THE BOOK--Playing The Percentages In Baseball

Here is a dysfunctional set of players if there ever was one:
Jimmy Rollins,S,.270,.330,.420,15,good,good,normal,not much
Utley,L,.290,.370,.500,28,yes,good,not much,normal
Abreu,L,.300,.425,.480,18,yes,good,not much,alot
Howard,L,.280,.355,.590,50,no,slow,normal,alot
Burrell,R,.260,.370,.510,33,no,bad,normal,alot
Rowand,R,.270,.320,.440,17,yes,good,normal,normal
Bell,R,.252,.320,.375,no,9,slow,ALOT,not much
Lieberthal,R,.265,.335,.400,13,slow,alot,not much

That one will definitely be a challenge, with 3 of the 4 best hitters being lefties.  Note that I’ll create a different lineup against lefties and righties.

Thanks....Any help you can give us here in Philly would be greatly appreciated!

The LHB-heavy lineup is definitely a problem.  Occasionally, the Phillies will come up against a team without a LHP and they will bunch Utley, Abreu, and Howard together, otherwise, Burrell splits them up somehow.

Step 1 - Take Forecast, and convert into “Vs L” and “Vs R” lines

I applied some basic adjustments so that I get a forecast against lefties and righties.  Using the forecasts supplied by Tom G, the best hitter against lefties is Burrell, followed by Abreu/Howard, and then Utley, Rowand, Lieberthal/Rollins, Bell, the pitcher.

Against righties, we get Abreu/Howard, and then Utley/Burrell, and then a chasm followed by Rollins/Lieberthal/Rowand, and then Bell, the pitcher.

Step 2 - Get Linear Weights for each batting slot per player

Take Table 52 from The Book, and apply the LWTS for each batting slot, for each player forecast.  For example, against Lefties, I get this line for Burrell, for each batting slot:

1 23.4
2 24.2
3 22.1
4 22.0
5 22.3
6 21.9
7 21.0
8 20.5
9 19.6

So, Burrell is +24.2 as a 2nd place hitter, making that his ideal spot...however...this is also the ideal spot for Howard. 

You start with the guy with the weirdest profile, which is Abreu.  His most obvious spot is #1, and maybe #2.  Not #3.  Next is Howard.  While he is their top hitter, and I don’t recommend normally having him #3, because of the lack of candidates, his optimal spot, in this lineup, is #3, followed by #4.  Not #1.

When you go through each combination, you get this as the preliminary “most-optimal” lineup slots against lefties:

1 Abreu, L
2 Burrell, R
3 Howard, L
4 Utley, L
5 Rowand, R
6 Lieberthal, R
7 Rollins, R
8 Pitcher
9 Bell, R

More to come…

Step 3 - Tweak

You want to try to minimize your high GIDP guy in the spots where you have lots of DP opps (see Table 59).  In this case, the high DP guys, according to the forecast, is also your three best hitters (Abreu, Burrell, Howard).  Ideally, you want those guys out of the #3 slot, and into the #1 slot.  If you can, put them in #2 or #5.  We’ll take this under consideration.

You want your speed guys to be followed by singles hitters, not power hitters.  Those runners are Abreu, Utley, Rowand, and Rollins.  We need to get Rollins away from the pitcher, and put the slow guy in front of the pitcher.

You want to break up your lefties.

And you want your lefty hitter to have guys on 1B.

Given these tweaks, I would propose this as the ideal lineup for the Phillies against lefties:

1 Abreu, L
2 Burrell, R
3 Howard, L
4 Rowand, R
5 Utley, L
6 Rollins, R
7 Lieberthal, R
8 Pitcher
9 Bell, R

Btw, I’m assuming that the hitters don’t have any appreciable change in approach with men on base, which may not be the case for Abreu.  If that is the case, you swap Abreu/Burrell, Howard/Rowand, Rollins/Utley.

Against righties, Abreu is an even clearer leadoff hitter.  Howard is a toss up at #3/#4.  In fact, I get the exact same lineup against lefties as against righties.

I estimate about an 8-run improvement over the season (0.05 runs per game), compared to their regular lineup, plus I’m able to breakup the lefties.  MGL did his through a sim, so let me find his post, and see where we differ.

Interesting stuff… You said, “Assuming that the hitters don’t have any appreciable change in approach with men on base...”

Actually, there is a big difference.  Only Abreu, and to a certain extent Utley, hit will with runners in scoring position.  Despite all their RBI’s, Burrell and Howard have a low BA w/RISP.

MGL posted his sim results here.

At the end of post 20, we see his best lineup (against Livan, a righty) is:
1 Abreu L
2 Rollins
3 Utley L
4 Burrell
5 Howard L
6 Rowand
7 Bell
8 Liberthal
9 Pitcher

I definitely don’t like Howard as #5.  Rowand/Rollins are fairly interchangeable (against righties).  MGL’s forecast for Bell must be alot higher than Tom G’s.  And of course, the pitcher should be batting 8, not 9.  All in all, I estimate MGL’s above lineup to be halfway between the current Phillies lineup, and the optimal one.

Tom G: now, it depends.  There is a huge amount of random variation if you look at a small sample (and by small, I mean a few years of data).  Burrell’s career line with men on base (forget this “RISP” business) is just like his overall line:
http://www.retrosheet.org/boxesetc/Jburrp0010.htm

If someone wants to say “this year”, well, your sample size won’t support you.  If you want to quote what you see visually, that’s he’s pressing like ARod, fine.  But, if you are going to quote me recent history numbers, then forget it.  You are quoting something almost meaningless.

Good point on the sample sizes, however, Burrell is definitely visually pressing at the plate right now [both with and without men on base].  The retrosheet link is interesting though, because despite Burrell’s reputation, he actually hits better, historically, with men on base as opposed to none on.

If Burrell is pressing with men on *and* bases empty, then we don’t need a “men on base adjustment”. 

In that case, you need to change your overall forecast downward.  However, if you do so, you then have to be aware of when to revise your forecast upward. 

That’s why I believe it is (mostly) foolish to use these visual clues, since the observer has to have real insight to know when things change.  Take ARod.  Would anyone really change their forecast for ARod downward for the next 7 games, with men on base, other than a smidge?

The Phillies regular lineup versus lefties (I used Glavine) is:

Rollins
Utley
Abreu
Burrell
Howard
Rowand
Bell
Lieberthal
Pitcher

Using my sim and in 100,000 game, that lineup generates 4.413 rpg. And you have the two lefties together which is always silly, as you can always break them up somehow.  Having them bat together costs you some win percentage in high leverage situations (when the opposing manager brings in a LOOGY when the game is on the line in late innings).

Using Tom’s optimal lineup above, but having the pitcher bat 9th (my sim does not easily adjust to the pitcher batting 8th), generates 4.418 rpg, which is only .005 rpg better, or .81 runs per season (assuming the Phils face all LHSP), which is essentially nothing.

Remember that my sim perfectly (more or less) accounts for the things that Tango does not account for perfectly, like GDP, baserunning, G/F ratios of the batters, K’s etc.

Maybe the only advantage with Tom’s lineup is the pitcher hitting 8th.  I forgot how many runs that is worth per game.

I would have to play around with some different lineups to see if there is a better one than Tom’s or the average real one (they seem to be the same according to the sim).

Pitcher batting 8th v 9th adds 2 or 3 runs, especially if you include PH. 

What I like about my process is that I can identify the reason that I want to put a batter in a certain spot, rather than doing various combinations to see what comes of it. 

As noted in The Book, swapping two equally talented players, but wildly different in styles (Abreu, Howard), will buy you a couple of runs.

Again, the other day, Manuel had 3 lefties in a row (Burrell was not playing), and again the opposing team brought in a lefty to pitch to them late in a close game, and again they all made out (one was a sac fly), and again the Phillies lost by one run.

He (Manuel) has to be the dumbest (and thickest) manager in the majors.  I feel sorry for the Philly fans.  My guess would be that Manuel has cost, espectation-wise, 4-5 wins in the first half.

What is puzzling is how he himself has talked about how he can’t let the “model” figure out the lineup because he has to think about the lefty issue.

Of course, as I’ve shown, my pretty simple model knows about the lefty issue, and you can come up with a pretty decent, and acceptable even by old book standards, lineup.

Yes, he is a puzzling manager for sure.  The other irony is that all managers know about and understand the problem with back-to-back lefties.  It’s not like it is a sabermetric secret.  I would give lots of money to hear his explanation for batting 3 lefties in a row.

The 2006 Mariners are quite a puzzle.  Their best hitter is also their only on base threat and their best baserunner.

I’ll give you my projections for the rest of the season and do a separate vs. LHP and vs RHP projection (we also have a platoon at DH and potentials for platoons at two other positions)

All lines presented in this order:

Ps - Name - BA/OBP/SLG/HRp600/SB (full season)/SB% (full season)/GIDP?/K (Extremely Low/Low/Normal/High/Extremely High)

Vs RHP

C - Johjima K - .290/.325/.465/15/0/NA/Normal/Extremely Low
1B- Sexson R - .260/.335/.560/39/0/NA/High/Extremely High
2B - Lopez J - .285/.320/.420/13/6/75%/Low/Low
3B - Beltre A - .260/.330/.450/18/13/80%/High/High
SS - Betancourt Y - .305/.325/.420/8/14/60%/Extremely Low/Extremely Low
LF - Ibanez R - .295/.360/.525/26/2/50%/Normal/Normal
CF - Jones Ad. - .250/.290/.380/8/3/75%/Normal/High
RF - Ichiro - .330/.385/.445/8/45/90%/NEVER/Low
DH - Broussard B - .290/.340/.500/26/0/NA/Normal/High
1B/3B - Dobbs G - .285/.300/.420/6/0/NA/Normal/Normal
DH - Perez E - .240/.280/.360/8/0/NA/High/High
UT - Bloomquist W - .225/.250/.300/2/18/80%/Low/High
BC - Rivera R - .200/.240/.300/4/0/NA/High/High

Vs. LHP

C - Johjima K - .275/.300/.460/22/0/NA/Normal/Low
1B- Sexson R - .265/.340/.580/48/0/NA/Extremely High/Extremely High
2B- Lopez J - .290/.325/.440/17/6/75%/Low/Low
3B - Beltre A - .275/.340/.475/24/13/80%/High/High
SS - Betancourt Y - .280/.310/.380/5/14/60%/Extremely Low/Low
LF - Ibanez R - .220/.290/.390/13/2/50%/Normal/High
CF - Jones Ad. - .270/.300/.430/12/3/75%/Normal/High
RF - Ichiro - .395/.460/.525/8/45/90%/NEVER/Extremely Low
DH - Perez E - .290/.350/.580/25/0/NA/High/Normal
UT - Bloomquist W - .275/.350/.350/3/18/80%/Low/High
BC - Rivera R - .240/.280/.380/6/0/NA/High/High
DH - Broussard B - .215/.250/.380/14/0/NA/Extremely High/Extremely High
1B/3B - Dobbs G - .200/.235/.260/0/0/NA/High/Extremely High

How would you order that line-up against either hand?

Correct format

Vs RHP

Johjima,R,.290,.325,.465,15,No,Normal,Normal,Not Much
Sexson,R,.260,.335,.560,39,No,Slow,A Lot,A Lot
Lopez,R,.285,.320,.420,13,Normal,Normal,Not Much,Not Much
Beltre,R,.260,.330,.450,18,Yes,Normal,A Lot,A Lot
Betancourt,R,.305,.325,.420,8,Normal,Fast,Not Much,Not Much
Ibanez,L,.295,.360,.525,26,No,Normal,Normal,Normal
Jones,R,.250,.290,.380,8,Normal,Fast,Normal,A Lot
Ichiro,L,.330,.385,.445,8,Yes,Fast,Not Much,Not Much
Broussard,L,.290,.340,.500,26,No,Normal,Normal,A Lot
Dobbs,L,.285,.300,.420,6,No,Normal,Normal,Normal
Perez,R,.240,.280,.360,8,No,Slow,A Lot,A Lot
Bloomquist,R,.225,.250,.300,2,Yes,Fast,Not Much,A Lot
Rivera,R,.200,.240,.300,4,No,Slow,A Lot,A Lot

Vs. LHP

Johjima,R,.275,.300,.460,22,No,Normal,Normal,Not Much
Sexson,R,.265,.340,.580,48,No,Slow,A Lot,A Lot
Lopez,R,.290,.325,.440,17,Normal,Normal,Not Much,Not Much
Beltre,R,.275,.340,.475,24,Yes,Normal,A Lot,A Lot
Betancourt,R,.280,.310,.380,5,Normal,Fast,Not Much,Not Much
Ibanez,L,.220,.290,.390,13,No,Normal,Normal,A Lot
Jones,R,.270,.300,.430,12,Normal,Fast,Normal,A Lot
Ichiro,L,.395,.460,.525,8,Yes,Fast,Not Much,Not Much
Perez,R,.290,.350,.580,25,No,Slow,A Lot,Normal
Bloomquist,R,.275,.350,.350,3,Yes,Fast,Not Much,A Lot
Rivera,R,.240,.280,.380,6,No,Slow,A Lot,A Lot
Broussard,L,.215,.250,.380,14,No,Slow,A Lot,A Lot
Dobbs,L,.200,.235,.260,0,No,Normal,A Lot,A Lot

Sorry for the weird format and confusion.

Looking specifically at the forecast for RHP that Matt provided, let’s hop to Step 2.

The guy with the weirdest profile is Ichiro, who is one of their three best hitters, and best player overall.  In this particular lineup, he is the ideal leadoff hitter, with #2, #4, and #5 also being not bad.  He’s terrible as a #3 hitter.

The next weird profile guy is Sexson, also a top 3 hitter.  He is ideal as #3 here, with #2 and #4 being not bad.

The other in the top trio is Ibanez.  He’s got a profile that is very adaptable to each batting slot.  #1, #2, and #4 are his best slots.  #3 and #5 are not bad.  Since we’ve got Ichiro and Sexson well-placed already, Ibanez is either #2 or #4.

Jones, the latest in the Mariners quest to replace Cameron without using Ichiro has another weird profile.  Normally, a guy with a weird profile is a good hitter.  In this case, he’s terrible, the worst hitter in the starting lineup, by far, according to Matt.  He’s pretty firm in the #9 slot (not bad enough to move him to the #8 slot).

The only other hitter who is qualified to hit in the top 4 is Broussard.  His profile is the most adaptable to any lineup slot (i.e., there’s just not any extra leverage in his profile, like there is with Ichiro).  Right now, it’s a toss-up between Broussard and Ibanez at #2 and #4.

#5 is Johima, followed closely by Beltre #6, Betancourt as #7 and Lopez #8.

So far, we’ve got this:
IchiroL
BroussardL
SexsonR
IbanezL
JohjimaR
BeltreR
BetancourtR
LopezR
JonesR

As you can see, three of the team’s best hitters are lefties.  And, since there is such a chasm between the #4 and #5 hitter, we are really being tested here to keep our lefties not clumped together.

On to the tweaking.  Broussard DPs alot and Ibanez doesn’t.  Better to put that high DP guy in the #2 slot.  So, we’ve got it right so far.  We want our fast guys to be followed by the singles hitters.  Doesn’t really affect us here.  You’d also like to have your lefty hitters have guys on 1B, which makes the Broussard/Ichiro case very appealing.

Let me propose two lineups.  One is the above one, which is a tough one to live with, with three lefties in four slots.  The other one is this one, which tries to break up the lefties:

IchiroL
JohjimaR
IbanezL
SexsonR
BroussardL
BeltreR
BetancourtR
LopezR
JonesR

Except for Ichiro as leadoff, and the quality/quantity of LHH, it’s really tough to make a bad lineup out of these hitters.

How does their typical order against RHP compare to what I proposed here?

On to the LHP.

Ichiro leaps to the forefront as to the team’s best hitter, by far.  Strange, seeing that he is a lefty.  But, Matt’s forecast is a huge reverse split, so that’s what I’ll go with.

(Remember, the optimizer is beholden to the data you as the forecaster believe in.  The optimizer simply tries to arrange those pieces.)

Anyway, he’s a clear #1, and 2,4,5 are not bad.  #3 is horrible.  Sexson is #3, though 2,4 are not bad either.

Perez, the 2nd best hitter against LHP, according to Matt, has #4 as ideal for him, and not much else.

After these three guys, there’s a huge drop in quality.  Beltre is next, and #2 is his best spot.  So far, this is great, as we’ve had no conflicts. 

Ibanez is the worst hitter in the lineup against LHP (I find splits that huge hard to believe, but this is Matt’s data we’re running), and is at home as #9.

Lopez, 5.  Johima, 6.  Jones, 7.  Betancourt, 8, which is in order of quality.  So, we’ve got:
IchiroL
BeltreR
SexsonR
PerezR
LopezR
JohjimaR
JonesR
BetancourtR
IbanezL

We’ve only got two lefties, so there’s no reason to have Ibanez and Ichiro so close.  However, now that I see the reverse split for Ichiro, you should actually be treating him as a righty hitter.  But, let’s swap Betancourt and Ibanez.  You’d like to have the high DP guy as #8 or #9, which is Ibanez and Jones.  So, I think I’d rather have Betancourt, Ibanez, Jones.  So, I’ll go with this lineup against LHP:

IchiroL
BeltreR
SexsonR
PerezR
LopezR
JohjimaR
BetancourtR
IbanezL
JonesR

Remember, if you are going to disagree, you have to decide is it:
1 - I disagree with Matt’s forecast, thereby making this exercise moot
2 - I agree with Matt’s forecast, but I disagree with the arrangement, thereby, we can continue this discussion

I note you’ve got a beef with my platoon split forecast for Ichiro and Ibanez.

I invite you to look at the real-world data form the first half of the year.  Ibanez is hitting .210 against lefties this year, and the Mariner blog-o-sphere is largely convinced this will continue, although perhaps not quite AS bad...he has opened up his swing this year (he’s trying to hit more HRs) which is costing him his self-defense hack he used to use against lefties, resulting in more Ks and weaker contact against lefties.  It could also be a warning sign that he’s lost some reaction time/bat speed and is about to decline, but right now, the common thought is that Ibanez has turned himself into something like Geoff Jenkins in his prime (note his HUGE platoon splits for most of his career).

The same is true of Ben Broussard.

And Ichiro is a career .356 hitter against lefties compared to .310-ish against righties...he does have a large platoon split, but it’s getting larger as you march through his career, and this year it’s about where I have it right now.  It might regress more toward his career norm, but I didn’t want to overcomplicate the projection.  Even if you give him .365 or .370 BA against lefties, he’s still EASILY the #1 hitter.

As for how your line-ups compare with what is typically done…

Vs. righties:

Ichiro
Lopez
Beltre
Ibanez
Sexson
Broussard
Johjima
Betancourt
Jones

Vs. Lefties

Ichiro
Lopez
Beltre
Ibanez (EVERYONE in Seattle is frustrated with Grover’s unwillingness to move Ibanez down in the order or platoon him with Bloomquist or something)
Sexson
Perez
Johjima
Betancourt
Jones/Bloomquist

Matt: I guess only time will tell regarding the platoon splits.

As for how things are really done:
RHP - what are Lopez and Beltre doing there?
LHP - Ibanez as you noted

It seems like the manager doesn’t take lefty/righty into account, which would explain Lopez/Beltre being as high as they are.

Seriously, this is pretty sad, on the part of both the Phillies and the Mariners. On the plus side, you’ll only save 5-15 runs.  On the minus side, that’s worth 2 million$.

GREAT way to put that.  Grover is costing the Mariners 2 million dollars a year the way he makes out his line-up cards (15 runs lost over replacement players...a 15 VORP player is worth about 2 mil)

Thanks for putting it in those terms...really helps me put it in perspective.

The RHP lineup that Hargrove likes to use currently:

Ichiro
Lopez
Beltre
Ibanez
Sexson
Broussard
Johjima
Betancourt
Jones

Lopez at 2nd was fine before his recent plummet back to Earth.  Before the year began I was a fan of batting Johjima 2nd, so I’m with you there, though prior to reading this I was led to wonder if a free swinging contact guy like Betancourt would make a better 2nd hitter.  But your analysis confirms Johjima being a better 2 hitter, which I like.

At this point, Lopez probably makes a good fit at 8, taking pressure to perform off him (which came on during his 1st half all star caliber run) and giving him chances to break his mini slump.

I’ve been sour on Sexson since his early slump and would like to see him moved down until he consistently drives the ball and gets pitchers to pitch around him again.  Sexson at 4 was where he was throughout last season.  Beltre is a good fit at 6 with Broussard and his steady power bat at 5.  Raul is Mr. Consistency and belongs at 3.  Betancourt deserves to get moved up from 8/9 as well.

Nice lineup, Tango.  I also like how Ibanez moves down in the LHP lineup due to his struggles with lefties.  I’m a fan of moving down good LHB who struggle against LHP but had too good a bat to remove from the lineup entirely against lefties.

Hmm I found this blog through Baseball Fever and instantly became curious on how well (or poorly) Grady Little was making out the Dodger lineup. I’ll try to post the numbers properly for you here if you have the time (or desire) to run them and optimize the Dodgers.

vs. RHP

Furcal,S,.274,.343,.356,4,yes,good,not much,a lot
Lofton,L,.316,.361,.430,4,yes,good,a lot,not much
Garciaparra,R,.307,.372,.487,16,no,normal,a lot,not much
Drew,L,.286,.394,.477,17,no,normal,not much,a lot
Ethier,L,.349,.401,.586,25,no,normal,a lot,a lot
Martin,R,.284,.342,.445,10,no,slow,a lot,normal
Betemit,S,.294,.365,.500,28,no,slow,normal,a lot
Lugo,R,.300,.363,.422,18,yes,good,normal,a lot

vs. LHP

Furcal,S,.337,.404,.551,20,yes,good,not much,normal
Lofton,L,.224,.309,.245,0,yes,good,a lot,normal
Garciaparra,R,.431,.516,.784,39,no,normal,a lot,not much
Drew,L,.256,.300,.378,13,no,normal,not much,a lot
Ethier,L,.350,.357,.450,0,no,normal,a lot,a lot
Martin,R,.400,.489,.550,15,no,slow,a lot,normal
Betemit,S,.204,.259,.310,21,no,slow,not much,a lot
Lugo,R,.300,.375,.463,30,yes,good,normal,not much

Also, if possible, I was curious what difference it would make in the lineups when Jeff Kent is inserted in the place of Lofton when Lugo plays CF?

vs RHP
Kent,R,.258,.357,.407,14,no,slow,not much,a lot

vs LHP
Kent,R,.296,.409,.593,44,no,slow,a lot,normal

Thanks in advance.

Before I run it through, can you confirm if your numbers is a FORECAST, or how they are actually performing.  For those who have not read The Book, there is simply not that large a split between how a better TRULY is against LP and RP.  (Big difference between sample/observed, and true talent.)

The numbers to provide are forecasts, as to what your best guess as to their true talent levels are.

Where to bat Soriano?

With bases empty, he’s .286/.328/.529 (2565 PA).  With men on 1B only, .293/.328/.526 (667 PA).  But, in all other situations (986 PA), he’s .253/.316/.447.  Included in that is the 31 IBB and 29 SF that should not count in his OBP and should be included in his BA/SLG.  That now becomes .244/.294/.431.

Obviously, the next step is to baseline against the league average.  (Someone else can do that here.)

So, his non-RISP wOBA is around .373, and his RISP wOBA is .320.  Without the men on base adjustment, that’s 3.3 SD apart.  With the adjustment, who knows, maybe still over 2 SD apart?

It’s very possible that Soriano simply is a better hitter when there is no one on 2B or 3B, than otherwise.

If someone is up for it:
1 - See who has the best/worst splits in RISP and non-RISP.
2 - See how Soriano’s components break down in RISP and non-RISP.

And, absolutely, treat the SF as an out, and not as a non-event.

I guess if we’re going to get picky, I mean treat the SF with the same run value as any other OF out with less than 2 outs and runner on 3B.

David Gassko chimes in with some interesting Lineup studies:
http://www.hardballtimes.com/main/article/lineup-balance/

In THE BOOK, I show the actual win value of all events.  The HR win value is as much in-line with its run value, as almost all other events.  Therefore, there is no “extra” punch that the HR produces.  There is one for the sac and the steal, as we’d guess.

Tom,

Are you saying my results are wrong?

Your results are what they are.  They are not wrong, but they have a certain uncertainty level around them.

In my case, I looked at each of the 20,000 HR from 1999-2002, and looked to see the impact of winning, at the point in time they occurred.  I *was* expecting to see something unusual, because of the “bunching of scoring”, and I didn’t see it.

***

One day I should update the chart here:
http://www.tangotiger.net/rc3.html
And include the win% for each HR class.  I would bet it would be rather linear until at least 3 HR in a game.