THE BOOK--Playing The Percentages In Baseball

I try not to get too worked up about it, but my Reds often bat two of their three worst hitting position players in the #1 and #2 slots. 

Our manager has stated already this year that when Jerry Hairston isn’t in the lineup, we will see Willy Taveras leading off, followed by Alex Gonzalez hitting #2.  Otherwise, Hairston hits second.

Way to write off the first inning and maximize PA’s for poor hitters.

Like I said, I try not to get too worked up about it given the relatively modest differences over a season.  It’s astonishing to me, though, that a company spending the amount of money that any ballclub spends on players can be so obviously inefficient.
-j

15 runs seems like a lot to me.  Thats like 6 million dollars.

1: The lowest power/highest walk player of your three best hitters. 
2: One of three best hitters.  Doesnt ground into many double plays.
3: 5th (maybe 4th) best hitter.  Doesnt ground into many double plays. 
4: Best power hitter of your 3 best hitters.
5: 4th (maybe 5th) best hitter.  Speed is a big plus (want to steal to get to scoring position in fornt of the low power hitters coming up).
6: 6th best hitter. Speed is a big plus (want to steal to get to scoring position in fornt of the low power hitters coming up).
7: 7th best hitter.
8: 8th best hitter.
9: Worst hitter.

In NL, the pitcher switches to 8th to be farther away from the good hitters at the top of the lineup.  (If the worst hitter is far, far worse than the second worst, flip them to 8th)

It might not have made a super huge difference, but seeing Bruce Bochy pencil in Jose Castillo for the #2 spot so much last season (so much = at all) made me want to scream my lungs out.

Alex/2: that’s the basic guidelines in The Book…

Yes, that was my recollection of what I read in The Book.  (Great chapter, everyone should read it!). 

Here’s hoping that the Mariners will follow it or something similar, now that they have embraced the sabermetric approach.

To me, the part about getting the 5 best hitters up in the first 5 slots is the most important to pay attention to. And that is even more true if your 5th best hitter is noticeably better than the 6th. Getting them up in the in the *ideal* sequence, vs whether you vary from that due to right/left or due to the comfort of the specific batters--I’ll go with the latter in the real world.

Which brings me to the Cubs. Their 5 best hitters are pretty clearly Bradley, Ramirez, Lee, Soriano, and Soto. No Raines type guys in the bunch. But Soriano is experienced and comfortable at #1.  Although he is stylistically the antithesis of an ideal leadoff man, his prior experience there and the team situation combine to make it an easy and probably correct move to keep him there.

So, Lou Piniella “stumbles’ into the correct move by “accident”. If Sori had not always been a leadoff man, Lou would not be putting him there.

What about Fukodome? Coming from Japan, he seems to have lost lots of the pop in his bat vs MLB pitchers, but retained his plate discipline and good baserunning. So, Lou is apparently going to try him in the 2nd slot. Not a bad idea, IMO.

I’ll stop here, pending anyone’s response.

I agree about the comfort/ego thing.  Unless the Cubs had Ichiro, Raines, or other similar player that Soriano could accept, then Soriano’s “comfort” will take precedence.  Basically, Soriano is calling the shots, overtly or not.

No, based upon the comments I’ve seen in the papers, I don’t think Soriano really cares anymore what slot he is put in, as long as they leave him there. He realizes that his age/speed/injury pattern does not match up that well with the usual demands of the leadoff slot. He would not mind being in an RBI slot, to silence his critics.

Now would be the time to move him, and Lou would be strong enough to do that--but it’s not clear at all that it would be an improvement…

Having run various lineups through my sim thousands of times, I can say several things:

1) You can see a difference of 10 or 15 runs a year between reasonable lineups but that is probably the exception rather than the rule.

2) There are other “tangible” factors that go into optimizing a lineup that are not captured by the “basic” model, such as base running, ability (intentional or not) or desire to hit behind runners.  For example, no matter how hard I “tried” I could not an NL pitcher batting 8th to be better than him batting 9th.  As well, I often get a good or even optimal overall rpg batting poor players at the top of the lineup.  It might be because of their base running and it might be because of other things that a basic model does not consider.

Not from running sims, but my opinion - and I have said this many times - is that there are lots of intangibles that could and probably do affect a lineup’s expected rpg, again, that a basic or even a complex model does not account for.  Not the least of these, as Tango admits, is the comfort or desire of the players.  If we are trying to squeeze out at extra 1-2 runs a YEAR by batting the pitcher 8th, can you honestly say with any certainty that the poor batter batting 9th might not have his feelings hurt just a tad, to the tune of 1-2 runs a game?  The same thing goes for other players and lineup slots.  It is hardly inconceivable that certain players prefer certain lineup slots that will net them a few runs a year more than the slots they do not prefer.  It is also hardly inconceivable that a manager might not be able to pick up on other things (as I said, hitting behind runners, bunting, and proper approaches at the plate) that would make a player more suited to one lineup slot than other, regardless of what the models say. 

All the above being said, if I were advising a team, that would be the least of my concerns.  I would give them several lineups that I thought were optimal.  I would look at their various lineups and tell them what the computer says about them.  And I would give them the basic “rules of thumb” for optimizing a lineup and explain to them how and why the conventional wisdom is wrong (fast player 1st, best hitter 3rd, power hitter 4th, scrappy bunter 2nd).  Then I would leave the manager be. I would advise him to do or not to do 2 things:  One, don’t spend more than 10 minutes a day constructing a lineup.  Don’t waste your time juggling your lineup based on how the players or team are doing from day to day, unless you think it might have some psychological effect on a certain player or the team, and even then, don’t spend to much time and energy on that - there are many important things to do.  Try, if you can, to have a fairly set lineup against RHP and then against LHP and that’s it.

I don’t disagree with anything MGL said.  I would not spend more than 5 minutes advising the manager, but the one thing I’d insist is that his top 5 players bat in the top 5 slots, in some order or other.  He’s free to do anything he wants with all 9 players, as long as he simply sticks to that number 1 rule of mine.

It is a real shame to put one of your three worst hitters in one of the two top slots.  Maybe I should do a study to see which managers break my rule, and how often they do that.

I don’t disagree with Tango, but I have “run” many lineups on the sim where having one of your poor hitters in a top slot does not cost anything and in fact, may be optimal.  That being said, it is not a bad general rule (what Tango said).  However, I would run the manager’s lineup, and unless it was costing more than 5 runs a year or so, I would not worry about it.

Of course, in reality, you CANNOT tell a manager to do or not do something like that, for two reasons.  One, they don’t like to be told how to run their team on that level, and two, there is almost a zero chance that he is going to “buy” the fact that having someone in the 2 slot, for example, who can bunt, move a runner over, hit and run, etc., is worse than some computer telling you/him that that player belongs in the 8th slot.  That is the bottom line.

Now, if you had a manager that was open to that kind of analysis/advice or actually sought it out, then all the more power to you.  Despite what you sometimes hear about managers like Yost, Maddon, Francona, Acta, etc., I do NOT believe that ANY present manager would be particularly open to this kind of advice other than perhaps giving it lip service.

I would say that if you take the media out of the equation, then at least one-third of the managers would be up for it.

I think their problem is that they have to justify every single decision they make to the media, and it’s certainly not worth the effort to do so if you have to answer why to bat Pujols 4th instead of 3rd.

I don’t think that is so.  I really don’t. If a manager has enough intelligence and moxie (or whatever you want to call it) to “go for it” wouldn’t it be pretty simple for him to appease the media (and fans)?  I don’t really buy your argument.  The managers that would be “afraid” to deal with the media are the same dumb**es who would not understand or accept a computer model of lineup value.

I mean, if I bat Pujols 4th and I get questioned by the media, how tough is it to respond:

“All of our players are team players and they will do anything to help the team win.  We have done extensive research on how different lineups affect our chances of winning each and every game.  We owe it to the players and especially the fans to put out the lineup that gives our team the best chance of winning each and making it to the post season and the World Series.  Our research indicates that that best lineup is one in which Albert bats 4th rather than 3rd.  And, as I said, he is all for it, and so is the rest of the team.  Now, do you have any questions about why I brought in my closer in the 7th, or why my 4th and 5th starters never pitch more than 2 or 3 innings and never seem to get an AB?”

Do right/left matchups make any difference at all in line-ups? Or is there a way to tell?

It matters only with the option of bringing in reliever(s).

Having three-lefties in a row, as opposed to LRLRL means that there is two guys who may lose out on the platoon advantage (maybe).  If that happens 50 or 60 times a year, that’s 2 runs.

I mean it “matters” as much as any little thing to consider, the propensity to GIDP, having speedsters in front of singles hitters, etc.

I’ve been getting a lot of questions along the lines of “If you start putting your best hitter second instead of third, won’t that result in fewer baserunners for the fourth and fifth spots, making the value of the PA go down?”

Looking through The Book, I don’t see anything addressing how changing the lineup will change the value of other lineup spots.  Is it something you guys looked at?  Should we be worried about it?

There was one spot where I swapped the #2 and #3 hitters, and the results were consistent with expectations.

***

The problem is that it takes tremendous effort to code the scenarios for batting orders, and with so little potential payoff, I think my time can be more effective if spent elsewhere than trying to capture the different number of baserunners each batting order spot will get.

Granted, I’d like to get a more definitive answer to that.  I’d like to know more.  But, at some point, I have to let it go.

The correct answer is a Markov, right? Essentially you build a transition matrix for each player, and you can see exactly what batting states a player would face based upon lineup order. I think - one day I need to sit down and pick apart your Markov model.

The problem is that it takes tremendous effort to code the scenarios for batting orders, and with so little potential payoff, I think my time can be more effective if spent elsewhere than trying to capture the different number of baserunners each batting order spot will get.

Granted, I’d like to get a more definitive answer to that.  I’d like to know more.  But, at some point, I have to let it go.

You don’t get good honest answers like that many places these days.  Much respect to you Tango.

That is why I use a pure (as opposed to a Markov-type) sim for analyzing BO and is why I often get very different answers than from other models.

Ron, I don’t think Tango was too clear about the L/R thing.  Since what goes into determining the optimal BO is a player’s component projection (s,d,t,hr,bb,so,gidp,etc.), as well as baserunning of course, and some other small things, like SB/CS, bunting, and general location and type of batted balls, whether a lineup is facing a LH or RH staring pitcher makes a huge difference.  If that is what you were asking.  Most teams would likely have 2 very different optimal lineups - one versus lefty starters and one versus righty starters.  As well, and none of the models even takes this into consideration, other than my sim, since there is a platoon differential for GB and FB batters and pitchers (opposites do better, like handedness), you would likely have some different optimal BO versus heavy GB or FB pitchers.

Peter: cool, thanks.

***

LRLRL: I thought he was talking about the sequencing, not about the differing production based on handedness.

***

http://forum.signonsandiego.com/showthread.php?t=96837

The difference between the absolute best and absolute worst lineups get between 5 and 15 runs over the course of a season.

I tried to post on that forum, but was blocked.  To correct the above quote: the difference between the absolute best and THE TYPICAL lineup is 5 to 15 runs.

Seeing that moving the pitcher to the 4th slot and keeping the rest of the order relatively the same costs 15 runs, then the difference between absolute best and absolute worst is alot more than 15 runs.

I think Palmer showed it to be 40 or 50 runs or so, and that seems perfectly reasonable.

Does lineup construction become more important as individual games become more important (sort of like how an at bat becomes more important in higher leverege situations)?  Could lineup construction be worth, say, 0.2 wins per game in a hot pennant race or a playoff game? Or even if it affects 0.1 runs per game no matter what, wouldn’t lineup construction become more financially important, given how financially valuable playoff wins are?

In a random game, you can hope for a .01 win improvement. 

In your case, you are asking about the chance to improve in getting the playoffs.  If it’s the last game, and each team has a 50/50 shot of making it, then your question is how much would a good sequencing perform over a typical sequencing. And that answer would be say .51/.49.

So, in this game, you would add .01 to your chances of making the playoffs, which is enormous for one move. But everything gets magnified here, like the IBB, sac bunt, etc.

To put in another way, it does not change the win percentage for that game - that never changes - but the impact of changing the win percentage in an “important” game can be much greater or lesser depending upon the utility of getting that win or not (making the post-season, getting a bonus if you are the pitcher, maybe “winning 100 games,” etc.).

Astros batting order:
http://www.crawfishboxes.com/2009/3/26/810400/optimizing-the-astros-line

***

MGL: you were saying how when you do your sims, you don’t get the same benefit for the pitcher as I do.  Can you do me a favor: grab a typical lineup (and typical hitting pitcher), and run two sims: pitcher bats 9th, pitcher bats 4th.  (In the latter case, just move the 4 through 8 guys down one slot.)

In my Markov process, I end up with 0.1 fewer runs per game.  What do you come up with?

I’ve created a hub for links to posts others have written optimizing team lineups based on the heuristics from The Book.  Here:

http://www.beyondtheboxscore.com/2009/3/26/811173/lineup-optimization-hub

Fantastic, thanks!  Saves me the trouble…