THE BOOK--Playing The Percentages In Baseball

I like it.  I had that thought the first time I watch a Strasburg start on TV and listened to Dibble.  I was thinking this was one of the few human beings on the planet who knows what it’s like to throw that hard.

To me Dibble, talking about Strasburg, often sounds like a parent trying to relive his glory days vicariously through his son.

Guys like Zumaya, Broxton, Daniel Bard might be able to stand in as comps for Strasburg’s strikeout ability.  None of them have anywhere close to his control though, or else they probably would be starters.

Just thinking about this in my head, Eric Gagne and Dick Radatz probably qualify, and like with Dibble, just a huge percentage of guys like that hurt their arms quickly after establishing themselves as dominant pitchers.

That’s the unfortunate comp result that is playing out too well for Strasburg.

The only thing against my theory is this:

Dibble and Gagne and the other relievers are probably pitching close to their 100% capacity.

Felix, Josh Johnson, etc, are pitching, say, close to 95% of their capacity.

One would think that Strasburg, being a starter, would think like a starter, pace himself, and pitch at 95% of capacity.

And so, it’s possible that relievers are more susceptible to injury because they take the bigger risk at pitching to capacity. 

So, that’s a strike against my theory.

Then again, it didn’t look like Strasburg was pitching like a (seasoned) starter.  IIRC, Mike Fast’s chart were showing his fastball speed as the game went on, and it was on a downward slope.  Basically, he wasn’t pacing himself like a starter would.  (Don’t quote me on that.)

Anyway, would the “pitching to capacity” essentially derail my thought?  On a per-pitch basis, who gets hurt more often, starters or relievers?

Wouldn’t that be difficult to measure since some (many?) relief pitchers were starters who either couldn’t stay healthy or lacked the durability to start.

Don’t the small sample sizes of relievers present an issue in comping starters to relievers?

Don’t the small sample sizes of relievers present an issue in comping starters to relievers?

"Wouldn’t that be difficult to measure since some (many?) relief pitchers were starters who either couldn’t stay healthy or lacked the durability to start.”

I thought starters generally turned reliever because they had 1 or 2 really good pitches, but not 3 or 4 that is required of a starter that goes through the lineup 3 or 4 times per game?

Also with Dibble, his motion and mindset are completely opposite of what you would want from a starter. You don’t enter a pit bull into a dog race.

Relievers can, if desired or need be, throw the same pitch or 1-2 pitches over and over again, and are out of the game before anyone gets to adjust or time it. I don;t think you could consistently throw fastball-slider over and over again as a starter and not get hit eventually ... or have your arm fall off.

"Don’t the small sample sizes of relievers present an issue in comping starters to relievers?”

Obviously we’d have better knowledge of a reliever’s talent level if he could throw 400 innings a year instead of 60 (and not destroy the arm).  But comparing Strasburg’s rate stats to starters leaves us with very few comps, pretty much Pedro and Randy Johnson.

Comparing him to relievers we have a lot more comparable players to generate an expected forecast range.  Strasburg pitched 68 innings, a pretty typical total for a modern closer.  Sounds like a great idea to me.

Great job by Rally:

http://www.hardballtimes.com/main/blog_article/relievers-compared-to-strasburg/

I hate to give you more work Rally, but when you say this:

If he were healthy enough to pitch 150, then the median projection should stay the same but the extremes would be closer to the mean.

I would think the median would ALSO go up, IF the sample selected is biased.  And, they might be.

In order to test that, all you have to do is grab their next 2-3 seasons, rather than next 1.  That should give you the best comp I think.  My guess is that the median will go up, and the ranges will be tighter (as you say).

In order to test that, all you have to do is grab their next 2-3 seasons, rather than next 1.  That should give you the best comp I think.  My guess is that the median will go up, and the ranges will be tighter (as you say).

***

You would have to remove the effects of aging before you could draw any conclusions.

I doubt that aging would play any role, if the average ages we’re talking about is in the mid-20s.

They were:

29
31
28
27
26
29
27
29
34
30
31
29
28
27
32
26
25
30
34

That’s an average age of 29. Now you want to look at their next three years, so we’re talking about an average age of 31 (or the aging that happens between 29 and 32). Aging most certainly matters.

Good stuff.  I wasn’t thinking they’d be so old.

Thanks David.  It’s hard enough to find comparable rate stats to Strasburg even without controlling for age.  With age, he’s pretty unique.  And none before 1987.  Before that strikeouts were lower across the board, pretty much impossible to find similar K/9.  And before that, ace relief pitchers had higher workloads, even they probably had to pace themselves, or else maybe lost a little effectiveness the 2nd inning of work.

You’re probably right that the mean FIP would be higher for 150 innings - none of these guys could maintain the same effectiveness for 150 innings that they do for 70.  But if we were able to look at 3 years, and aging were not a factor, I doubt the results would be much different.

Looking at year X+1 stats, I don’t think is a biased group.  Year X is obviously a selective sample, since I selected it for rates similar to Strasburg.  But year X+1 should be an unbiased estimate of the true talent of the group, shouldn’t it?

This is an interesting exercise, but I’m not sure what it tells us.  If we want to project Rally’s relievers as starters, we would have to add at least 1.00 R/G to their projections, even leaving age aside.  Yes, 150 innings will reduce volatility, but it’s also a fundamentally different job than short relief.  Only 2 of Rally’s pitchers ever started a non-trivial number of games.  (Did you know that Hoffman and Wagner have zero starts between them?  It’s amazing to me how many of these guys are never even tried out as starters.)

Nathan:  4.60 as starter, 2.17 relief
Gagne:  4.68 starter, 2.62 relief

Please don’t scream “selective sampling” at me—I’m not claiming these as true talent measures for these guys as starters.  On the other hand, what would we project guys like Mantei, Nenn or Wagner to do as starters?  (Honestly, I would add about 2.00 to their ERA.) In any case, we would have to bump up their projected ERA by at least 1.00.  So what does these pitchers’ X+1 performance as relievers really tell us about the plausibility of Strasburg maintaining this same level of performance as a starter? 

And I disagree with Rally’s assertion that Strasburg’s projections are now “moot.” Rate projections and playing time projections can’t be—or at least shouldn’t be—treated as separate for pitchers.  What is the value in a projection which tells me that a starting pitcher will have an ERA of 2.50, if in fact there is virtually no chance he can do that and stay healthy?  One-quarter of Rally’s guys got hurt, but approximately 100% of starters who resemble Strasburg get hurt.  Isn’t that the more meaningful reference point?

At some point I think we will move toward evaluating pitching projections based on WAR, not rate stats, as it is aggregate value that we mainly care about.

You would think it should be, and, if that’s the case, then year x+2 should give you similar results, aging notwithstanding.

That’s why we test… if you find decreased performance in the year x+2 relative to x+1 that is not age-explained, then we likely did have bias.

If we want to project Rally’s relievers as starters…

But that’s NOT what we are trying to do here.  What were doing is looking for pitchers who pitch like Strasburg:  pitchers who pitch all-out, gas-it-all-out, get similar results to a paced-Strasburg.

So, the question on the table is what happens to those pitchers who gas-it for 1 or 2 innings at a time in year T+1 and T+2?  Could that be a good forecast for what you’d get from a paced-Strasburg in year T+1?

"Could that be a good forecast for what you’d get from a paced-Strasburg in year T+1?”

I guess it “could.” But why would it?  If we learned that a player who hits 40 HR in A ball one season will typically hit 38 HR in A ball the next year, would that help us better project a 40-HR rookie in MLB?  I’m not sure it would..

The other question is, how do we know Strasburg is “paced?” Maybe he’s been going all out for 5-6 innings, and we should expect his ERA to rise by .80 or 1.00 if/when he becomes a “real” starter.  How would we know?

I don’t want to make any assumptions about whether Strasburg is pacing himself or going all out.  I don’t know.  Maybe if he had been in relief he would have hit 104 like the Chapman kid.

I see the question is what results you can expect from players who pitch X amount of innings and have Rate stats of W, Y, and Z.  And the empirical evidence shows that they put up results consistent with what you expect by feeding the stats into a projection system.

"And the empirical evidence shows that they put up results consistent with what you expect by feeding the stats into a projection system”

Yes, but the evidence also suggests these pitchers will deliver, what, 55 innings to their teams (you don’t report average IP, so I’m guessing).  So if you want to project Strassburg at those rates and 55 IP, I guess I can live with that.

Here’s the way I would put the question:  if the proposition on the table is that Strasburg will be a 2.50 starting pitcher, was his 2010 season evidence for or against that proposition (stipulating that one season doesn’t prove anything)?  The consensus here seems to be that it’s at least weak evidence in favor.  I would say it’s weak-to-strong evidence that the proposition is false.  The actual result—68 IP followed by TJ surgery—is completely consistent with the historical record, which is that no 22 yr-old starter has ever been close to that good and stayed healthy.

Start with that proposition:  Strasburg is a 2.50 SP.

Add 68 innings with his results (and observation) and I’d be more likely to agree with the proposition than I would have been before the year.

But the TJS is the kicker here.  Now we know for certain that he will not be a 2.50 pitcher, or any kind of pitcher, in 2011.  And the uncertainty of recovery, possible changes in mechanics/pitching style to prevent recurrence, etc. puts a lot of doubt that he will ever be that pitcher.

”...possible changes in mechanics/pitching style to prevent recurrence, etc. puts a lot of doubt that he will ever be that pitcher.”

I don’t think we disagree much, but if Strasburg has to make changes that reduce his rate performance in order to remain a starting pitcher, then I would say it raises doubts that he ever WAS that pitcher.  That is, if his level of performance made it virtually inevitable that he would either get hurt or change his approach (or both), then he never really was a 2.50 starting pitcher.  Or more precisely, we could never project him at that level for 180 or 200 IP, which is what we expect from a starter.