A’s Lineup Permutations 2007
by Ken Arneson
2007-03-31 12:23

Billy Beane was quoted in the Los Angeles Times this morning (via Jon) as follows:

On a recent morning, Beane took note of a website that attempted to calculate how many runs the A’s could score with every possible lineup combination.

“With the Internet and blogs, you have so much more,” Beane said. “You’ve almost got a ‘wisdom of the crowd’ situation.”

Hey, I resemble that remark!

I had kinda forgotten about it until then, but since it seems Billy Beane took note of it, I suppose it’s time to dust off the old formula, and see what it says about the 2007 A’s. I plugged in PECOTA’s projections into the formula, and ran all the possible lineup combinations. The results are kinda weird, mostly because I think PECOTA’s OBP projections are too high for Mark Ellis (.349) and too low for Mike Piazza (.324). So you may want to look at the results, and mentally flip-flop Piazza and Ellis, because the formula likes Mark Ellis batting cleanup. Cleanup!

Also, I used Bobby Kielty’s projection, because he was the only name I could fill in for Dan Johnson that I was sure would be on the team. For what it’s worth, Erubiel Durazo’s projection (.333/.409) is nearly identical to Kielty’s (.336/.408). Todd Walker has a higher projected OBP, but a lower SLG (.345/.402), but that projection was for Petco Park–the numbers might project higher in Oakland. Still, all three players project similarly, so it probably won’t make much of a difference for this exercise.

The number beside the lineup is runs/162 games. The number will seem higher than actual team season run totals, probably because the optimal lineup never plays all 162 games. Lesser players get a lot of plate appearances besides the top nine players.

Top 10 Lineups
816.65 – Bradley Swisher Chavez Ellis Kielty Piazza Crosby Stewart Kendall
816.65 – Bradley Swisher Chavez Ellis Crosby Piazza Kielty Stewart Kendall
816.39 – Bradley Swisher Chavez Kielty Ellis Piazza Crosby Stewart Kendall
816.34 – Bradley Swisher Chavez Crosby Ellis Piazza Kielty Stewart Kendall
816.17 – Bradley Swisher Chavez Kielty Crosby Piazza Ellis Stewart Kendall
816.12 – Bradley Swisher Chavez Crosby Kielty Piazza Ellis Stewart Kendall
816.03 – Bradley Swisher Ellis Chavez Crosby Piazza Kielty Stewart Kendall
816.02 – Bradley Swisher Ellis Chavez Kielty Piazza Crosby Stewart Kendall
815.91 – Bradley Swisher Ellis Kielty Chavez Piazza Crosby Stewart Kendall
815.87 – Ellis Swisher Chavez Bradley Crosby Piazza Kielty Stewart Kendall

The projections are about 40 runs lower than last year’s projections. It still likes Kendall batting ninth. Stewart is this year’s Kotsay, which it likes eighth.

Last night on the A’s broadcast of the Giants game (where a miked Rene Lachemann was a hoot), Bob Geren said the top six lineup spots would probably be Kendall, Stewart, Bradley, Piazza, Chavez, and Swisher. So let’s look at that lineup:

798.54 – Kendall Stewart Bradley Piazza Chavez Swisher Crosby Ellis Kielty

That’s pretty much right in the middle of the possible combinations.

And for fun:
Bottom 10 lineups
772.85 – Piazza Kendall Stewart Crosby Kielty Ellis Bradley Chavez Swisher
772.84 – Piazza Kendall Stewart Kielty Crosby Chavez Bradley Ellis Swisher
772.82 – Piazza Kendall Stewart Kielty Ellis Chavez Bradley Crosby Swisher
772.79 – Piazza Kendall Stewart Crosby Kielty Chavez Bradley Ellis Swisher
772.77 – Piazza Kendall Crosby Kielty Stewart Bradley Ellis Chavez Swisher
772.69 – Piazza Kendall Kielty Crosby Stewart Bradley Ellis Chavez Swisher
772.65 – Piazza Kendall Stewart Kielty Ellis Bradley Crosby Chavez Swisher
772.61 – Piazza Kendall Stewart Crosby Ellis Bradley Kielty Chavez Swisher
772.43 – Piazza Kendall Stewart Kielty Crosby Bradley Ellis Chavez Swisher
772.39 – Piazza Kendall Stewart Crosby Kielty Bradley Ellis Chavez Swisher

The slowest player in baseball history leading off? Yeah, that seems like a bad idea.


P.S.   Go fill out TangoTiger’s community projections, and show PECOTA a thing or two.

Comments: 10
1.   Greg Brock
2007-03-31 13:13

1.  Congrats, Ken. Very cool to know that the man upstairs (the GM guy, not the deity guy) is reading your stuff.

Hopefully this will lead to a substantial decrease in the casting of porosity league-wide.

2.   doppelganger
2007-03-31 14:35

2.  I went to Fan Graphs and did a quick and dirty combination of Marcel, Chone, Zips, and Bill James' projections. Combining that with some of the batting order permutations suggested in The Book (Tango, MGL, Dolphin), here is what I have. (OBP/SLG in parenthesis)

1-Bradley (359/445)
2-Chavez (355/460)
3-Kielty/Durazo (Kielty at 339/409, but that should improve in a platoon. Durazo at 361/445, but only two systems even did a projection for him)
4-Swisher (362/484)
5-Piazza (340/460)
6-Stewart (347/407)
7-Ellis (342/411)
8-Crosby (328/413)
9-Kendall (363/351)
I also think Stewart and Bradley will do better than their projections, while Ellis and Kendall will do worse.

3.   D4P
2007-03-31 14:49

3.  It would be "neat" if a team had the stones to actually use the lineup that these run-predicting formulas suggest should be used

4.   Ken Arneson
2007-03-31 14:52

4.  After watching Walker and Durazo today, I doubt that Durazo's even going to make the team.

Walker came up early in the game with a runner on second and no outs, and got a pitch early in the count he could make contact with, and hit a grounder to second to advance the runner. Durazo comes up later with a similar situation, just stares at a fastball right down the middle for strike two, and then then whiffs on strike three, doing nothing to advance the cause. If Walker and Durazo are projected to put up similar numbers, I'll take the guy who does so making contact over the guy who whiffs and walks for a living.

5.   Ken Arneson
2007-03-31 14:53

5.  I think that if teams actually did use this formula, it would make the formula change, and we'd probably end up with something more like a traditional lineup.

6.   Ken Arneson
2007-03-31 14:55

6.  Plus, Walker can play three positions (badly), while Durazo only plays one (badly). Versatility is useful, badly as it may be.

7.   Tangotiger
2007-04-01 07:08

7.  Thanks for the link.

Lineup optimization should at the very least breakdown between LHP and RHP.

I'm not crazy at all about the lineup optimizer program, if it's the one based on the regression work of Cy Morong.

Here's an analysis of the Phillies and Mariners of last year, which you can try to apply to the A's.

8.   Tangotiger
2007-04-01 07:09

8.  http://www.insidethebook.com/ee/index.php/site/comments/optimizing_the_batting_order/

9.   bsinhaucd
2007-04-01 09:01

9.  the a's are batting stewart second because they expect him to be better than those projections. if he's not, i'm sure he won't be batting second.

like with tangotiger's community projections, the a's believe they have inside information that pecota does not.

10.   Ken Arneson
2007-04-01 12:07

10.  7 Well, I would have run separate lineups for LHP and RHP if someone would run projections for batters against LHP and RHP. But I don't have that data.

I agree that the projections don't match common sense much, so I take it with a big grain of salt.

Comments on this post are closed.
This is Ken Arneson's blog about baseball, brains, art, science, technology, philosophy, poetry, politics and whatever else Ken Arneson feels like writing about
Original Sites
Recent Posts
Contact Ken


Email: Replace the first of the two dots in this web site's domain name with an @.
Google Search
Ken Arneson

10   09   08   07   06   05   


08   07   

06   01   

12   11   03   02   

12   11   10   09   08   04   
03   01   

12   10   08   07   06   05   
04   01   

12   11   10   09   04   

12   11   10   09   08   07   
04   02   01   

10   09   06   01   

12   02   01   

12   11   10   09   08   07   
06   05   04   03   02   01   

12   11   10   09   08   07   
06   05   04   03   02   01   

12   11   10   09   08   07   
06   05   04   03   02   01   

12   11   10   09   08   07   
06   05   04   03   02   01   

12   11   10   09   08   07   
06   05   04   03   02   01   

12   11   10   09   08   07   
06   05   04   03   02   01   

12   10   09   08   07   05   
04   03   02   01   

05   04   02