THE BOOK--Playing The Percentages In Baseball

DSG reviews Diamond Dollars:

http://www.hardballtimes.com/main/printarticle/diamond-dollars-reviewed/

I wonder which cover Ned Yost has…

http://www.brewcrewball.com/story/2007/3/7/22311/75137

I got the feeling Ned didn’t like it!

***

For those who have never used Google Checkout, you can get The Book for $7.37 (including shipping, it’s 17.37, minus 10$ instant savings from Google):
http://www.buy.com/retail/product.asp?sku=203528794&adid=17662

Just “add to basket”, and then on the next screen click GOOGLE CHECKOUT

Not sure if Buy.com offers free shipping over a certain amount, but you can’t beat that price.

Reds and blues reviews BP:

http://www.redsandblues.com/?p=241

***

Every year, I’m in the same boat.  I want to like the BP annual, and I hope for some great stuff.  And, I’m always left disappointed.  The last great BP I read was the 2002 or 2003 edition.... the one with Phelps on the cover.... checking Amazon now… ok, 2003.  That was also my first BP I think.  Alot of excellent full-length articles in there… Clay I think had a great one on the draft, Woolner had a great one on replacement level, and Wolverton had a great one on peak value.  (I’m hoping I’m remembering right.)

Coincidentally or not, these are all takeoffs of Bill James works.  James has a fantastic piece on the college and HS drafts in… I think the yellow Abstract.  James introduced replacement level when discussing Roger Clemens and Don Mattingly MVP bid (and Rice and Guidry).... I guess that makes it the 1987 annual, which must have been the yellow one.  And in the Politics of Glory, James had a lengthy study on peak pitchers and long-careered pitchers.

I guess in 2003, BP spoke my language.  Since then?  No so much. 

There was Pappas’ marginal wins/marginal dollar, work that I had already done, so I enjoyed it, but it was lacking.  Woolner did his Win Expectancy and Leverage, which again, I’ve worked on, and he did a good job overall (though his Leverage should be shelved immediately).  Click’s baserunning work was pretty good.  And, I’m tempted to buy the latest BP annual, just to see Keith’s latest work on PAP.

Can’t remember the others right now, which is really my problem.  I can remember all of James’ pieces, no matter how obscure.  Kinda like knowing the words to all the Beatles songs.  Prince, Oasis, Rage… they all succeed and fill the void.  But not for everyone of course.  They don’t necessarily speak my language. 

That the core of BP itself has changed, though, is what’s bothering me.  The Beatles are the Beatles.  Prince is Prince, and Bill James is Bill James.  While one may turn to drugs, change his name, or become corporate, at their core, they are who they are.

To me, BP does not have that same core.  Changing your periphery means you are adapting and learning.  Changing your core means that you’re moving on to something else.  That’s not necessarily a bad thing.  I just don’t have to like it.

My core has not changed.  I like where I am, and that means I can’t move with them, though they keep me around like a lovesick puppy dog with guys like Dan Fox and Nate Silver. 

BP has an obvious business plan.  I’m not even sure Bill James even knows the words.

Keith Woolner had an interesting chat session:

http://www.baseballprospectus.com/chat/chat.php?chatId=269

Talking about 1st/2nd half splits:

While I haven’t performed a rigorous analysis of 1st/2nd half split statistics, I do think there are times where they can be useful. But there has to be an underlying reason *why* you’d want to look at those splits first.

Agreed.  Most of the time, people look at splits, and try to figure out something out of them.  But, you’d be better off having a prior reason for looking at the splits in the first place.

As an aside, there is a small carryover effect from 2nd half performance, but nothing to get excited over.

He and his wife are involved in an MS fund drive:

I am active in raising money for multiple sclerosis, a disease my wife has dealt with for a decade. http://www.dessert-first.org

BL (Bozeman): Your five favorite non-BP baseball books?
Keith Woolner: Hidden Game of Baseball
The Diamond Appraised
Baseball and Billions
My Turn At Bat
The Fix Is In

Loved the first and third books.  The second one was pretty good.  Don’t know about the others.  Surprised not to see any Bill James books in there.

His take on Sarge, Jr:

I think that given the amount of money poured into Matthews contract, and the cloud that now hangs over him, that unless the air is cleared quickly, he will be suspended, and that being suspended will hurt his performance (now there’s going out on a limb).

I’m rather surprised that he thinks a player can get suspended because of “clouds” and “air”.  I think Seligula won’t act unless there’s hard evidence.

On the best CF:

It’d be hard to argue with Carlos Beltran, Grady Sizemore, and Vernon Wells as the top three.

I’ll argue it rather easily: Andruw Jones, Ichiro, Mike Cameron.  Except for Beltran, you can argue any of the other 5.

On the Quality Start:

Why doesn’t 8+ innings of 4 runs allowed count as a quality start?

Keith Woolner: Because that’s the way the guy who invented it decided to define it. No real reason other than that.

As I’ve said in the past, I agree with Bill James assessment that the issue is really the name “Quality”.  If they called the “Maddux game”, no one would care how it was defined.  That said, making a definition like “ERA 4 or under, with at least 6 IP” would have been better.

Excellent illustration on regression of data, and increase in talent level:

Andy Phillips ((New York)): Josh Phelps? Meh… “Everybody” is saying that Joe Mauer will regress, yet he’s only 23. Shouldn’t he still be improving?

Keith Woolner: ...In other words, he’ll go from a .347 BAVG (.310 ability + .037 luck) to a .325 BAVG (.325 ability + 0 luck)

On PBP for fielding:

TomTerrific (chicago): Why doesn’t Prospetus go the pbp for analyzing defense for the years that information is available while using the Davenport numbers for cross era comparisons?

Keith Woolner: The lack of PBP information on BP to date has much more to do with the difficulties of securing the appropriate affordable commerical licenses to use the data as we’d like to than lack of faith in the data itself. And it’s something that we are (and in particular, I am) working on, and have been for some time.

It’s a very political statement: verbose, yet vague.  I’m pretty sure if BP gives BIS or STATS $30,000, BP will have what they need.

And something that I’ve been talking about:

Greg Tamer (Go Boilers!): Does it bother you to see what BPro has become? It’s less about statistics and analysis and more like what you railed against years ago. It’s rumors, buzz, prospects, and bloviating. Defend the change in direction and convince me that BPro’s wand isn’t broken.

Keith Woolner: Does it bother me? Not at all. I disagree with the premise. ... But we haven’t abandoned the statistical and sabermetric front either. BP07 has my study on historical pitch counts and the effects on starters. We’ve covered the economic and business side of the game in depth. Dan Fox has done numerous studies on BP.com, including a fine series on baserunning last year. Clay Davenport has run translations on Cuban leagues, and the WBC. We’ll be introducing new customizable statistics reports that will have BP stats like EqA, VORP, SNLVAR, and WXRL split by lineup slot, position, home/away, by team, by league, and vs. opponents. This is a continuation of the BP tradition of stat and analysis, augmented by additional content that attracts readers and brings them into the BP family. And so far, it’s working well.

I’m usually with Keith on most things, but I’m on the opposite side here.  Maybe I’m in the minority of customers who feels this way, which is why he says “it’s working well”.

If you follow the link in post #4, and read the reviews to that link, you will see alot of unhappy BP customers.  There’s obviously a disconnect between what BP says their customers are seeing, and what the customers who post on blogs know they are seeing.

I’ll be getting my copy of BP this week, so we’ll see how good it is.  In the meantime, I noticed this funny review on http://www.amazon.com/gp/discussionboard/discussion.html/ref=cm_rdp_st_rd/002-7318564-4281653?ie=UTF8&ASIN=0452288258&store=yourstore&cdThread=TxDRNT2YIBK48S&reviewID=RG0HV0IOQ5PD6&displayType=ReviewDetail where the reviewer says:

Billy Beane, the world-renowned author of Moneyball, is back at it again. This time he has come up with a magical new invention known as the BABIP.

Just so that you know the reviewer was having fun, he ends his review with high praise and 5 stars.  And yet, there are two people who posted on Amazon to correct the reviewer that the author of Moneyball was not Billy Beane.  Do some people have no sense of humor?

As Lewis himself said that it’s an author’s dream that everyone who hates his book don’t even know that he in fact wrote it!

I purchased The Book a couple months ago, and it’s fantastic. I was pleasantly surprised at how much statistical information was in there without making it read like a math book.

Anyway, I’m in the same boat with BP, I found them a couple of years ago (i’m only 18), and I feel that their work has declined since then. The review in post #7 was funny, though.

Josh, thanks for the kind remark. 

The “readability” of The Book was critical for us.  There were a few things that we did to assure that we met that goal:

1 - Put a human face on everything we said.  For example, the nine ace relievers (Hoffman, Rivera, Wagner, Percival, et al) brings out the point that we are talking about 9 specific individuals, and not just some blanket “ace”.  Once that was established, we can jump off that to talk about 31 aces.  The same applies to when we trot out those lists, so that you can associate the topic to real players.

2 - Explain and elaborate so that virtually everything we did was reproducible (for anyone crazy enough to want to spend the time).  There was very little “trust us”, so we force the reader to come along for the ride.  It’s the journey, not the destination.

3 - The ride had to be at the same level, meaning that we learn when you learn, rather than we teach/assert and you learn/accept.  Other than the intro and appendix, it was a discovery process in every chapter.  It’s ok for the reader to conclude you are an expert, but not ok for the reader to be “told” that you are an expert. 

The above three were all Bill James-inspired.

4 - Create “The Book says...” Box every few pages, so that a reader that did get lost or bogged down would be able to jump over any part, and just read the Box.

This was inspired by Earl Weaver’s “laws” in his book.

5 - Use wOBA (or OBP).  All the statistical significance tests are alot easier when we just stick to the binomial.  Creating a metric like wOBA allowed us to do that.

6 - Stick to the strategy theme, making it a book, rather than a collection of essays, even if, in fact, they were a collection of essays.

Other than these basic rules, each author worked independently.  It was a pleasant surprise that the chapters flowed as well as it did.  MGL did the streaks, bunt, and game theory.  Andy did the clutch, walk, platoon, appendix, and last couple of pages of the toolshed.  I did the matchups, batting order, pitchers, stealing, and the toolshed.

For each chapter, the two non-authors would proofread and peer review.  This was the worst part of the whole process, and we spent months on this, and likely the reason I never want to do this again.  I much rather prefer to write on my site, and move on, rather than go over excruciating detail on every thing.  But, now that the final product is there, it’ll be our legacy, and will have its own life, whatever that will be.

Remembered another one.

7 - MGL and I (Andy was late to the party) had brainstorming sessions that were very productive.  The first thing we did was create a Yahoo group, and we just spit out all the possible topics.  All those topics were grouped and itemized.  Then, we took each topic, one-at-a-time, and we spit out 10 to 20 questions on that topic.  “When do relievers tire?  Do runners distract batters?”, etc, etc.  Then, we assigned out the topics, and the author picks-and-choose the questions he wants to handle.  Nothing worse than being told what to write, so this allowed the author to be inspired by whatever.

Which reminds me… since it’s unlikely that we’ll write a “part 2” to the book, maybe we should release the brainstorm document to the public, and let others pick up the torch.

Book for Padres fans:
http://www.lulu.com/content/716799

***

Ok, finally got around to reading Woolner’s PAP article in the book.  A couple of good nuggets, a couple of holes, one rather gaping I think.  I’ll post my thoughts later today.  It definitely required peer review, especially because it’s in book form.

If you guys want to comment first, go ahead.

Intro
Keith Woolner wrote an article in BP 2007, focusing on pitch counts and PAP.

The first thing he did was create a pitch count estimator.  In one of the STATS Scoreboard, they showed the number of pitches per event.  It was that work that led me to create the basic pitch count estimator:
3.3 PA + 1.5 SO + 2.2 BB, where PA = 3 IP + H + BB.

It’s rather crude, and linear, and it works fairly well:
http://www.tangotiger.net/pitchCounts.html

Keith included more events, and also an intercept.  This means that even without facing a batter, the pitcher starts off with 10 pitches.  I don’t like it, but I agree that at the level he’s looking at, it doesn’t matter.  Still.

Changing levels
Anyway, he also notices that the number of pitcher per PA has been increasing over time.  He runs regressions against the rates of events (more K and more BB means more pitches obviously), but he couldn’t find enough to explain it.  But, he sees the sloping trend over time, and decides to simply add a time parameter.  That I think is a silly thing to do.  What does “year” have to do with the number of pitches you see?

Certainly I can believe that players may be more patient today and they take more pitches.  But, again, why is there some linear relationship between 20 years ago and today?  Are we going to see this trend continue for the next 20 years?  What if it falls?

In any case, if players are more patient, then we’d see that in their stats: K and BB, a fact Woolner acknowledges.  So, we have a situation where pitcher per PA increases, and K and BB also increase, and we know we get more pitches per K and per BB… why would we need more?

Estimator
If you go to my pitch count estimator http://www.tangotiger.net/pitchCountEstimator.html and just page down to the graph, you see an interesting chart: the more BIP per PA, the fewer pitches thrown.  It’s a given.  If 99% of your PA ended with the ball in play, chances are, you did not go deep in the count.  If only 60% of your PA ended with the ball in play, then we know you had alot of K and BB, and those times that you didn’t, you probably were deep in the count too. 

So, the better process would have been to take that model, and expand and refine it with empirical data.  My model still makes it too linear.  And as Keith shows, the rising levels are more than even what I’d show.  But, the answer lies in creating a markov model of balls and strikes, and not on regressing against time.

In short, transport all the players of today to 1968, with the mindset of 1968, and the parks of 1968, and the changes in pitches per PA would be completely inferred by their new stat line, without an extra time parameter.

Sampling
The next problem that Keith encounters is selective sampling.  He uses the whole year of pitch data for a pitcher to put him in quartiles of pitch counts, and then uses those groups to see how they perform.  What he should have done is looked at grouping the pitchers based on only the first two or three months, and then seeing how they perform afterwards.  Or, selecting them based on the prior year. 

When he went to the historical data, he redeemed himself to me.  He focused on pitchers who threw at least one 150+ pitch game, and see how they performed following those outings.  And, it seems to me, that I didn’t see any real change in performance.  (The scale of the graph was not well done, as it captured some very extreme points on some very limited data.) Certainly, if there was a change, it would be minute.  He also broke it down by various rates (BABIP, K, etc).  So, that was a great part of the study. 

Where’s the rest?
Where he left me wanting more was at the end.  We all know how pitchers in the 70s pitched way more than before or after. 
http://www.tangotiger.net/top4gs.html

So, Keith introduces us to the average “Stress” level for each season.  A clear and huge difference between the 70s and today.  But, pitchers in the 1970s, I presume, were no more lost to injury than pitchers of today.  This was the frustrating part.  This should have been the next step. 

In The Book, I showed how relievers in the 1970s threw more pitches than relievers today, that they were up to it, that pitchers can handle it.  There wasn’t a breakdown of relievers, even though they were pitching 100+ innings back then.  With Keith’s work above, and the idea behind PAP, it would have been the logical step to take: Does PAP really work?

PAP “works” based on the data in the sample.  But, PAP was developed against 1990s/2000s data.  But, pitching pattern of the 1990s is far different than the 1970s.  Now that Keith had his model for pitch counts, why not extend PAP back to the 1970s?  With those huge “stress” level he reported, I would have expected a huge breakdown in pitcher performance or pitchers themselves.  The latter didn’t happen, and I doubt the former did either. 

What I’m left to believe is that PAP doesn’t work.  But, I would have hoped to see a study on it.

A friendly reminder: if you buy ANYTHING at Amazon, go to Amazon through the link at the top of this page.  We actually get referral fees on anything bought at Amazon, if you go through our link.  The amount of the monthly fees is pretty tiny… enough to cover the monthly cost of this site.  But, it’s a nice thing for us to not worry about.

Look at the list of books I have at the top of this blog entry (made Feb, 2007), and now look at the list of recent books at the bottom half of this page:

http://www.baseballprospectus.com/book/?list=all