I tried to study the impact of the batting order change using ZIPS projections. I was a little disappointed in the results.

I used the average plate appearances by batting order position in the AL last year. The second spot had 103 more plate appearances on average than the 8th spot. That made sense as it is close to 6x18.

I then compared a Dozier with 735 plate appearances to a Dozier with 632 plate appearances. (Yes... I know he is not going to play 162 games and neither are any of the others but this should give me the extreme.)

I added to Dozier's 632 plate appearances all of the increases of the other 6 players using yesterday's batting order to compare with Dozier in the 2 hole.

Using ZIPS

Dozier batting 2nd 247/349/300

Dozier 8th (plus others) 249/359/305

This difference over that many plate appearances amounts to about 5 runs.

Darn! I expected more.

Why was the change so small?

The batting average hardly changes because six of the guys getting more at bats (Willingham, Plouffe, Doumit and Parmelee) are projected to have about the same average as Dozier. Mauer's extra 18 plate appearances helps some but not much in the context of 735 plate appearances.

I know the study has flaws

- If Dozier isn't hitting second, a few less outs will be made in that spot and overall the team should have a few more plate appearances on the season. That can't account for a significant change.

- I used ZIPS.

2013 ZiPS Projections ? Minnesota Twins | FanGraphs Baseball

Maybe ZIPS does not project the Twins hitters well or over projects Dozier.

- It isn't realistic that the best hitters will play 162 games. They will be replaced by guys that are not much different than Dozier and reduce the impact of the extra at bats.

I am disappointed. I thought I would see a difference of over 20 runs and then scale it back due to playing time. It didn't happen.

There are many studies written about optimizing a batting order. A well optimized lineup is shown to be better than a poorly optimized order. Moving just one guy in the context on this team and their ZIPS projections didn't make difference I expected. I will take the 5 or so runs though.

Thanks for posting, I always enjoy your analysis.

One of my favorite qualities of research is that the results speak for themselves. You can massage the results to fit a narrative, or you can present results as they are. In this case, the batting order doesn't mean a whole lot, but the it also doesn't mean nothing. I think that moving Mauer to 2 and Dozier to 8 will have more impact on a game-by-game basis, rather than a season-long basis. This means that sample sizes will always be too small to use for predictive value, but the results of the individual games can still be impacted.

Of course, you are right about the ZIPS projections. If Dozier performs better than these projections, then the move will result in even fewer runs.

Um, those projections of batters 2-7 seem pretty low.

I am not sure there would be significant difference had I used Bill James, Marcel, Oliver, or Steamer.

I decided to choose one before I looked at the data. Otherwise, I might pick the one that happens to have Dozier lower and some of the others higher to better attain the results I expected.

Dozier's OPS projections range from 646 to 672. ZIPS had him at 649.

Willingham's were 816 to 833. ZIPS had him at 822.

ZIPS was least favorable to Doumit and Morneau.

I don't think the results would be significantly different had I chosen different projections.

The projections seem low, but three of those players are in the decline phase of their careers with some injury history.

It is strange. I would bet that there would a 5 run difference in the first inning *alone*.

There is probably a way to study where the extra runs came. It might be 5 in the first inning.

We would probably give a couple back when a couple guys are on base mid game and 8-9 are due up.