Optimizing Lineups: Chicago Cubs
The Book has a nice approach to assembling a lineup. Simply put (details here), you split your 8 starters up three ways - the top 3 hitters, the middle 2 and the rest (AKA The Professor and Mary Ann + 1). You can also estimate the production of a lineup with David Pinto's tool, which 'll use to compare three lineups. Or, really, compare two and let the tool pick a third. Casey Stengel makes a cameo towards the end - stick around!
Here are the inputs, from CHONE.
| R150 | AVG | OBP | SLG | |
| Bradley | 27 | 0.298 | 0.407 | 0.520 |
| Ramirez | 16 | 0.291 | 0.366 | 0.524 |
| Lee | 15 | 0.293 | 0.376 | 0.494 |
| Soriano | 8 | 0.278 | 0.337 | 0.530 |
| Soto | 7 | 0.279 | 0.363 | 0.474 |
| Fukudome | 3 | 0.274 | 0.368 | 0.429 |
| Fontenot | -4 | 0.27 | 0.351 | 0.422 |
| Theriot | -13 | 0.282 | 0.355 | 0.368 |
Building by The Book
The bottom three take care of themselves, rather easily. Fukudome projects mildly positive, Fontenot takes it negative and Theriot goes a baker's dozen under.
6 Fukudome
7 Fontenot
8 Theriot
I'm not going to go with the pitcher batting 8th, sorry Big Z.
6 Fukudome
7 Fontenot
8 Theriot
9 pitcher
Actually, I'll take that back. Zambrano should hit in the 9th spot, but Harden, for example, would go in the 8th slot. If you check out this thread, you'll get the idea that bad hitters, really bad hitters, should be "isolated" from the better hitters found at the top of the order. And a reminder that egos need to be tended. Flip a coin on Big Z - an ego spot or 9 spot. Even though putting him 9th does mean he's a better hitter than his rotation mates.
Back to the bats at the top ...
The best three hitters should fill the 1,2 and 4 spots. Bradley, Ramirez and Lee are clearly the guys - Lee goes first, since he's the least reliant on home runs, Aramis is the most reliant on the bomb, putting him 4th, leaving Milton, appropriately, with his high OBP:SLG ratio in the second spot. You could justify a switch of Lee and Bradley. If you want Lee to continue to GIDP like crazy.
1 Lee
2 Bradley
3
4 Ramirez
5
6 Fukudome
7 Fontenot
8 Theriot
9 pitcher
The Book says to put the best remaining hitter 5th, unless he's basically a home run guy and not much else. It's hard to say, based on CHONE, who the better hitter is, but Soriano fits the home run guy mold better than Soto. So, Fonzie goes third.
1 Lee
2 Bradley
3 Soriano
4 Ramirez
5 Soto
6 Fukudome
7 Fontenot
8 Theriot
9 pitcher
Measuring Up to the Alternatives
I ran the lineup in Pinto's tool. Click for full results. The lineup by The Book projects to 5.379 runs per game.
The best line-up, according to Pinto's tool, using CHONE again, yielded 5.614.
1 Bradley
2 Lee
3 Fukudome
4 Ramirez
5 Soto
6 Soriano
7 Fontenot
8 pitcher
9 Theriot
This tool-generated lineup flips Lee and Bradley, Fukudome and Soriano, and the pitcher with Theriot. I really think the only significant difference is the flipping of Fukudome and Soriano.
Lou Piniella has told the media a bit about his 2009 lineup already. This likely Cubs lineup projects to 5.317 runs.
1 Soriano
2 Fukudome
3 Lee
4 Bradley
5 Ramirez
6 Fontenot
7 Soto
8 Theriot
9 pitcher
Platooning
Mr. Stengel has a bone to pick. Whether it be my take on the The Book's technique, or Lou's or Pinto's, old Casey might suggest changing some stuff against left-handed starters.
As far as my Book-based creation, I don't think anything would change order-wise. Simply plug in Reed Johnson and Aaron Miles, and possibly re-order the 6-8 spots a little. I don't think anything useful would come of that, so I'd probably leave it alone.
Lou is faced with a bigger challenge - or at least I think so. The 2 and 6 spots will probably get filled by Miles and Johnson, respectively. I would hope Soto gets moved up a spot. Even against righties.
Pinto's tool doesn't let you analyze platoon effects. I'd guess restricting pitchers to lefties would cause a bunch of changes, moving things around more than the other options discussed.
In Closing...
Without regard to the platooning, the production for the three lineups over a full season:
Pinto's Best: 909 runs
The Book: 871
Lou: 861
Taking the output of the tool at face value, The Book will beat Lou by one win (assuming they both manage the pitching staff the same, hah) but falls short of Pinto's Best by almost four wins.
12 comments
|
0 recs |
Do you like this story?
Comments
Very nice.
Love reading this sort of stuff. Hope someone does one for the Braves, as I’m far too lazy to attempt it.
by Smoltz's Beard on Mar 20, 2009 11:07 AM EDT reply actions
I don't like Pinto's tool much.
Once you piece together how it works, you get an idea of how imprecise it is. It’s impossible to say that any of the lineups mentioned in this article are better than the other based on his model.
I ran an analysis similar to this using my own lineup-based LWTS on GROTA a while back, and came up with a different-looking result. What really stuck with me is how little it all matters, though. So long as you’re keeping a really crappy batter out of the #2 spot, you’re doing fine.
Colin, I've been trying to hunt down discussions about the faults of the Baseball Musings tool for a few days, but can't find them
Can you remember any, or do you care to share what you’ve found?
Beyond the Boxscore // Calling BJ Upton lazy is lazy.
by Sky Kalkman on Mar 20, 2009 11:45 AM EDT up reply actions
I'm working from memory here.
IIRC, it’s based on Cyril Morong’s weights for OBP/SLG:
http://www.beyondtheboxscore.com/story/2006/2/19/192135/078
Regression-based run estimators always tend to have pretty high error bars, and that goes extra for ones that use OBP/SLG as inputs instead of components, I think.
Of course, there’s also the question of the fact that the weights are based upon the other players in the lineup; the more unconventional the lineup the tool presents, the less confidence we have that the weights maintain their value.
And yes, I'm aware you could present the second criticism...
…for any static lineup optimization weights. I think the problems compound.
Yes, Sky said the same thing yesterday
the good news is, I can definitively declare my Book lineup the best. Prove me wrong.
by Harry Pavlidis on Mar 20, 2009 4:02 PM EDT up reply actions
In other words, the typical average error of a linear run estimator is +/- 20 runs at the team-season level. I’d wager on a larger error for the Morong weights, although I haven’t tested this. So when we say that two lineups are 40-runs apart using these weights, really what we’re saying is that they could be essentially identical.
And that’s not taking into account the idea that some of the weights are just plain wrong; it’s possible that in fact one or more lineup spots are being over/underweighted as a consequence of the regression.
The Nationals
If I understand correctly, the Nats optimal line up would be:
1. JOHNSON
2. DUNN
3. MILLEDGE
4. ZIMMERMAN
5. DUKES
6. GUZMAN
7. HERNANDEZ
8. PITCHER
9. FLORES
Do I have that right? I’m a little confused as to how to value Dunn. CHONE projects him to be .232/.382/.482, so how much should I factor his AVG in, if at all? Would Zimmerman (.296/.364/.488) be more valuable than Dunn, or is Dukes (.263/..373/.462) more valuable than Zimmerman?
Should more emphasis be placed on OBP or OPS or a mixture of AVG, OBP and SLG?
AVG is important to help figure out ISO and BB%
Why those are important considerations is outlined in the original article.
Dunn walks a lot, hits a lot of homeruns (and doubles) and doesn’t single often at all. He’s the second-best hitter on the team. Here’s everyone projected to be above average:
25 Nick Johnson
16 Adam Dunn
13 Ryan Zimmerman
11 Elijah Dukes
8 Josh Willingham
I’d probably go Johnson, Dunn, Willingham, Zimmerman, Dukes, etc.
Beyond the Boxscore // Calling BJ Upton lazy is lazy.
Suggested Chicago Cubs lineup
Check out the suggested Chicago Cubs lineups at http://sportspowerrankings.wordpress.com .
A new suggested lineup is posted prior to each game. The suggested lineups are based on a combination of statistics and time honored managerial principles.

by 


















