THE BOOK--Playing The Percentages In Baseball

I guess this is semi-related to a book review.  In case you hadn’t seen it, Bob Geren seems to be a disciple of “The Book”:

http://sports.espn.go.com/mlb/preview09/columns/story?columnist=kurkjian_tim&id=4018271

Ken, very cool!

Geren is amazingly organized. He has a board in his office on which he has three weeks of games planned out, and what pitchers the A’s are going to face. He is a real student of the game, and big believer in the statistical end of the game, some of which he gets from a book called “The Book.” He knows, from reading, that over the course of the season, a No. 3 hitter is going to get 15 more plate appearances than the No. 4 hitter.

“If you have a strikeout guy, sometimes it’s best to hit him second, not fifth,” Geren said. “There are more productive outs at the fifth spot. And did you know that the No. 3 spot in the order comes to the plate with no runners on base more than any spot in the batting order?”

I presume he was misquoted, since obviously the leadoff hitter wins this one.

Here’s the relevant chart from The Book:

Table 46. Number Of PA, By Batting Order And Men On Base

Batting Order
PA empty
PA men on
% with men on
Number of Runners On

1 3.11 1.72 36% 2.39
2 2.63 2.09 44% 2.77
3 2.38 2.23 48% 3.00
4 2.19 2.31 51% 3.20
5 2.28 2.11 48% 3.10
6 2.29 1.97 46% 2.84
7 2.20 1.94 47% 2.74
8 2.17 1.85 46% 2.61
9 2.13 1.77 45% 2.48

Formatting is tough here.  The five column headers are listed just above the five columns of data.  The first and last columns are the key pieces of data.

The surprising one for me when I first did this was that the number of runners on base per game was higher for the #5 hitter than the #3 hitter, even though the #3 hitter gets 0.22 more PA per game.

There’s alot of dependency of course.  The above was based on 1999-2002 data.  Here is that same table, but based on Markov and that every hitter in each slot was identical:

Table 49. Number Of PA, By Batting Order And Men On Base
(Theoretical)

Batting Order
PA empty
PA men on
% with men on
Number of Runners On

1 3.01 1.80 37% 2.55
2 2.64 2.06 44% 2.78
3 2.42 2.17 47% 2.96
4 2.27 2.21 49% 3.06
5 2.34 2.03 47% 2.93
6 2.29 1.97 46% 2.79
7 2.19 1.96 47% 2.74
8 2.13 1.91 47% 2.70
9 2.08 1.85 47% 2.62

As you can see, there is a change in the 3-5 spots, all based on the talent level preceding them.

Fascinating sh!t that amounts to a couple of runs here or there.

Ideally, you create such charts for the actual players you are considering.

By the way, look at this data:

3 2.38 2.23 48% 3.00
4 2.19 2.31 51% 3.20
5 2.28 2.11 48% 3.10

The third column is number of PA with runners on.  the last column is the total number of runners on.  So, the #3 hitter has 2.23 PA with men on, totalling 3.00 runners, for an average of 1.35 runners (if there is at least 1 runner on). 

For the #5 hitter, if a runner is on, he’ll see 1.47 runners per PA.

What this means is that the #5 hitter will have runners in a better position to score already (i.e., the more runners per PA, the more likely that the runners are on 2B or 3B).  This is why you prefer your non-HR hitter in the 5-hole, and the HR hitter in the 3-hole.

Not to mention that the #3 hitter gets more PA with 2 outs than any other spot (maybe that’s what Geren was trying to say?!?).  Again here, the HR has more value.

That’s the basic rule here: if you have two equal hitters, put the HR guy in the 3-hole.

I understand you want your best hitters up when there are more men on base.  It’s kind of counterintuitive to say you’d prefer your HR hitter to bat when there’s fewer on base--but that’s just compared to another hitter of equal wOBA and fewer HR, correct?

I bet Geren was trying to say that the #3 hitter gets the most PA with no one on base and 2 outs.

I’m not sure how Geren reads The Book and then ends up deciding to bat Orlando Cabrera second every day.  “Disciple” was probably far too strong a word.

Ken, that is because one, this is just an article on a web site. What is written in an article and reality are often two completely different things.  Two, a manager being a “student of the game” may be simply that he has actually heard of the word sabermetrics.

It is also one thing to read a book, like it, and agree with things it has to say, another to actually understand it, and still another to actually know how to use the material and information to your advantage.

Any manager who puts more than 10 minutes a day into constructing lineups does not really understand what it is we are trying to say about lineups. More or less…

Jeff Luhnow reads the book too.

http://www.beyondtheboxscore.com/2009/4/7/824829/q-and-a-with-jeff-luhnow

I love it that he calls MGL “MGL” and not his real name!

Jeff is a stand-up guy.  Very nice words for us; and for the rest of it, a very insightful way to put the issue:

First of all, I am constantly impressed with the quality of the publicly available research on the game of baseball. It is clear that there are numerous very talented and smart people who study the game and develop amazing and true insights. MGL and Tango’s book is a great example of this, as is the work shared and discussed on their website. I’ve had a chance to talk to and work with both of them and they are first rate baseball thinkers. To me, these two and many more like them are an important part of our industry… both in terms of helping shape the game in some ways as well as generating public interest in the analytical part of the game.

As a baseball club, we read much, if not most, of the thoughtful research done on the topics that matter to us. When we know the people and the quality of the work, we consider the findings useful and may even act on them. However, since we don’t control the work and can’t do the quality checks to make sure the data was clean, the methods were up to our standards, etc, it becomes more difficult to act on the conclusions. That drives us to develop our own capabilities, which we have done for the past six years, since fall of 2003 when I arrived at the Cardinals.

Our analytical group, currently led by Mike Girsch and under the supervision of John Abbamondi, produces very hiqh quality work that is both highly relevant and actionable. When I compare it to the quality of work I used at any of the world class companies that I worked for prior to joining baseball, it is equal and sometimes better. Is it better than the public work done? For our purposes it is, because we trust it more and it helps us answer questions that we need in order to make decisions. Our analytical group includes four full-time, highly experienced employees and a number of other part time contributors. This is what they do, so you can be assured that they do it well.

That has to be one of the best and most-insightful interviews I’ve read in a very long time.

What would the A’s optimal lineup look like last night?

Cabrera at 2nd seems to be a surrender to tradition, the OC hits like most #2 hitters throughout most of baseball history. 

Geren’s got the ultimate strikeout guy in Jack Cust, but he hit 8th, not 2nd.  But then again, they were facing a lefty.  Against a lefty and with his defensive limitations, you kind of wonder what Cust was doing in the lineup at all.

MGL/6: While it’s true that spending so much time on lineups is silly, and the opposite of your main point in The Book, I think it’s equally silly to do something as blatantly stupid as batting Cabrera 2nd and Cust 8th. Last night, Geren basically used the two worst hitters in the lineup (considering a left-handed starter) 1-2. That’s pretty inexcusable, even if it’s just one game and they would’ve lost anyway.

From what I’ve seen of Geren over the past 2 years, he’s actually the exact opposite of an enlightened manager. It’s strange to see Billy Beane sticking with a guy who has proven to be just as bad as Dusty Baker (anybody else wondering why Sean Gallagher seems to be unable to pitch these days? Maybe it has something to do with Geren leaving him out there to die last year when Gallagher was in pain, throwing 5 mph less than normal, and giving up 9 runs in an outing).

Yeah, really good interview with Jeff.  Not the usual “fluff” piece and canned answers.

I’ve said this before, but you are not going to get a more intelligent, open-minded, ego-less, and reasonable guy then Jeff (in a baseball organization).  He will make a helluva GM one day soon, in my opinion.

I’m not sure what this means though:

In some ways, it’s the SS position for left handed hitters, of which there are so many in today’s game.

Unless he just means there happen to be a lot of LH playing 2B.  I don’t think there is any good reason for that to be true.

Anyone know who Mike Girsch is?  I’ve heard Sig mention him before, but I don’t know anything about him.

JD, if you say that you don’t want your manager to spend a lot of time working on lineups, you cannot also criticize him for having a sub-optimal lineup. I don’t see him batting Holliday 9th or Ellis 4th, so he has a “reasonable lineup.”

Then again, no manager can “figure out” an optimal lineup, which is another reason not to sweat it.  (And neither can you, JD, nor I, for that matter).  So any criticism of a manager’s lineup by a fan or even by me, is swinging a sword through the air.  Without a sim or some fairly complicated mathematical analysis, there is not way to figure out the an optimal lineup. 

Plus, I have NOT seen ANY managers who go with anything other than the traditional lineup more or less, regardless of how “sabermetric” they consider themselves.  They all put their best hitter 3rd or 4th, with a power guy in the 4th and 5th slots.  Most of them pay little attention to the overall quality of the #2 hitter.  Etc.  Managers also don’t like to switch their lineups too much against RH and LH pitching - maybe for good reason (they think that players get comfortable in certain slots and they don’t want to offend good LHB by moving them down in the orcder).

FWIW, against a LH starter like Saunders (and don’t forget that a starter like Saunders is only going to pitch 5+ innings), the A’s score 5.007 rpg with that lineup from last night. If I switch #1 and #9 (bat Ellis 1 and Sweeney 9), they score 5.072, which is 10.5 runs per 162, which is a lot, even for only one game.  If we also switch #2 and #7, to get a better hitter in the 2 hole, they score 5.076, another .65 runs per 162 - negligible of course. The problem with that lineups is that there are just too many LHB in that lineup.  For example, I don’t know if they have an alternative, but Chavez should probably not be batting versus a LHP with any kind of a platoon split.

Lineups are tricky, which is why I say to a manager, just do what you want and don’t spend too much time on them.  I mean, if he wants to listen to some sabermetric recommendations, that’s fine.  But if he thinks that a player is more suited to or comfortable with a certain slot, for whatever reason, that has to trump the sabermetric answer I would think - or at least cancel it out.

Now what the A’s really should do is get some pitching.  Eveland is NOT a good starting pitcher.  I have him with a projected ERA of 4.70 and I think that is in relief or at least in partial relief…

MGL,

There’s a middle ground between taking 10 minutes to calculate all the lineup possibilities and tinkering and putting a guy who is pretty clearly one of the two or three worst hitters (and against righties, THE worst) in the 2 spot.

I’m not quibbling over Giambi 3 instead of 5 or something like that. Batting Cabrera 2nd is about as stupid as batting Holliday 9th. Or batting Cust, your second best hitter, 8th.

As for starting pitching: I do wonder why people always think young/inexperience = bad. Sure, all the young guys could be bad, but isn’t it hard to call a pitching staff bad when nobody really knows how the young pitchers will perform? There’s a pretty good chance that Eveland is out of the rotation in a couple months if things go right.

Sure, all the young guys could be bad, but isn’t it hard to call a pitching staff bad when nobody really knows how the young pitchers will perform?

Not it is not.  If the young, inexperienced pitchers are ones who excelled in the minors and thus had good projections, then “we” would not call them bad.  Calling someone bad or good has not much to do with the uncertainty in that description, at least as the words bad and good are used in this context. If you like, I can say, “He is likely bad,” or “He may end up being good, but as of now, our best guess is that he is bad.”

I know nothing about Eveland.  I have him with a bad projection.  Not terrible, but about .3 runs worse than league average.  Some other projections have him much better, so who knows?  I probably picked a bad example of a pitcher who I “think” is bad.

When you say “we,” you are talking to the wrong audience on this site, JD.  No one says things like that here, certainly not the authors of The Book.  If I make a comment about a player, it is generally based on my projection or the consensus projection for that player.  If that projection turns out to be inaccurate, then so be it.  My comments about a player have nothing to do with their age or experience, unless I am commenting specifically about their age or experience (and I am exceedingly unlikely to throw “experience” into the mix, although with David Gassko’s research on playoff experience, I may have to change that).  If I say a player is bad or good, that player could be young, old, experienced or inexperienced.  Gallardo and Lincecum are very good pitchers, young and inexperienced.  Ponson and Livan Hernandez stink, old and experienced.  So on and so forth…

Bringing up an old thread…

In some ways, it’s the SS position for left handed hitters, of which there are so many in today’s game.

I think what he means is, RHH tend to hit ground balls to SS and LHH hit them to 2B. If there were more LHH than RHH, the second baseman would probably get more total chances than the SS.

Actually, looking at Chris Dial’s figures, the 2B gets 95% as many total chances as the SS. That’s more than I would expect given that LHH are about 44% of all batters.

dcj - Chances for a second baseman include assists on DPs and force outs at 2B on balls hit to the left side of the infield and most caught stealings whith a RHB at the plate.  The true split on ground balls fielded by 2Bs is 84% of the balls fielded by shortstops.  That is 45.6% of all balls fielded by SS + 2B compared with 41.7% of all non bunt PAs by LHBs.  A slight edge but about what you expect since the 2B can play deeper and therefore get to more ground balls plus he also has to cover a larger area when the 1B is holding a runner at first.