Lineup Optimization for the Baltimore Orioles

Now that the regular 2011 Oriole lineup is pretty much set, I thought it was time to mess around with lineup optimization. What, you have something better to do until Spring Training?

Using ZIPS projections and BaseballMusings.com 's lineup optimizer, I constructed what I thought would be the optimal "traditional" style lineup. Here's the lineup and the projections I used initially:

Player 1:
Player 2:
Player 3:
Player 4:
Player 5:
Player 6:
Player 7:
Player 8:
Player 9:

* I couldn't find a ZIPS projection for J.J. Hardy yet...I used the FanGraphs.com fan projection instead.

According to the lineup optimizer, this lineup will produce 5.0 runs per game, 810 runs for a season. By contrast, the Oriole offense only scored 3.8 runs per game in 2010. 5.0 runs a game in 2010 would have put Baltimore among the offensive elite in the American League, only behind the Yankees (5.3) and tied for second with the Rays and Red Sox. Hopefully, these projections have some merit.

Anyway, here's the optimal lineup given by the calculator. It would produce 813 runs over a season.

Markakis
Scott
Jones
Lee
Reynolds
Hardy
Wieters
Pie
Roberts

I would really like to think that Roberts power has been sapped so much that he would be a bottom of the lineup hitter in 2011 but that's my heart speaking more than my head. But I can't find a projection that has him slugging .450 again.

Regardless,the offensive talent has improved greatly which as we've seen in the past matters much more than the order of the lineup. With health and a little luck (like Matt Wieters finally starts to hit like we all think he could), this offense could actually be something special.

Excercises in Oriole Lineup Optimization

Just so we're clear, this post is NOT in response to last night's heartbreaking loss to Tampa. The creeping fatalism, once so identified with Red Sox fans, that has taken hold in Birdland will not breach the ramparts of Dempsey's Army. I so swear.

No, I started looking at this before the game began, when I looked at the lineup and saw that Dave Trembley had designated Miguel Tejada as our Opening Day cleanup hitter. That's right, the guy who hasn't ISO'd north of .200 since 2005, the man who has only slugged more than .450 once in the last three years (.455 last season).

So, I wanted to see how much that was hurting the team. So I went to the Baseball Musings Lineup Analysis site to check it out.

First, I entered last night's lineup with 2010 CHONE projections for the lineup:

Brian Roberts
Adam Jones
Nick Markakis
Miguel Tejada
Luke Scott
Matt Wieters
Felix Pie
Garrett Atkins
Cesar Izturis

This lineup is projected to produce an average of 5.128 runs per game or 831 runs over the course of the season.

Now, I entered the lineup I would like to see. (We'll forget about the personnel for a minute and just rearrange them...)

If you check in regularly,you know I have endorsed Luke Scott for cleanup hitter to start the season. I have framed this argument using slugging percentage or ISO but I think I need to break it down to more basic levels. Here's what I like in my cleanup hitters...With runners on base, I want him to hit the ball a long, long way, as far away from home plate as possible so that the base runners have the most time possible to round the bases and reach said home plate. The further he hits it, the more likely the base runners are to score. He might even hit it over the fence. And I want him to do it more often than anyone else on the team. Silly me.

Right now, that man is Luke Scott.

Brian Roberts
Nick Markakis
Adam Jones
Luke Scott
Matt Wieters
Miguel Tejada
Felix Pie
Garrett Atkins
Cesar Izturis

How does this version shake out? This lineup would score 5.148 runs per game or 834 runs per season. That's only three extra runs per season. Better but probably not that significant.

What's the optimum lineup?

Roberts
Markakis
Scott
Wieters
Jones
Pie
Atkins
Tejada
Izturis

This lineup would score 5.157 runs per game or 835 over the course of a season, only one more run than my lineup.

However, what if we moved Luke Scott to first base and plugged in Nolan Reimold at first base?

The best lineup would look like this:

Roberts
Markakis
Wieters
Reimold
Jones
Tejada
Pie
Scott
Izturis

That lineup would produce 5.274 runs per game or 854 runs over the course of the season.

So as you can see, lineups are fairly overrated, at least in comparison to upgrading the talent which moves the needle much more.

As long as you keep Izturis in the 9th spot and Roberts in the 1st, you can jumble around the rest of the lineup with only two or three runs difference over the course of the season. Could that mean a win or two? Sure. But I think this difference could be made up for with luck on either side. But replacing Atkins with Reimold instantly adds 19 runs over the course of the season. Now that's a huge difference.

So maybe we Oriole fans should worry less about the order of the lineup and more about the personnel that comprises it.

Could Nolan Reimold or Luke Scott Play First?

This comment under my Luke Scott post got me thinking...

Replace Atkins with Pie and that lineup is a thing of beauty. It would go



S L R L R S R L S


No consecutive batters from the same side. Man I wish Trembley would let Reimold or Scott play 1st.

math_geek

First, I hadn't realized how well balanced the handedness was in my sample lineup.

But really it got me thinking about first base. Why couldn't Nolan Reimold or Luke Scott play first on a regular basis? You could still sub in Atkins to add flexibility and give Reimold his reps in left. But replacing Atkins with Pie would give the O's the best offensive lineup.

CHONE Projections for the players involved:

OPS
Atkins      .739
Reimold     .832
Scott       .808
Pie         .767
 

Even Pie is projected to be a significantly better hitter than Garret Atkins. And even if we assume average fielding from Reimold in left, Pie is a big upgrade with the glove.

Reimold and Scott would both be big offensive upgrades over Atkins but how big is the difference in their defense? For his career at first base, his UZR is -6.3. That's only over 708+ innings and is a bit of a small sample size. We'll bump him up to -.3 UZR and give him the benefit of the doubt.

Now, we can measure this in terms of WAR. How badly would Scott or Reimold have to play at first to offset the offense they bring to the table. Plugging some values into the WAR spreadsheet, this is what I came up with. (I am assuming all will be equal baserunners.)

UZR
Scott     -19.0
Reimold   -26.0

The worst defensive first baseman (of regulars) in baseball last year was Billy Butler. His UZR was -7.4. You will occasionally see a guy have a season with -16, -18 UZR (Mike Jacobs had a -19.4 UZR in 2008) but it would be virtually impossible for Scott or (especially) Reimold to play first base so badly that they would offset their much improved bats.

The best choice would be to make Scott the everyday first baseman, rotate Pie and Reimold through LF/CF/DH with Adam Jones and bring Atkins off the bench.

But that's not happening.

An Excercise in Optimizing Lineups...

I'm stealing this idea from Beyond the Boxscore but I don' think they are in any hurry to do this exercise with the Orioles so I'll do it here.

It's all about optimizing lineups, getting the most bang for your buck out of the players that you have. I think batting order is very overrated when it comes to run production (I think Sky at BtB has said the same thing) but it is interesting to see how unorthodox a lineup can look and, theoretically, score more runs per game.

Like BtB, I'll be using the standards found in The Book: Playing The Percentages in Baseball as well as David Pinto's lineup tool found at Baseball Musings.

Here are the main players for the Oriole lineup and their CHONE projections for OBA and SLG for 2009:

                OBA   SLG
B. Roberts     .359  .423
M. Mora        .328  .414
N. Markakis    .376  .474
A. Huff        .354  .466
L. Scott       .352  .462
A. Jones       .337  .444
G. Zaun        .313  .355
F. Pie         .320  .407
C. Izturis     .287  .326

Using the rules, the first step is to identify the top three hitters on the team. For the Orioles, that projects to be Brian Roberts, Nick Markakis and Aubrey Huff. These guys will take up the top three spots in the lineup.

For leadoff, OBP is king. You also want the batter with the least power among those three elite hitters on the team. Speed is icing on the cake. Hello, Brian Roberts.

In the two hole, it's the best hitter between Huff and Markakis. This hitter comes to bat in more big situations than any other hitter in the lineup. #2 belongs to Nick.

Cleanup should be one of the best hitters on the team and the guy with power. This would be Huff. You could quibble that Markakis may have more raw power next season and you could swap the two without much argument from me. But Huff is #4 for now.

The next two best hitters on the team are Adam Jones and Luke Scott. More value is gained in the 5 slot with a batter who hits singles, doubles and triples rather than living and dying with the home run. Scott doesn't live and die with the homer but Jones is probably a more well-rounded hitter is this regard. Scott goes to #3 and Jones is #5.

Melvin Mora, Gregg Zaun, Felix Pie and Cesar Izturis are left. From here, the order goes from best hitter to worst with a caveat that if you have someone who can steal bases, they add some value in the 6 hole. Only Izturis is a fair stolen base threat at this point in his career but the bat is so bad that I don't think he will leverage that skill too often. Mora is the best of what's left, followed by Pie, Zaun and Izturis.

Your "optimized" 2009 Orioles lineup:

1 Roberts
2 Markakis
3 Scott
4 Huff
5 Jones
6 Mora
7 Pie
8 Zaun
9 Izturis

Plugging this information into the Lineup Optimization Tool, this lineup is estimated to score 4.885 runs per game. If you swap Huff and Markakis, the total is 4.879 so I was barely correct with Markakis in the 2 hole.

Pinto's best lineup? It scores 4.898 runs per game.

1 Roberts
2 Markakis
3 Jones
4 Huff
5 Scott
6 Pie
7 Zaun
8 Mora
9 Izturis

Mora and Pie, as well as Scott and Huff are interchangeable with the same production.

One wrinkle: what if we add Matt Wieters? Wieters' CHONE: .349 OBA and .439 SLG.

Wieters would fit in at the 5 slot and bump everybody down:


1. Roberts
2. Markakis
3. Scott
4. Huff
5. Wieters
6. Jones
7. Mora
8. Pie
9. Izturis


Just the addition of Wieters...brings the runs per game up to 5.031

Pinto adds this lineup for a runs per game of 5.041 moves Wieters to number three.

How about the run production for a "traditional" lineup?

1. Roberts
2. Mora
3. Markakis
4. Huff
5. Scott
6. Zaun
7. Jones
8. Pie
9. Izturis

This configuration gives you a runs per game of 4.857.

Swap Wieters in for Zaun in the 6 spot? That gets you 4.999 rns per game.

Difference in runs per game using The Book rules vs. the traditional approach: .028

The Book vs Pinto? Pinto wins by .013

The Book vs The Book with Wieters? Wieters improves the rate by .146 runs per game.

Traditional vs. Traditional with Wieters? Wieters adds .142 runs per game.

Conclusion? I would love to see a team try one of these non-traditional lineups sometime but as you can see, it matters little. I suppose it might let you win a game or two over the course of a season which could be a

But adding a better catcher? It improves jumps by at least a factor of 5.

So improving the actual members of the lineup means far more than the shuffling of said lineup. Go figure...