Thanks, SG.

Thanks, SG. That is really interesting. It might be interesting to compare that rate (48%) to the rate of pitchers who haven’t pitched as many innings in their 25-27 age seasons. Maybe look at guys who threw 400-599 or something.

Wow - SG, that’s great.  And bill… has a point: what would be a good standard by which to judge the effect of these two high-inning groups?  Are the drop-off higher, lower, or equal to what we would expect absent the greater pitching load?
Thanks yet again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again again!

SG, great work as always. Can you also look at the number of pitches thrown. I know Sabathia crossed the 3800 mark this year, which is considered to be risky and abusive to pitchers by many.

Also, do you have information on pitcher weights (just to satisfy Don’s curiosity, of course)?

Once those two things are there, I am thinking whether with the data you have one can run a logistic/probit regression to see the impact of these factors on the attrition rate.

Another great post!  I only hope the Yankee’s analysis is as sophisticated as SG’s.

I wonder whether regression to the mean should be a part of this analysis.  The sample group had very good results.  A decline would be expected, since a part of the good results may have been due to some random element.

Of course, the same can be said of Sabathia, so maybe regression to the mean cancels out.

Can you also look at the number of pitches thrown.

I don’t have pitch data for everyone, but I can pull a subset of the pitchers who I do have data for.

Also, do you have information on pitcher weights

I can pull in the listed weights from the Lahman database, but I’m fairly certain they’re not very accurate.  I’m pretty sure C.C. hasn’t seen 250 in about 6 years.

Actually, one thing I’m wondering about is if there is anything into more distinct groupings.  E.g. SG mentioned below-average performers such as Mike Maroth.  Can we find any logical division into say, “elite performers”, who typically pitch a lot of innings w/ great rate #‘s, and “innings eaters”.  “Innings eaters” are the Andy Pettitte’s and Tim Wakefield’s of the world, who consistently throw a lot of innings at around league-average performance.

I guess what I’m looking it is, can we find a good separation, and then do the two groups age differently?  E.g. do “innings eaters” age 25-27 quickly become 5th starters during age 28-34, and “elite performers” stay elite longer?  Would this tell us anything useful in regards to Sabathia (who clearly is an elite starter right now)?

SG,

As usual, terrific work.  Very interesting stuff.  I do have a question regarding the design you’ve chosen for the analysis.

Essentially, you’re using a retrospective cohort design similar to what’s often done in survival analysis.  Not exactly, but roughly. 

That design does give a valid answer for the probability of the pitcher still being in the major leagues after 7 years.  However, I think there are two confounding issues being addressed here:  workload and time.  From a longitudinal cohort design I’m not sure if you can distinguish the effects of one vs. the other.  I think some kind of quasi-control group would be needed to make that determination.

The question is this - if you followed a cohort of pitchers who threw an “average” workload from ages 25-27 for the next 7 years how many would still be pitching past age 34?  Then I think you’d need to compare this with the higher workload group that CC is involved with.  The attrition rates need to be compared to get a sense of how much the additional workload hurts the pitcher.

The kind of retrospective cohort design you’ve used here indicates that it is very risky to sign a pitcher to a 7 year deal.  But I’m not sure if it answers how much more risky is it to sign a pitcher who had thrown a lot of innings from ages 25-27 to a 7 year deal as there is no baseline control group. 

I think it’s that added marginal risk from workload that’s the real key here in terms of making trade offs when it comes to whether or not the yanks should sign CC to a long term deal.

Perhaps I’m asking a different question than what you’re interested in looking at.  Your analysis gives the bottom line answer of what the probability is that CC will still be pitching at the end of the contract.  However, intuitively I think we know that a 7 year deal for a pitcher is very, very risky.  I guess I’m wondering what the additional workload related risk is on top of the baseline risk to signing a pitcher to a seven year deal?

Again - this isn’t a criticism in anyway.  Your analysis is spot on in terms of the analytic method you’ve chosen. Quantifying that 48% probability of CC still pitching at age 34 is very insightful.  Just a question regarding baseline risk and control groups.

Good points all.  I know this is a complex question and there are a lot of different ways to look at the data, so I’ll read through the comments today and then take a look at the same question again using some of the suggestions that can be somewhat easily incorporated.

One observation that really belongs in the earlier thread about whether the Yankees can afford both CC and Tex is that this is an unusually important year for the Yankees to plan to contend.

The most expensive seats in the new stadium (in suites, and premium sections of the infield) are being sold based on a 5 year commitment.  If the team approaches this off season from the perspective of shrinking payroll, they might not be able to get commitments for all of those seats (and therefore be forced to reduce prices or sell them one year at a time).

I don’t have the details on the marginal revenue which could be at risk by having this year’s team appear to not be a contender during the off season, but I would expect it to be far larger than for a normal year.  Because of that, I would expect the team to err on the side of overspending this year, even if it means looking for ways to save money in 2010 (like, for example, pocketing the $13 million when Matsui’s contract ends rather than replacing him with a free agent DH).

Some analysts (e.g. in the book Diamond Dollars) have tried to quantify the marginal revenue which is associated with additional wins by team (since it varies greatly both by market and by how many wins the team is already expected to have).  Unfortunately, I haven’t seen anything specific to the Yankees in 2009 (including the impact of the new stadium and the associated long term deals).  Have any of you heard something along those lines?

Thanks SG.  It’s a very tough chore and no one right answer.  It’s more related to the inherent limitations when working with large data sets in retrospect than anything else.

Even setting up a quasi-control group as I suggested would be challenge.  How do you select the pitcher to include if you’re going back all the way to 1954, especially with the shift from 4 to 5 man rotations, etc. I suppose one could define a minimum number of innings thrown between ages 25-27 and maximum (less than the criteria for the group CC belongs in) and use that as the baseline control.

Just a hunch but I’d guess that even with the pitchers who threw some kind of “average” workload from ages 25-27 the attrition rate by age 34 will be very high.  Perhaps not 52% but still high.  If it turns out that 40% of those pitcher are out of the league it would change CC’s value proposition.  Then his workload (which is the main variable people are worried with him vs. say Santana last year) could be quantified as a 12% marginal risk due to workload.

Either way - your analysis is the best take I’ve seen anyone do on the risks associated with signing CC.  Much appreciated.

However, intuitively I think we know that a 7 year deal for a pitcher is very, very risky.

Lets put it another way:

Whats the longest contract for a pitcher that has turned out well for the signing team?  Are the Pedro Martinez, Barry Zito and Mike Hampton signings typical?

Telnar, Vince Gennaro wrote about the Yankees’ marginal win value during the Santana talks last offseason (link).  He estimated around $6.25M for a marginal win I think.  Not sure if we should drop that some based on the economic downturn or not.  Most analysts are saying a marginal win should be worth between $5M and $5.5M in 2009 for most teams, so $6.25M’s in the ballpark for the Yankees I guess.

I just found this prospective look at the impact of the new stadium on the Yankees marginal revenue per win written in 2007 (to analyze the Arod contract) http://sports.yahoo.com/mlb/news?slug=ys-gennaroarod112707&prov=yhoo&type=lgns

It suggests that the MR per win will be down sharply in 2009, but that was without considering the 5 year license effect.

Whats the longest contract for a pitcher that has turned out well for the signing team?

Seems like the Moose signing has to be considered up there(assuming long > 4 years).  Randy Johnson’s first contract with the D’Backs has to be the best free agent pitching signing ever, but it was only four years.

“Whats the longest contract for a pitcher that has turned out well for the signing team?”

If you’re just referring to contract length rather than specifically free agent signings, then I’d say that the 6 yr $75M extension Pedro signed with Boston in 1998 was amongst the best deals ever. He threw 185 innings or more in 5 of the 6 year and pitched at arguably the highest level any starting pitcher has in the history of the game over an extended period of time.

Greg Maddux?

[17] was in reference to [15], by the way: the issue of best FA signings, not contract length.

Yeah, you may be right Frog.  I’d have to look at the terms of both the Maddux and RJ signings in terms of cost/value, but Maddux may have been a better one.

Unless I missed something, the OOL category has a huge contamination from (aged-to-) RL pitchers getting dropped.  I’d want to see something like the group of pitchers starting at OPS+ - delta_OPS+_expected-aging - buffer > RL(OPS+, innings) at t=0.  And of course medical/training advances make me skeptical about such a deep dataset.

Unless I missed something, the OOL category has a huge contamination from (aged-to-) RL pitchers getting dropped.

I don’t think so.  I don’t think a single pitcher in the initial group of 600 innings pitched over ages 25-27 was a replacement level pitcher, or they wouldn’t have been allowed to pitch 600 innings over three seasons.

Now if they eventually moved towards RL, they should still probably be included in the OOL bucket, since their skills eroded to the point where they were no longer viable major league pitchers, which is no different to the bottom line than if they got hurt and were unable to pitch at all.

And of course medical/training advances make me skeptical about such a deep dataset.

Yeah, there’s a fine line here.  I didn’t necessarily want to include everyone I included, but getting an adequate sample size necessitated pulling from a larger group.  I can look at something like the last 10 years and see if that tells us anything differently.

Hes a risk. If we sign him we should look to throw out a 6 man rotation at times next season. Keep Hughes Joba and CC’s innings low

Andy Pettitte threw 740 innings, including postseason, from age 24-26.

from age 25-27, he threw 697 including postseason.

i don’t know if that is appropos of anything, but i don’t think there would have been much hand-wringing over his inning totals were Pettitte a FA before his age 27 or 28 seasons.

maybe we’re just smarter now, i don’t know.

Considering this year we had about 3 starters at all times I’m not really worried about throwing out this mythical and costly 6 man rotation.  I’ll be pleasantly surprised if we can keep 4 of your 6 around at the same time for a majority of the year.

“I don’t think a single pitcher in the initial group of 600 innings pitched over ages 25-27 was a replacement level pitcher, or they wouldn’t have been allowed to pitch 600 innings over three seasons.”

The claim is that some are somewhat above RL (maybe even league-average) innings-eaters; as they age and lose OPS+ and innings, they get dropped.

But aging to RL isn’t a risk for CC, so the dataset may be contaminated.

If you look at the previous-year innings and OPS+/OPS+_rl of pitchers who fall out, you might see the above.

“i don’t think there would have been much hand-wringing over his inning totals were Pettitte a FA before his age 27 or 28 seasons.”

For one thing, it wasn’t a question of CC-standard money.

Wow, great post SG, as always.

Can we find any logical division into say, “elite performers”, who typically pitch a lot of innings w/ great rate #’s,

So, the name that jumped out at me right away was Sandy Koufax. From age 25-27, Koufax had a huge amount of innings and probably a sub 2.5 era. He remained dominant, but he was gone at 30. Now, he’s only 1 of 178 pitchers who are included in the spreadsheet, but it makes you wonder, without going through the pitchers one by one, if it makes sense to break this group down into smaller groups. 

You really could go nuts with this if you wanted. Maybe removing pitchers who were injured before they were 34 in non-pitching related injuries (or maybe just non-shoulder/elbow related).  Maybe make a group of Sabathia’s top 20 comparables through 25-27 and see how those numbers shake out. I forget when they started to go to the unbalanced schedule, but that will obviously factor in as Sabathia would be pitching to more Rays, Red Sox and Orioles and less Royals and White Sox.

Also, seems like there is quite a jump going from 33 to 34. I’d hate to see what age 35 looks like.

For one thing, it wasn’t a question of CC-standard money.

good point.

“But aging to RL isn’t a risk for CC, so the dataset may be contaminated.”

Am I missing something, or does that not depend upon the length of contract in question?

The shift from a 4 man to 5 man rotation also seems to play a role in this. Also, what about the strike seasons?

Maybe the IP number should be a percentage of the league leaders IP total?

“Am I missing something, or does that not depend upon the length of contract in question?”

Will CC age to RL before he’s in his mid-40s?

IP+.

There ya go.

“Will CC age to RL before he’s in his mid-40s?”

Isn’t that assuming the result of the study?  Aren’t we asking whether very high-IP seasons in the given age zone affect performance (rather than simply leading to a binary injured-and-out or uninjured-and-performing result)?  And wouldn’t it be interesting to know whether or not that does make a pitcher age to or towards RL faster?  What if he aged to ML avg., rather than RL, within a few years?  Seems relevant to me…

The study as designed assumes no pitcher ages to RL.  But, ok, my claim is that no sensible aging model (assuming a clear distinction between aging and injury) will take CC from his current WAR to 0 during his contract; and the above’s 0.3/year is consistent with that.

Now, he’s only 1 of 178 pitchers who are included in the spreadsheet, but it makes you wonder, without going through the pitchers one by one, if it makes sense to break this group down into smaller groups.

Yeah, that’s why I’m asking if it does, and if you would still have meaningful groups.  To pull a guy out, look at Roger Clemens.  He threw a TON of innings by age 27.  Between 28-34, he had some extremely high inning years (271 at age 28, 264 at age 34), and some very low inning years (140 at age 32).  Ditto with ERA+ (160 or better 4 times, including 221 in 1997, but also a low of 104 in 1993).  All of this is supposedly before steriods as well.

So…would CC be more like Clemens, or more of a smooth downward trend like the group average?  If more like Clemens, would we be perfectly happy with something resembling (understand IP totals lower to reflect the way pitchers are treated now) his age 28-34 seasons for a ton of $$?  I guess the big thing is if you look at a pitcher like Mo, we KNOW we have it hard to find a group of pitchers similar enough to him to have a significant sample-size.  Would CC be more like the all-time greats, or would he fit normal (understanding normal is difficult with an already small sample) distribution?  Or should I get more sleep at night?

If we sign him we should look to throw out a 6 man rotation at times next season. Keep Hughes Joba and CC’s innings low

Then why bother signing CC? Yeah he’s a very good pitcher but a lot of his value comes in his ability to eat innings like tasty, tasty chicken wings.

Might similarity scores be a quick and dirty way to approach this discussion?

For instance, Baseball-Reference tells us that the ten most similar pitchers to CC Sabathia through age 27 are:

  1. Dave McNally (939)
  2. Greg Maddux (934)
  3. Ken Holtzman (931)
  4. Denny McLain (931)
  5. Dennis Eckersley (931) *
  6. Alex Fernandez (917)
  7. Lefty Gomez (917) *
  8. Milt Pappas (916)
  9. Carlos Zambrano (914)
  10. Steve Carlton (914) *

Of those ten, you’d be thrilled to have signed Carlton or Maddux to a seven-year deal.

Pappas would be okay.  He managed to last in the big leagues for seven years before retiring, and was effective by his own standards for most of that time.

McNally and Gomez had four good years in them before their retirement.

Holtzman managed three good years before his career went south.

Eckersley had two or three decent seasons as a starter before transitioning to the bullpen.

Fernandez had less than a season of effective pitching ahead of him.

McClain had nothing left and was out of baseball after his age 28 season.

If more like Clemens, would we be perfectly happy with something resembling (understand IP totals lower to reflect the way pitchers are treated now) his age 28-34 seasons for a ton of $$?

i would be ecstatic.

Part of the difficulty with having a smaller sample is that we are already dealing with a bunch of outliers to begin with. The reason why it is hard to find a comp for Sabathia is that he is an outlier in terms of his place among pitchers. It is extremely difficult to project such individuals within acceptable degrees of statistical error.

That is why doing it in a regression framework (survival analysis, or a probit/logit binary outcome model) makes more sense, as you can introduce controls to account for a bunch of factors (including time period indicators, which take differing eras and medical improvements into account), and can also get reasonable sense of the extent of error of such projections. 178 individuals spread over quite a few seasons is enough data to conduct such an analysis. If you go to the “control group” mode, the statistical accuracy quickly vanishes.

This data is interesting but also less useful out of context. What is the average drop rate (OOL) for all pitchers in the same time frame? I think that would make the 48% remainder a more valuable number than it is on it’s own.

Also, how would it break down if there were a cutoff based on production? Pitchers with, say, 600 combined innings and an ERA + over 148 (Sabathia’s over the last 3 years)? This would seem more relevant as we’re trying to evaluate how a talent such as sabathia will age, and would do an even better job of creating a comparable playing field (eliminating the Mike Maroth types).

I know this would take time and also leave a small sample size, but it would still be interesting data.

Via RAB, Giambi not typed?  Dunno if that affects people’s thoughts re 1b at all.

from that same link:

If 62 or fewer such players file (which is possible) then each team will have a quota of three Type A or B signings.

is this true?  i didn’t know it was 3 type A OR type B.  i thought the limit was just on Type A’s.

Giambi missing Type B probably doesn’t change the Yankees thinking at all.  i think their decision to let him walk or not would not hinge on getting a sandwich pick back.

Thanks for the link to the $6.25 million marginal revenue/win number, SG.  If we assume that it’s accurate for the average of a typical long term contract (aside from inflation), then the fact that we still have to reduce it by the luxury tax, justifying only $4.46 million in (2009 dollars) contract spending per win (admittedly, that ignores marque value, which both CC and Tex would have).  That might not be quite enough marginal revenue to justify offering what it will take to sign then, although it should at least be close.

Go fillys?

Goodnight Mrs. Parker

Chubby catchers make you say uh oh.

I kind of want to see Lidge blow the save…wouldn’t mind if the phils ended up winning in the end though.

Come on Phils, let’s get this offseason started.

Didn’t Hinske hit a pinch hit homer?

“McClain had nothing left and was out of baseball after his age 28 season. “

God damn the election, I didn’t see the “L.” Last night during a commercial break there were 6 ads. All 6 were political.

Only one more week till it’s all over…

I thought you guys might want to know that for Halloween I am considering going as the scariest thing I can think of: Comma Guy.

Thank you!

Did we sign Sabathia yet?

I know, I have a funny sense of timing.

Congratulations Phillies, Phillies phans, and my Pop.

Wow, I didn’t realize that Jesus was on the Phillies coaching staff.

Holy Moley. No one walked off a Lidge.

And congrats, Old Phillies Fan.

My Da just called me from the ole folks center in Toms River. Twice.
Crazy old mick cant tell a double play from doubleheader.

You too You son of an OPF.

The Rays should have dressed up like Albert Pujols. That would have scared the beejezus out of Lidge..

Are the phans booing Selig?  That’s hysterical.

May have scared a darker, fouler substance from him J.

did he just say something about putting pieces of pie in a bucket

I think that’s what OTF was saying.

this is fucking boring.

This whole business of “the drought is over” is a little overblown considering the Red Sox and White Sox victories of a couple years ago.

Selig is the guy in the 9 year old Buick that always seems to be in front of me when I’m running a wee bit behind.
F you Sonny! I’m doing 25 inna 40 zone!

http://tinyurl.com/http-teehee-com

SG- I tried to load the spreadsheet into my venerable PC.
William Gates called and said ante up lad.
Thurm the Katt, blittzed on philly sardinesteaks told him to go watch the Bucky Boone game on low def.
More later.
Yay Phils.
Eat tonight Met fans.

Anyone else think Maddon blew it by not starting the 7th with Price?

At any rate, ‘twas nice not really caring about the outcome in this series.  Emotional detachment oddly makes games more enjoyable to watch.  Congrats, Phils!

Now to text my Met-loving friends…mwahahahahahaha!

Transcript of chat with long-time, long-suffering, morbidly-enjoying-the-long-suffering Phillies fan friend, who consistently insists that they will somehow, some way, lose:

[9:58:09 pm] WP says: You’re running out of outs if you really intend to lose, Al.
[9:59:29 pm] WP says: Press tomorrow: What a spunky, inspiring loss for the Divils!  Oh, they’ll surely be back every year for years to come!  This WS was an emphatic statement by the Rays!  (Was there another team involved, by the way?)
[10:00:15 pm] WP says: Ah, here you go.  The kind of hits usually given up by Rivera, when he gives up hits at all.
[10:02:53 pm] WP says: I presume you’d expect the Phillies to get within one out, even within one strike of winning it all before the run scores…
[10:03:19 pm] WP says: And there it is, the second out…
[10:05:04 pm] WP says: And Hinske, a typical WS hero - unheralded, not even on the team when the WS started, already one PH home run in the only pitch he’s seen in the WS… WS MVP?
[10:05:15 pm] WP says: OK, strike one.  Let’s get to strike two so things can happen.
[10:05:46 pm] WP says: All right, now we’re within ONE STRIKE of victory.  Home run?  Or do we go to 3-2 first?
[10:06:06 pm] WP says: All right - explain THAT!
[10:14:33 pm] Long-Suffering Phillie Fan says: Yes.  Pass the red pill.  Alice is going bye-bye.  (Ref:  The Matrix)

Great job SG.

I don’t want to spoil this, but how many of these guys have pitched in the last 20 years, when it cannot be taken for granted that any longevity/workload enhancements were the result of PEDs?

I see Clemens’s name in the comments… does nobody remember that MLB tests for drugs now?