least 1000 PA in the last 50 years. (If not lowest, then in the bottom 10). Just for fun, go look at any active pitcher's (BB+K) / PA, and you will see a direct correlation to pitches / PA. MLB has the data from 1998-2003 on their site. Benitez is I think the #1 guy in (BB+K)/PA and pitches/PA (or pretty close).

Is this what you are getting at? Otherwise, please restate the concern.

- higher OBA means more batters / game meaning tougher to get a complete game

I'm not sure why you are talking about the 50s, when the low OBA was probably in 1967/68, but let's say there are 25 batting outs / 27 outs (I realize that a pitcher with few baserunners will be closer to 27/27).

Anyway, if your OBA is .300, that means .3 safe plays per .7 outs, or 10.7 safe plays per 25 batting outs, or 35.7 batters. A .350 OBA works out to 38.5 batters. (Just an example here). So, as the OBA rises, so does the number of batters per 27 outs. With more batters available because of the run environment/playing conditions, the more pitches needed overall.

I agree with your last point that there are more variables at play for pitches thrown across era. I've got a working model, though I have no data to validate it against. Essentially, it shows Cy Young with, if I remember right, 2.8 pitches / batter, and Nolan Ryan with 3.9 pitches / batter. (or somewhere around there.) Eventually, I'll present a mathematical proof that shows the relationship between K,BB,BIP and pitches thrown. This is as much as I can say for the moment.

Posted 4:34 p.m., June 19, 2003 (#8) - tangotiger
This article would have been much more interestiung if Joe had gotten a hold of a pitch count estimator like Tango's and actually checked how many pitches were being thrown per start by today's aces versus yesteryear's.

To expand on this point, and tie it in to Ross's point, our estimators are based on "all things being equal". I am reasonably confidant that the extended pitch count model (unpublished) that I have works with the 1990s to today pitchers. I wasn't certain how well it would hold up to the 1960s pitchers, and so I was very happy to see how well it matched to Koufax.

So, for Koufax, and for the 90s pitchers, the way batter/pitcher matchup exist in terms of "all things" we can say that they are pretty equal. Because it worked on Koufax, will it work on all other pitchers of his era? I would bet yes, but I'm not sure. If it works in the 60s and 90s, should it work in the 70s? I would say yes, but Sheehan brings up a good point about maybe baseball changed the way they played in the 70s. I can't just discount it, especially since Ryan and Carlton and a few others had kinda high pitch counts. If you look at the Ryan progression, he completely tails off at around age 30. It's kind of remarkable, and I don't know if he was injured or what happened. Maybe the Angels didn't want to wear him out so he can resign with them as a free agent? But again, I would bet that if we really looked at it, baseball was the same from the 50s to today, in terms of batter/pitcher matchups, and how often balls/strikes were thrown, etc.

But how about earlier? Well, again, the more you go back, the less likely things are the same. You get to the point where you have very few BB+K per PA across the whole league. Those pitcher/batter matchups may not be "all things equal".

To get back to the point, I agree the article would have been enhanced if he used my estimator or someone else's (it seems Nate/BP also has an estimator). However, the article is excellent, and gives a good push for people to tackle some of the issues, most notably, the talent distribution across eras.

Posted 4:37 p.m., June 19, 2003 (#9) - RossCW
there are about 5.4 pitches thrown per BB, 4.8 per K, and 3.3 per BIP

Has that been consistent across era's?

So, if you have a pitcher with lots of (BB+K) / PA, you can bet that that pitcher will have more pitches / PA than league average

Well you can bet on it, but it may not be true. The average pitches thrown for each category may vary widely by pitcher.

Is this what you are getting at?

No. Although it is part of it. The question is what are the effects of the difference in K's and walks between eras, which is the way Sheehan is using it. It is perfectly possible that 50 years ago more 2 strike counts ended in a ball in play rather than a third strike. That does not change the number of pitches at all.

I'm not sure why you are talking about the 50s, when the low OBA was probably in 1967/68

The point is today's OBA are not really historically higher than other era's - they are about average. I don't have the numbers in front of me, but MLB OBA appears to be quite volatile from one year to the next.

So, as the OBA rises, so does the number of batters per 27 outs.

Assuming that the number of double plays, caught stealings and runners thrown out advancing are all the same. I don't think they are.

With more batters available because of the run environment/playing conditions, the more pitches needed overall.

Assuming the number of pitches per batter is constant for all pitchers and regardless of the number of batters.

I agree with your last point that there are more variables at play for pitches thrown across era

I think there is a fundamental problem with using average results for comparing elite pitchers. The question is does Roger Clemens have to face more batters and throw more pitches to accomplish the same things Steve Carlton did. The fact that Sean Bergman gave up a lot of baserunners or Bobby Witt struck out a lot of batters and walked a lot of batters doesn't seem to have much to do with that. And yet when we compare era totals they are included in the analysis - it may be there were fewer Bobby Witt's and Sean Bergman's pitching in the mid-60's but that doesn't make Steve Carlton's job easier than Clemens.

Posted 4:51 p.m., June 19, 2003 (#10) - RossCW
Of course Carlton didn't really pitch in the mid-60's but the point it the same if you use mid-70's or Bob Gibson.

Posted 4:58 p.m., June 19, 2003 (#11) - tangotiger
there are about 5.4 pitches thrown per BB, 4.8 per K, and 3.3 per BIP

Has that been consistent across era's?

It's not even consistent among pitchers of the same era. However, I've got a working model that takes the "type" of pitcher you are, and spits out the pitches / event. It's quite logical, but there's room for error. Because of the way that model works, it's transportable to other eras, assuming that the way batters/pitchers approach each other in terms of throwing/looking for strikes is the same. For this reason, I'm not too crazy about publishing this until I get some data to at least point me to the right answer.

So, if you have a pitcher with lots of (BB+K) / PA, you can bet that that pitcher will have more pitches / PA than league average

Well you can bet on it, but it may not be true. The average pitches thrown for each category may vary widely by pitcher.

It actually does vary widely. But, I think eric's point explained it very well in post 6. I won't extend my answer more than that, and my bet would be there. Of course I could be wrong, but that's why betting is fun!

Is this what you are getting at?

Sure, it's possible. As I said, we don't know what the strike/ball matchups were for pitcher/batter.

So, as the OBA rises, so does the number of batters per 27 outs.

Assuming that the number of double plays, caught stealings and runners thrown out advancing are all the same. I don't think they are.

I agree, which is why I added my provision. But again, what's the impact here? I'm sure I can come up with an r of over .95 between OBA and PA/27 outs historically (if I had the data back then).

With more batters available because of the run environment/playing conditions, the more pitches needed overall.

Assuming the number of pitches per batter is constant for all pitchers and regardless of the number of batters.

Assuming the same K and BB rates, I'd say yes, you are right about your first part.

I agree with your last point that there are more variables at play for pitches thrown across era

I think I understand what you are saying, but then I don't. I agree that you don't necessarily want to use the basic pitch count estimator, as it won't work too well on the extreme pitchers. But then again, that basic estimator was 2% off for Clemens and Koufax compared to their actual totals.

Posted 7:51 p.m., June 19, 2003 (#12) - tangotiger
Ross, just so I'm not going to go nuts again, here's what I'm going to do. You let me know if this will satisfy the issue.

Select
- all starter seasons
- from 99 to 02
- min 400 PA in that season

Calculate
- seasonal OBP
- seasonal PA / 9IP

Then run a regression of OBP against PA/9IP. I expect to get an r over .90, and more likely over .95.

My contention is that the more runners on base / batter faced, the more batters at the plate / 9ip, regardless of pitcher. This I think is rather obvious, so rather than wasting my time again, you tell me what you want me to run, and what contention you are positing. Thanks...

Posted 10:54 p.m., June 19, 2003 (#13) - Chris Dial
Part of the flaw, unless I missed an explanation, is that in 1993, the league got juiced. Then again in 1994.

These raw numbers don't mean much to me. What is the change *relative to the league* at these positions. Heck, the league ISO and SLG has gone up more than these positions have.

I looked it up - I'm right:
as a function of league average (2B Slg/Lge SLG)*100

Yes, it is better to subtract the group in qquestion, but that's too much work for this.

2B SLG pretty much has always been 93. 2B ISO has *dropped* by 10% from the earlier part of the century (the last century) - it averaged 88 from 03-48 and 75 from 53-03. SS SLG - 88-90 SS ISO - 72, but has fluctuated from 50 to 100 (in 1908). C SLG ~94 since 1938. C ISO - jumped into the 90s in 1953, and has stayed up there.

Stat Avg SD
2B slg 93 3.0
2B iso 81 7.8
SS slg 89 4.7
SS iso 75 11.2
C slg 92 3.8
C iso 88 8.1

And only the K rate has increased. The BB rate in Joe's article is pretty close to constant. Adding the two means little.

So, whether or not pitchers have to work harder, I don't know, but Joe's dataset doesn't appear to say anything like that *at all*.

Posted 10:59 p.m., June 19, 2003 (#14) - Chris Dial
My data
Year 2b slg 2B ISO SS SLG SS ISO C SLG C ISO
1903 95 94 93 91 87 83
1908 98 91 100 101 87 78
1913 99 88 89 80 89 81
1918 91 85 92 80 84 72
1923 97 93 87 70 89 81
1928 94 86 85 71 88 80
1933 94 87 91 83 91 80
1938 92 83 86 71 95 90
1943 95 86 91 75 91 78
1948 92 82 92 81 89 86
1953 92 78 89 75 94 92
1958 88 71 86 73 96 98
1963 90 73 87 68 97 96
1968 91 69 82 61 92 86
1973 92 74 77 50 95 94
1978 88 69 86 60 95 97
1983 93 76 88 82 95 95
1988 90 73 90 74 94 92
1993 92 78 88 68 97 100
1998 93 78 88 73 93 89
2003 96 85 95 87 98 97

Obviously the 2003 data may be skewed by sample size.

Posted 11:44 p.m., June 19, 2003 (#15) - RossCW
I think I understand what you are saying, but then I don't.

Well I'm losing track of the issue I orginally raised - the question whether you can conclude that pitchers in earlier era's had an easier time completing games.

One argument for this idea, as I understood it, was that there are higher OBA today and so pitchers had to pitch to more batters to complete a game. Another was that players get more walks and strikeouts. My point was this is pitcher specific. There is no reason to think that batters are more likely to get on base against the same quality of pitching today than they were in earlier eras. The same is true of K's and BB's. So there is no reason to think that a pitcher of the same quality will need to throw more pitches to complete a game today than they did 20 or 30 years ago.

I agree there is a close relationship for an individual pitcher between their opponents OBP and PA/9IP. There may be some variations in CS, pickoffs, double-plays and batters thrown out but I doubt they are significant. When you compare different eras I am not sure that is as true. When I looked at CS per base runner the rate was not constant.

Posted 9:36 a.m., June 20, 2003 (#16) - Rally Monkey
"So there is no reason to think that a pitcher of the same quality will need to throw more pitches to complete a game today than they did 20 or 30 years ago."

In other words, a specific pitcher who allows 8 hits, 3 walks, and 7 strikeouts per 9 innings in 1920 (if there was one) should be expected to throw about as many pitches per inning as someone in 2003 who has the same numbers. Is that what you are saying?

Sheehan says it was easier to complete games back when. True, for an average pitcher who might strike out 3.3 per 9 innings. For Bob Feller, who one year walked 153 and struck out 348, he had to be throwing many more pitches than pitchers of today. The only example I have for him is a 1 hitter where he walked 6 and struck out 6. That game was 136 pitches. Since Feller had many games with 10+ k's and some with 10+ walks, and completed most of his games, he was bound to have some incredible pitch counts.

Posted 11:30 a.m., June 20, 2003 (#17) - tangotiger (homepage)
So there is no reason to think that a pitcher of the same quality will need to throw more pitches to complete a game today than they did 20 or 30 years ago.

I don't think I was really talking about this, but I agree. The number of pitches a pitcher has to throw is dependent on the style of the pitcher/batter, in which both are dependent on the skillset of the pitcher/batter and the "run environment", which is partially provided by the pitcher/batter in question.

This is neatly evidenced by Bartolo Colon 2001/2002, who must have changed his pitching style drastically (assuming he maintained the same quality) in order for him to have that change in pitch count, walks and K changes, etc.

See above link for his stats. Looking at his 2002 to today (about 1400 batters), his pitches/batter was: 3.6. In 1999-2001, he averaged 4.0 pitches/batter.

His (K+BB)/PA rates in 1999-2001 was: 32%. In 2002 to today, he's at: 23%.

Looking at his HR/H:
1999-2001: 12.5%
2002-today: 10.6%

His (BB+H)/PA:
1999-2001: .319
2002-today: .297

So, Bartolo changed his style to the point that batters were no longer going deep in the count, probably because Colon was giving what looked like more "hittable" pitches (though they probably weren't). The net effect is that by not going deep in the count (or by batters not taking Colon deep, who knows), Colon:
- reduced his pitches/batter
- reduced his walk and K rates per batter
and as a side effect
- improved his overall performance level (though not necessarily to a statistically significant degree... I didn't check).... that is, maybe Colon skillset remained the same, but he found a more optimal pitching style to increase his performance level.

Posted 12:22 p.m., June 20, 2003 (#18) - tangotiger
In 2002, Colon went to 3 balls or 2 strikes: 501/966: 51.9%

In 1999-2001, he did that: 41.6%

Boy, that's not at all what I was expecting. Assuming I didn't make a mistake somewhere, Colon managed to get to 3 balls or 2 strikes MORE when not concentrating on striking players out. Yet, somehow, he ended up with fewer pitches/batter. Not sure about 2-strike fouls changing. I can only guess he had more pitcher's counts, so that when he got to 2 strikes, he did that with getting fewer balls.

Let's see.... ok, here's how many balls/PA thrown, when Colon managed to get to 2 strikes:
year.. PA with 2 strikes... balls/PA
2002 455 1.58
2001 459 1.77
2000 432 1.77

Interestingly, he managed to get to 2 strikes while throwing fewer balls, which may explain some of his good performance (more pitchers counts, and fewer pitches to get there).

Posted 4:19 p.m., June 20, 2003 (#19) - RossCW
In other words, a specific pitcher who allows 8 hits, 3 walks, and 7 strikeouts per 9 innings in 1920 (if there was one) should be expected to throw about as many pitches per inning as someone in 2003 who has the same numbers. Is that what you are saying?

It seems obvious that that is true. But a pitcher of the same quality would also be expected to have the same number of hits walks and strikeouts per 9 innings in both eras.

It may be that any overall increase in OBA is an effect of the best pitchers pitching a smaller percentage of the total innings rather than a cause.

Posted 4:56 p.m., June 20, 2003 (#20) - Rally Monkey
"But a pitcher of the same quality would also be expected to have the same number of hits walks and strikeouts per 9 innings in both eras."

This has nothing to do with pitch counts, but I disagree here. Two pitchers of the same quality should be expected to have the same k's, walks, hits only if they pitch in the same circumstance. The differences in level of competition, size of hitters, hitters approaches, ballparks, fielding ability will affect all the pitchers numbers. How many homeruns would Ramon Ortiz give up in the dead ball era? I'm sure it wouldn't be 40. How many K's per inning would Randy Johnson have in the 1930's, when hitters used heavier bats and many were more concerned about making contact than hitting for power? Probably less than he has now.

Posted 7:37 p.m., June 20, 2003 (#21) - RossCW
Two pitchers of the same quality should be expected to have the same k's, walks, hits only if they pitch in the same circumstance. The differences in level of competition, size of hitters, hitters approaches, ballparks, fielding ability will affect all the pitchers numbers.

I agree.

How many K's per inning would Randy Johnson have in the 1930's, when hitters used heavier bats and many were more concerned about making contact than hitting for power? Probably less than he has now.

But then what is the common measure of quality? Its possible Randy Johnson wouldn't have been at all successful in the 30's. Its possible Walter Johnson wouldn't be successful today. It seems to me that we need to assume that pitchers with the same numbers in different era's were of similar ability.

Posted 7:51 p.m., June 20, 2003 (#22) - Rally Monkey
I agree with what you say about Randy and Walter. We have no way of knowing how successful these guys would be in different eras. Only Dr. Emmett Brown has the means of findng out.

But assuming that pitchers with similar numbers in different eras are of similar ability, are you talking about absolute or relative numbers?

I would assume a pitcher with a 4.00 ERA in 1996 would be a much better pitcher than one in 1968. A pitcher who strikes out 6 per 9 innings would be a fireballer in 1920, but run of the mill in 2003.

If you are talking about ERA+, then its a reasonable assumption that 2 pitchers with a 120 ERA+ in different times are of similar ability, although it may not be right.