THE BOOK--Playing The Percentages In Baseball

It is hard to tell from these interviews what is fluff and what is not.  It really is.

I can’t say that I didn’t raise my eyebrows several times while reading this one.

Was it really necessary to ask Will which is worse, playing on turf or on grass?  Seriously, is there ANYONE reading this interview that doesn’t know the answer to that question?  “Gee, will, is it better to swing at a strike or a ball?  Do you wear your glove on your throwing hand or the other one?”

No statistical took has ever had more impact on the game than PAP?  I like Will too and he writes lots of things of value (and writes well), but can we take anything else in the interview seriously after reading that?

What about ‘Pitcher injury does not involve luck, especially in the long run? (I paraphrased)’ Huh??

What would constitute disproving pap?  Or proving it, for that matter?

We’ve got Randy Johnson, who put up pitch count totals that push his pap off the charts, yet at age 44 he’s never had an arm injury, its his back and knees giving out on him.

Nolan Ryan’s pitch counts from the mid 70’s would make Randy look like Greg Maddux.

PAP is predicated on high pitch counts over a series of starts being detrimental to a pitcher’s health (or at least constuting “abuse").

But, this was developed and tested with pitchers since around 2000 or so.  But, that’s right around when pitchers stopped pitching such high pitch count games.  If you looked at 1970s pitchers (more pitches per start, more starts per time period), it would knock PAP for a loop. 

And yet, in the time period where pitchers were most “abused”, we still see pitchers facing as many batters then (if not more) as today.

My problem is that BP has never gone back to the 1960s and 70s to test PAP.

There’s very much an ESPN type atmosphere to some of BP, where if they know their stats are not reliable, they’ll just keep them and not mention the downsides of the stats, thus why there’s WARP with a replacement level that’s too low, PAP not being shown to work in the 70s, and defensive stats that come from some Davenport Translations but they’ve never shown them to be reliable. It’s kind of like when ESPN came up with their arbitrarily created player valuation system, but didn’t explain how the numbers came about nor show that they correlated with anything.

Would it be that difficult to run some statistical tests (e.g., regressions) to see whether some version of PAP correlates with either poor future performance or injury (using DL days or missed starts or career-ending injury, etc.)?

I vaguely recall Rany (or whomever) doing that in his origianl (or ONE of the original) articles.

And yes, it would be nice to run those tests on various eras in baseball, as there is a difference between, say, 120 pitch games being dangerous when pitchers don’t throw 120 pitch games anymore (more or less) and when pitchers routinely throw 120 pitch games.  In fact, it would be ironic (if that is the right word) if the only reason why high pitch count games (high PAP) are detrimental is because pitchers don’t throw that way anymore.

IOW, let’s say that throwing a lot of pitches was not detrimental.  Now, let’s say that you thought it was and you limited your pitchers, more or less, to 100 pitches max per game.  Now let’s say that occasionally a pitcher throws one or more games with more than 100 pitches and that ends up causing a problem.  You just nedlessly screwed yourself.

Of course, one thing that people, especially the MSM when they complain that starters do not go deep enough into games anymore, seem to forget or not realize, is that is generally a GOOD thing for a team to bring in a fresh reliever as early as possible, especially if the starter is not a premium one.  And even if the starter is a premier one, a fresh closer and sometimes another short reliever, is usually BETTER than the starter who has faced the lineup 3 or 4 times already.

So if I am running a team, it is not so much that I want my starters out of the game early to preserve their arm, but that I generally want a fresh reliever in there as soon as possible, especially if I have a good bullpen.

"whomever” should be “whoever” of course, in case there are any grammar police lurking…

I don’t understand why BP is considered the gold standard if they don’t have to prove anything they do. They’re using outdated rating systems (FRAA, VORP), and haven’t improved on them in quite a while. A site like the Hardball Times does things the right way, in proving their information and testing it, but doesn’t get nearly the respect it should.
Bill James had some kind of list in SI using his new young talent inventory stat, but didn’t explain anything, not even the basis behind it. Every team had a seemingly random listing of young players (the yankees list didn’t include Hughes or Joba), with a point value next to it. Of course James is given the benefit of the doubt and his YTI will probably be explained, but people accept it as fact without an explanation, just like BP does.

Adam, good comp on ESPN/BP. BP seems to be interested more in promoting itself than promoting sabermetrics.

When you do something for money EVERYTHING changes.  The bigger the money, the more things tend to change.

The goal becomes how to make more money (and/or get more recognition) at the expense of integrity.

MGL/irony: there is some indication of this.  For example, when I ran my study for 99/02, it showed pitchers were not as good on 3-days rest as 4-days, and even better at 5-days.

Woolner ran his test for the 1970 or 1980s in BBTN, and he showed, IIRC, just as good (maybe better?) performance for 3-days and 4-days rest.

This can be explained with MGL’s posit that you pitch as well as you are conditioned.

***

Dan, the YTI is explained in the BJ Handbook 2008, but it’s just a simple metric, with no theory or basis behind it.  It’s even difficult to figure out what the idea behind it is, as it all depends on how young you think Pujols/Howard/Utley are, and whether the amount of “young talent” they have left is more or less than Nick Markakis.

Sometimes I get confused when I come on here and see comments I don’t remember posting, only to realize that apparently there are more people than myself named Dan. 

Perhaps the tangotiger’s and Pizza Cutter’s have the right idea with respect to handles.

Names are intended to identify someone… uniquely, preferably.  The only other Tango I know is someone else’s pet.  So, I’m pretty safe.

Actually, just an other Dan.

You folks are being waaaay too kind here.

PAP = Rubbish.  Nonsense on stilts.

What precisely are we to make of a system that tells us the Nolan Ryan, Livan Hernandez, and Randy Johnson would never pitch past the age of 30?

Oh, I see.  They’re “different.” We have to throw them out of the analysis.

And of course, we’ll know they’re different because ... well, because they don’t fit the model.  Because they throw hundreds and hundreds of pitches and keep coming back for more.  They’re different because they’re genetically gifted, with the perfect power pitcher’s body (Ryan), or because they’re freaks of nature (hey, we’ve never had a 7-foot tall pitcher, so Randy’s an outlier!), or because they’re cruisers who can turn it on and off at will (Livan).

So knowing this information, a savvy GM could take Will Carrol’s Pitcher Abuse Points spreadsheet and do ... and do what?  You tell me.  Except he won’t.

I kind of feel sorry for him because he published that faux “insider” book on steroids that he titled “The Juice” approximately two hours before Jose Canseco was on every talk show ever known to mankind pumping his “Juiced.” Kind of sorry, because his book was just like his PAPs:  nonsense.

Just in case anyone thinks I’m being unfair to Will Carroll, here’s the PAP leaders for 2006:

http://www.baseballprospectus.com/statistics/sortable/index.php?cid=100934

Livan:  same old Livan in ‘07
Zambrano:  same
Willis:  poor season, but no apparent arm trouble/loss of velocity, etc.
Schmidt:  Bingo!  A winner!!
Arroyo:  no ill effect
Smoltz:  ditto
Zito:  poor season (although not really way out of line with reasonable expectations), again not because of any arm trouble
Cain:  excellent season
Peavy:  superb season
Meche:  very good season
Myers: Bingo!  A semi-winner!!
Redman:  crap, but that was expected, and no arm trouble
Lowry:  same as always
Lackey:  really sucked, didn’t he ... Scioscia totally ruined him.  Artie Moreno should’ve traded him for Wilson Betemit.  If only he’d listened to Will Carroll
Schilling:  pitched to expectations
Sabathia:  another guy ruined and useless the next year!
Vazquez:  his best season since 2001
Davis:  no ill effects
Pettite:  very good season
Batista:  nice season
Cabrera:  very nice season
Suppan: same as always
Francis:  best season yet
Lilly:  ditto

So those were the “25 Most Abused Pitchers” in 2006.

Out of those 25, Schmidt’s arm blew out.  One out of 25 pitchers.  Four percent.  Out of any randomly selected 25 starters who make, say, 20 starts in a season, how many will have serious arm trouble the following year?  Could it be more than one?

If any GM is using PAP, he ought to be fired.  It really is that bad.

Other Dan,

I’ve done that so many times on this site and I get confused every time, “hhmmm, I couldn’t have been that far gone last time i came here?”

I’m gonna start using a more unique name (sdanne) following this post.

The problem I have with studies of so-called pitcher abuse and pitch counts, etc., especially when comparisons are made with earlier eras, is that they fail to take account of the damage that could be done to the arms of pitchers before they even become pros. A few years back ESPN did a feature on their Outside The Lines program about the abuse being done to kids’ arms in their high school years. A number of things that are different in this era were pointed out. For example, Tom House mentioned that nowadays kids often only play ball in organized settings and many of them only throw when pitching. He claims that they don’t just go out and play catch. Thus their arms aren’t strengthened. They also talked to some teenaged pitching prospects who said that they didn’t worry about hurting their arms because, if they did, they would just go get surgery to get them fixed. (A stupid idea, but there it is.)

My point is that it could be that the environment in which the young play these days has changed, leading to more arm damage that affects the durability of pitchers when they become pros. The trouble is that there’s no real way to test such an hypothesis since the data isn’t available but, as far as I’m concerned, it’s impossible to make legitimate comparisons with previous generations of pitchers and obtain any clear results without access to complete histories of the pitchers.

Leaving aside PAP, why is it that there are so few complete games these days compared to, say, three or four decades ago (even among those “top 25 in PAP")?  Because managers believe in PAP (or are just counting pitches) and are making the stupid decision to ‘save the pitcher’ from risk of injury? Because pitchers decline in effectiveness at an accelerated rate after a certain number of pitches? Because batters are behaving differently? Or there some other reason or combination of reasons?

In short, what happened to the complete game, and why?

I do recall Dodger broadcasts back in the early 60’s in which pitch counts were routinely reported, at least at the end of the game. I don’t know how common that was, but I think Alan Roth must have been compiling such stats at that time and feeding them to Vin Scully and Jerry Doggett.

You can find pitch counts for Dodgers pitchers of old at my site (click my name), and look for “pitch counts”.

The average number of pitches Koufax threw per start is EXACTLY the same as pitchers in the 1980s/90s, and LESS than in the 1970s.

The standard deviation of pitches per start is far far higher in the 1950s/60s pitchers than today.  That is, they not only went 130+ pitches far more, but they went 70- pitches far more too.  Even Koufax.

At least managers today see fit to let pitchers pitch out of jams, rather than back then, when managers will routinely pull pitchers because they “didn’t have it”.

Added note.  It would be circular reasoning to argue that the reason there are fewer CG’s in recent decades is the rise of relief specialists.

(It would also be ahistorical to attribute the decline in CG’s to “PAP,” since the latter is a more recent invention than the decline in CG’s. Also, I think we started “saving the arms” of young pitchers, i.e., in Little League, at least a couple of decades ago not just by outlawing the curve ball; only more recently have strict rules been implemented about pitch counts but I’m not sure when that happened.)

So how do we come up with a non-circular and non-tautological explanation for the decline of CG’s?

Thanks, Tango for the link to your pitch count summary. (I wrote my previous message without seeing yours.) The figures for Koufax comport with my memory.

But it’s also likely true that the average starter (or the average highly successful one?) was more likely to complete the game with that ca. 110 pitches, whereas the later-day pitchers do not.

So wouldn’t a CG/Pitches ratio for starters would be far lower today than in Koufax’s day?  And if so, is it just a change of style of the game, or perhaps, as you are suggesting, the avoidance of extremely high pitch counts for ineffective pitchers?  (Again, trying to avoid a circular argument here.)

http://www.baseball-reference.com/d/drysddo01.shtml

Open up the PITCH DATA SUMMARY section.  3.50 pitches per PA.  That’s a nimbler today, a Brad Radke type.

And instead of pitching 34 starts per season, he’s at 39.

From 1962-1965, he faced 1270 batters per year on 41 starts, or 31 batters per start, as opposed to the 27 of today.  He faces an extra 4 batters, which is one inning.

31 batters x 3.50 = 108 pitches per start.  In the mid 90s, that would not lead the league.  These days, he’d just be right at the top.  Pedro in the 2000s had 107 pitches per start a year or two.

So, he gets to pitch an extra inning, just because hitters aren’t waiting for more pitches, he throws 5 or 6 extra starts,and managers are more likely to let a guy who put in 8 innings go for 9, because they are ready to take him out after 5 innings if they have to.

Tango/#17:

One of the things that we are seeing today - which probably contributes to the reduction in complete games - is an increase in the number of starts in which a pitcher struggles early but stays in the game for 5-6 innings.

I should note, also, that it’s not just starters who are seeing innings reductions - relief pitchers are facing fewer hitters per appearance, too.

Thanks, Tango.  So (1) hitters are more patient,” which in effect means that if pitchers throw 108 pitches they will throw one fewer inning on average today; (2) managers of yesteryear reached early decisions to pull pitchers who started poorly, thereby assuring that they had more starts than they might have if they were allowed to finish 5 or 6 innings (to what extent can this be shown in the data? are “modern managers” maybe realizing that back luck can lead to a poor first couple of innings?); (3) if a pitcher gets to 8 innings, the manager is likely to give him a chance to complete the game (contingent on size of lead?). Perhaps another factor is the number of pitchers in the basic rotation, and hence the number of starts the top pitchers can expect to have.

Seems plausible that these are explanatory factors, but perhaps begging the question about the changing use of the quick hook over time, and how much it may be contingent on the pitcher’s performance and the score of the game (or the overall run-scoring environment in different eras). 

Also, the drop in total CG’s still seems rather more steep than might be accounted for in the terms explained above.  Perhaps need to do both intertemporal and comparative cross-sectional analysis (within a season in 1960’s and within on in the 1990’s or 2000’s).

Anyway, this is all very interesting.  Thanks a lot.

I’m

The “in for 8, go for 9” is a supposition of mine. The rest is basically fact.

Nobody thinks it odd/worrisome that Matsuzaka lead MLB in PAP this year given all the concerns about the longer schedule and not having as much rest between starts? Or that I shouldn’t be worried as a Jays fan that Burnett was so high especially given the missed time?  I’m just saying…

I think Tango’s suggestion to look at PAP scores from the 70s would be VERY interesting.

MGL/#5 has it bang on.  Wise management will realize that the fresh arm is more effective, separate and apart from any injury risk. 

What has not fully been appreciated in the move to the LaRussa bullpen, with declining number of batters faced per relief outing, is the effect on the “pull the starter” decision.  The LaRussa bullpen usage is predicated on pushing starters in the middle innings, as, over a season, if the manager doesn’t, he will wear out the pen. 

I am still waiting for a team who bucks the trend and goes to a smaller bullpen with longer outings the norm, as we had in the late 70s.

Complete games have been declining for 100 years. It’s a long-term trend and pitch counts per se have only become an issue recently.

There are no doubt a number of factors. Probably one of them is that managers have observed that a fresh reliever is better than a tired starter and there are great benefits to getting a platoon advantage in key late-inning situations (both of which Tango and mgl have pointed out). Complete games used to be a big deal in evaluating pitchers, and it was a stigma to be a 7-inning pitcher. Over time, this has eroded.

I’m no fan of PAP, but I don’t think it’s appropriate to judge its effectiveness on 1970s pitchers. IMHO, the mindset for starters back then was completely different than it is today. If pitching a complete game is more or less expected every start (most of which are on just 3 days rest with the four man rotation), pitchers would naturally adjust their output to compensate. Just as a runner adjusts his pace according to the length of the race he’s beginning, so too would a starting pitcher adjust his effort.

Now, I’m not a big league pitcher, so I don’t KNOW the above is true. However, I highly suspect it is. John Smoltz regularly hit the mid to high 90s on the radar gun as the Braves closer. Since being moved back to the rotation, he rarely tops 95mph and regularly sits in the 93-94mph range. Josh Kalk’s work with pitch/fx puts his avg fastball at 93.5mph. Smoltz is clearly regulating his effort differently as a starter than he did as a closer. If he knew he was expected to throw 130 pitches on average, instead of just in rare exceptions, I have little doubt he would regulate his effort even more.

Drydale in 1963, with 17 CG in 42 starts, ERA+ of 114:
http://www.baseball-reference.com/pi/gl.cgi?n1=drysddo01&year=1963&t=p

In those 17 CG, he averaged 114 pitches per start. In his other 25 starts, he averaged 92 pitches per start.

In his 17 starts with the most pitches thrown, he averaged 117 pitches per start.  In his other 25 starts, he averaged 90 pitches per start.

Let’s take the 1991 Greg Maddux, with 7 CG in 37 starts in 1991, and ERA+ of 115:
http://www.baseball-reference.com/pi/gl.cgi?n1=maddugr01&t=p&year=1991

In his 18 longest outings (min 8 IP), he averaged 112 pitches per start.  In his other 19 starts, he averaged 88 pitches per start.

In his 15 starts with the most pitches thrown (40% of his starts, like Drysdale’s CG), he averaged 117 pitches per start.  Just like Drysdale.  In his remaining starts, he averaged 88 pitches per start (compared to 90 for Drysdale).

In all, 100 pitches per Maddux start in 1991, and 101 pitches per Drysdale start in 1963.  The standard deviation of pitches per start was 18 for Maddux and 18 for Drysdale. Drysdale faced 30 batters for those 101 pitches, while Maddux faced 29 for 100 pitches.

In short, you have two pitchers throwing the exact same number of pitches per game, with the exact same distribution, facing virtually the same number of batters, and yet one pitcher completed 40% of his games, while the other completed 19%.

Don’t look at “complete games” as describing something that you think it should.  It simply doesn’t.

Note: these were actually the first two pitchers and the first two seasons I selected.  I did not try to canvass for anything other than having matching ERA+.

tango: was #28 in response to #27 or #26? I’m not trying to be difficult, I just want to understand the point you are trying to make.

Kyle: are you suggesting that Tom Seaver, Steve Carlton, Nolan Ryan, and all the other 250 game winners of that time period could have been even more effective if they pitched harder than they did, that they somehow coasted more on 120-pitch starts, than starters today do in 105-pitch starts?

It’s possible.  The data is there to prove it if you like.  Comparing Smoltz-starter to Smoltz-reliever, while accurate in your assessment, doesn’t shed the light we need.

Here’s what I found when I looked at 1977 pitchers:
http://www.baseball-reference.com/pi/psplit.cgi?lg=ML&team=TOT&year=1977

Go to the section that says:
Times Facing Opp. in Game

Realize that the “1st time” is polluted with relievers and starters, so jump to the “2nd time” which is mostly starters and “3rd time+” which is virtually all starters:
.328/.411 2nd time
.336/.425 3rd time+

And in 2007:
.336/.433 2nd time
.352/.467 3rd time+

So, there is some indication here that 2007 pitchers are not pacing themselves like the 1977 pitchers.

***

However, what does that have to do with PAP?  PAP is saying pitchers arms are abused when they reach certain threshholds after a series of games.  What you are adding as a qualifier is “expectation of pacing”.  That depending on how much pacing you have, that uppers or lowers your threshhold.  That the threshhold for non-pacing starters is 100, and for pacing-starters is 115 or whatnot.

In order to know if an arm is being abused, you need to know the percentage of “full non-paced” usage there is.  That if Smoltz pitches to “95%” full speed, that his threshhold is 100, but that if he pitches to “92%” full speed, his threshhold is 115.

Again, what does this have to do with PAP?  Nothing.

If you want to talk about pitch counts, performance, and pacing (PPP), that’s fine.  I’m on board with that.  PAP?  No thanks.

28 was to 26.

30 was to 27.

This post (31) was to 29!

I understand your argument, but I think my point is still actually somewhat germane (you may, of course, disagree). If the context of PAP is the 1990s “Go at 95% of max” environment, and the thresholds are set based on that environment, applying it “as is” to the 1970s might not be fair.

To use an analogy, say we want to rank the best fielders of all-time and we decide to use a terrible metric to do so (fielding percentage). If we aren’t careful, we might say that all the fielders of the 1910s were awful because their FPs suck compared with those of today. Of course, the game as played today bears little resemblance with how it was played then, so such a comparison is obviously unfair.

All that said, I agree with your premise. PAP is completely arbitrary, doesn’t seem to be based on any actual medical or physiological data, and has essentially zero predictive worth. Carroll’s hyperbole is well deserving of the skepticism you have heaped upon it here.

You can create a threshhold pattern like the following:
Pace AbuseLevel
0.99 10
0.98 20
0.97 30
0.96 40
0.95 50
0.94 60
0.93 70
0.92 80
0.91 90
0.90 100
0.89 110
0.88 120
0.87 130
0.86 140
0.85 150
0.84 160
0.83 170
0.82 180
0.81 190
0.80 200

So, this chart (completely madeup) says that you can pitch at 99% effort for 10 pitches, and then you are taking a higher risk.  You can pitch at 88% for 120 pitches or 90% for 100 pitches, etc.

Obviously, it’s not so straight, and each person is different, but again, this has nothing to do with PAP.

Various abuse levels for various levels of pacing is commonly accepted. This is what lets you play catch for an hour straight without your arm hurting.

How do I know what level of pacing Clemens had in 1986, 1996, and 2006?  And was Maddux matching the same pace?  What about all the other pitchers?  And what about in the 1970s… did some pitchers really not bother pacing themselves (fear for their job maybe)?  There’s tons of questions here.  And I don’t see PAP answering them. This is a PPP question.

Here was Sean Forman’s critique of PAP at the time:
http://www.baseballthinkfactory.org/files/primate_studies/discussion/forman_2001-03-07_0/

Tango/#30:

I don’t think we can look at stats each time through the order without adjusting for batter quality. As an example, using the two years in your post, starters averaged 27 BF in 1977 and 25 in 2007.

Now I don’t know the exact distributions of BF, but let’s say every single starter faced 27 batters each time out in 1977 and 25 each time out in 2007. For the third time through the order, the 2007 pitchers wouldn’t face the 8-9 hitters, who will almost certainly be the worst in the lineup.

The average of 1-9 hitters in 2007 was .268/.336/.423. For 1-7 hitters, it was .277/.346/.443. That makes up about two-thirds of the difference in the second & third time through stats.

The quality of batters faced should go up each time through the order because of starters getting pulled in the middle of the order...the best hitters at the top of the lineup will receive a disproportionately larger chunk of the PA each time through. I’d guess for 2007, over 90% of the fourth-time-through-the-order consists of the first four hitters.

Wouldn’t we want to weight the 1st, 10th, 19th & 28th batters faced equally, the 2nd, 11th, 20th & 29th batters faced equally, etc.?

Anthony/35: there’s no question you want to adjust for the quality of batter (and pitcher).  This is exactly what I do in The Book (1999-2002).  I was simply offering a very quick first glance in a blog post.  If someone wants to repeat my study from The Book using 1977-1980 data, please, do so.  I’d love to see the results.

I’d actually love to see the results of EVERYTHING in The Book replicated for 1965-1968, 1977-1980, and 1989-1992.

If (starting) pitchers of the past threw around the same number of pitches per start as today’s pitchers, and it only seemed like they threw more because they threw more IP per start, then that is the end of the discussion I guess (about why starting pitchers today don’t throw as much as they used to).

Sean’s critique of PAP was fairly interesting.  I have not read the original PAP articles in a while, so to comment on Sean’s article is not completely fair to PAP and its authors, but I’ll give it a shot.

One, as Sean mentioned, I don’t know why they used 100 in the cubic regression formula with no constant, if the curve is flat between 100 and 130 which it appears to be, at least according to the data given by Sean. 

I don’t know much about curve fitting, but I don’t know whether the data is robust enough to use a cubic equation as they did and for us to be confident in that (maybe the “real” relationship is linear and the “curve” is just noise).  Although intuitively it would seem that the damage of high pitch counts might be exponential to the number of pitches thrown.  Of course we need to know how statistically significant ANY reasonable regression line is, don’t we?

Assuming that the formula, at least above 130 pitches, is significant, I disagree with Sean that .2 or .3 runs per 9 (the amount that a pitcher “suffers” for 21 days after a high (>140 I think) pitch count) is not significant (not in a statistical sense).  That is a lot.  If you are doing ANY studies in baseball regarding the effects of just about anything other than a gunshot to the arm, if you try to get effects more than a tenth or two of a run, you are going to be waiting a very long time!  Now, even if the observed effect is a couple of tenths of a run (which, as I said, is a lot), we do need to know the confidence intervals.

Finally, he is nitpicking in his critique of the long-term effects (injuries) of high pitch counts.  Although it would be nice to control for pitches per start in their data, it is not necessary.  Using total career pitches is fine, at least as a starting point.  (If I got interesting results, then I would look at pitchers with high and low pitches per game or some combination of pitches per game and total career pitches, as there probably is a dual effect in terms of injury.)

If indeed they found a strong correlation between PAP and injury, after controlling for career pitches, Sean should have congratulated them on that finding, as it would be important.

As far as the discussion between Kyle and Tango, I am not sure what Tango’s argument is.

Kyle is spot on.  If we apply the current PAP formula to pitchers in an era where higher pitch counts are the norm and we don’t find that it “works” that might be because it is not the right PAP formula.  If pitchers do indeed pace themselves based on the number of pitches they are expected to throw (and we think that that is the case actually), then we probably need to use a PAP formula with a different threshhold if we want to test PAP at all on these different class of starters.

Of course, even if starters who are used to a different average pitch count do NOT pace themsleves any differently (maybe it is just relievers and starters that have different “paces"), we STILL might need to use a different threshhold in a PAP formula because those pitchers might be physically conditioned to pitch with higher pitch counts and not be affected until they go even higher.

I think that is what Kyle is saying.  I am not sure what Tango is disagreeing with.

While the relationship between injury risk and pitch counts interesting, it is also largely moot at this point.  And that’s because we have figured out that, at least in today’s game, the less you use a pitcher, the better he will generally pitch.  Very short relievers (4 outs and under) are more effective than long relievers. Relievers in general have a big advantage over starters.  Starters are more effective the first two times through the order than after that.  Starters pitch better on 5 days rest than 4, and better on 4 days rest than 3. 

The most important changes in the game over the past 20 years have exploited this fundamental truth:  universal use of the 5-man rotation, reduced IP for starters and the death of the complete game, expanded pitching rosters, and shorter appearances for the best relievers.

The starter/reliever gap, as well as the advantage of long rest, was documented in The Book.  If you doubt that less use and shorter appearances matters for relievers as well as starters, consider the following data.  Baseball Crank compiled the IP and ERA for teams’ 2 best relievers in (roughly) 5 year intervals.  I’ve converted ERA to ERA+, to make it easier to compare over time:
Year IP ERA+
1920 217.5 97
1925 249.1 93
1930 233.1 97
1935 253.2 101
1940 196.5 101
1946 212.1 108
1950 219.3 97
1955 201.7 115
1960 200.2 113
1965 190.1 125
1970 178.2 121
1975 186.4 116
1980 200.9 127
1985 182.2 125
1990 163.7 138
1996 148.1 138
2000 146 135
2004 144.9 138
The trends couldn’t be more clear:  teams use their best relievers less, and get a much better performance out of them as a result. 

It may be that the original motive for reducing pitch counts was avoiding injury (though I’ve never seen clear evidence for that).  But even if it now became clear that reducing starter pitch counts did NOTHING to reduce injuries, the game would hardly change at all.  Starters would still get pulled after about 100 pitches, because a decent reliever is simply a better pitcher at that point than all but the very best starters (especially with the reliever often having platoon advantage). Any team that pushed their starters for more innings would quickly find that they won fewer games, and would stop. 

Yes, a few things would change on the margins:  some young studs might be allowed to pitch the 7th more frequently, and veteran aces might pitch the 8th more often.  But starters would still average about 6.0 IP, complete games would still be rare, and teams would still carry 11-12 pitchers.

I DO think it’s possible that the advantages of short and less frequent work are greater today than in the past.  Tango’s data in #30 suggests that may be true, as does the difference between Woolner’s and Tango’s findings on days of rest for starters. Most likely, those differencs are because of the huge importance of strikeouts in today’s game.  The central advantage of relievers over starters, and short relievers over long relievers, is higher K rates (presumably because they can throw harder).  Could pitchers in the 1960s and 1970s gained a comparable advantage with less frequent use and/or shorter outings? Perhaps; hard to say.  But I think it’s clear that a team that tried a 4-man rotation today, or even a rigid 4-day-off rotation, would see a decline in Ks. So too a team that asked its best relievers to routinely get 6 or more outs.  (I realize The Book reached a different conclusion on reliever usage, but I think the selective sampling problem with those relievers asked to go long raise doubts about that finding.) And given today’s HR rates and high BABIP rates, pitchers who stop striking out hitters won’t win a lot of games.

Guy,

There’s a huge selective sampling issue.  Taking, after the fact, a team’s two best relievers (out of 4 or out of 6) is a huge skew.  Imagine there’s a time when you have 10 relievers, and you take the top 2, after the fact.  Of course, they’ll look great.

Take the top 2 relievers in year x, and tell me how they did in year x+1.

The Crank wasn’t 100% clear on his methodology for selecting the two “best” relievers. It appears to be based partially on ERA, but also GF and IP.  I don’t believe he’s including Loogies with 40 IP and a great ERA.  And he used the same approach in all years, so every season has some selective sampling.  Look, I agree it’s not a perfect methodology.  But I also know that Gossage had a career ERA+ of 126, and Rollie Fingers was 119, while Billy Wagner is at 180.  I just don’t think there’s any serious question that the top 8th and 9th inning pitchers of the past 15 years have posted much better ERAs than top relievers of the past.  Do you?

I agree that the top OPS+ and ERA+ has been posted by Mariano, Wagner, Percival, et al recent pitchers.

But the methodology of selecting pitchers is far from “not a perfect”.  It is downright wrong.

Furthermore, Mariano, Wagner, Percival are great pitchers.  You won’t find a plethora of great pitchers subjected to the bullpen for their 10 prime years prior to 1962 or so.  So, any of that old data is completely explained because the gap between the average starter and average reliever is enormous back then.

Whether it’s the 4-batter, 6-batter, or 9-batter reliever that can sustain himself at a high pace, I don’t know.  I think the success of Tom Henke from the outset as a 5-batter reliever was probably very telling, in the mid-80s.

So I agree with you, but I reject the data from Crank.

Tango:  Fair enough.  BTW, Steve Treder has tracked data for “closers,” defined as the pitcher with the most saves on each team.  While that doesn’t control for the possibility that teams are using better pitchers in the closer role, I think it’s a reasonable way to identify the pitcher each team thought was its best reliever (highest LI would be better).  Here are IP/ERA+:
1960-72 94 130
1973-85 97 140
1986-92 77 146
1993-04 67 149
Looking at the top quintile of closers over this period (most saves + wins), IP drops from 110 to 74, and ERA+ rises from 157 to 186. Of course, a 20- to 30-point increase in ERA+, in high leverage usage, is huge:  about .50 to .75 R/G. 

It would be interesting to look at the 2nd and 3rd most trusted relievers, to see how their workload and performance have changed.  The reduction in length of average outing might be even greater.

That list is a bit better, if only because the closers don’t turn over every year.  I would still have preferred he takes the closer of year x and look to see how he did in year x+1.

Anyway, the 1973-1985 time period (I wouldn’t have gone all the way to 1985… I think I would have stopped in 1982) has an ERA+ of 140 with 97 IP.  So, in a 4.00 ERA league, that’s a 2.86 ERA.

The 1993-present is 67 IP and 149 ERA+.  That’s an ERA of 2.68 with 67 IP.

To recap: 97 IP, 31 ER compared to 67 IP, 20 ER.  The difference is 11 ER and 30 IP (3.30).  I don’t know where you will find a replacement like that today to make up the difference.

I don’t doubt today’s pitchers are more effective.  But, I also think they can easily add to their workload and maintain a similar effectiveness.

Given the choice between 97 IP, 2.86 ERA and 67 IP , 2.68 ERA, I think I would much prefer the former.

BTF has a link to this blog, and it has spawned its own thread:
http://www.baseballthinkfactory.org/files/newsstand/discussion/the_book_blog_tango_pap/

One reader there commented how the pitch count estimator doesn’t work because of his “testing”:

20. Best Dressed Chicken in Town Posted: November 05, 2007 at 06:26 PM (#2605898)
Estimated pitch counts for nowadays are too low, I’m fairly certain. When I’ve tested Tango’s estimator against recent games for which we have actual counts, it’s almost always low.

And post 26 has one conclusion:

Quite simply, every thing have ever seen tells me that pitch count estimators:
A: Over estimate how many pitches were thrown in the 60s and 70s;

This was based on presenting the changing rates of BB, K. 

Both of these posts show an ignorance as to what the pitch count estimator does.  The xPCE in fact looks at the rates of BB and K per BFP to change the number of pitches per batter.  You wouldn’t get that impression by post 26 (which was otherwise very thoughtful).

Post 20 however is the worst kind of post, passing a conclusion without evidence, against a theoretical framework that was tested by a few other researchers.

***

What is likely true is that the xPCE still overcredits pitchers of yesteryear (not by much though, say 5 pitches in a start), because the mindset of the hitter has changed.

One day, I’ll improve the implemenation.  But the basic framework still will stand because of its logical underpinnings:
http://www.tangotiger.net/pitchCountEstimator.html

I don’t have the numbers to back it up, but it seems to me there have been a lot more relief aces with longer careers since they started pitching 70 innings instead of 100.

Maybe Bruce Sutter would have pitched to age 40 if he had the workload of Trever Hoffman.

To me, I judge the length of a career by the number of innings or batters faced.  And The Book made the results there quite clear that the pitchers of today have no advantage. 

Including the post-season, Mariano Rivera has pitched less than 1100 innings, virtually identical to Bruce Sutter.  And Sutter is considered someone with a very short career.

I consider cars to give me mileage, not years.  I can keep my car running for 20 years, if I only drive it 3000 miles a year.

I just looked.  Hoffman hasn’t even reached 1000 IP.  He won’t even approach Sutter’s innings next year.

Tango/44:  I agree I’d rather have 97 IP, 2.86 ERA.  The question is whether those 1973-85 numbers could be sustained today (they weren’t from 1960-72), given the need for strikeouts.  Perhaps, but I don’t think we know for sure. 

On Rally’s point:  spreading the performance over more seasons could have more value, if it allowed teams to use a closer in higher-leverage situations.  And it certainly has more value ($$$) to these pitchers!

Bruce Sutter’s LI was 1.9.  Since 2002, Hoffman is at 2.10, Mo is 1.95, Wagner is 1.74 (average of 1.93).  Percival is right in there, too.

So, what we have here are relievers that face just as many batters with just as much leverage.  The difference is that Sutter’s batters weren’t spread out into his late 30s.