David- are these trends consistent with last years data as well? It would be interesting to see the breakdown in DER by direction for the sabermetric teams vs the non sabermetric teams.

Boy there are a lot of cool questions you can ask when you have nice big quality datasets like this. As a bioinformatition, I have to say this looks like fun.

>>Maybe, for some reason, the model does not value the balls hit at Millar as highly as the balls hit at Mientkiewicz.

Mientkiewicz probably just had an off-year defensively. Both Defensive Win Shares and Baseball Prospectus's RAA show a significant decline from last year.

No. Mientkiewitz is much better than Millar. He just had the good sense to let Reese or Bellhorn take the tweeners in the hole while he is properly covering firstbase. Millar would go after everything, often playing himself out of posiition.

This metric also apparently skips a criticala skill of the firstbaseman-saving the other infielders throwing errors. Minky kills Millar here.

Could it be something like Millar playing behind the runner on 1st more often instead of holding him on? I have no idea if this is true at all, I'm just fishing for an explanation.

Where's slice 16 at? Millar is getting .844 in slice 16 while Minky only .533. It can't be straight away 1B because the predicted DER is higher for both players in slices 17-19. They both are over .77 in slices 17 and 18 and then Minky gets .846 to Millar's .500 in slice 19. And there are a lot more balls hit into slice 16 than 19. Maybe Millar plays further from the line.

Didn't Millar and Ortiz play 1B the most for the Red Sox in 2002-2003? And wasn't Millar hurt a lot? This could make it easier to have a high rating in Fenway in 2004.

Perhaps Minky's improvement on grounders going from the Twins to the Sox has to do with artificial turf vs. grass? Just a thought.

I don't know, Lefty. In general, fake turf is supposed to give you truer bounces than real grass. At least when fake grass was prevalent, fielding appeared to be better on the artificial surfaces.

kevin: look at the data again. Millar's getting the biggest win in regions 15 & 16. Based on the number of balls in play and predicted DER, these appear to be "straight away first base" regions.

Nick may have the best explanation. The model is trained on 02-04 data. If the model determines "predicted outs" by effectively dividing outs by in-play balls for the respective park, then predicted outs are biased according to who played 1st base in those parks. What would be interesting would be to look at how Mientkiewicz fared at the Metrodome vs. Fenway to see if his Metrodome numbers are being penalized for playing in a park where 1B's are good fielders.

Nick: Millar played the majority of 1B in '03 (853 inn). Ortiz (378 inn) and Hillenbrand (164 inn) also played significant amounts in '03. Tony Clark played the most 1B in '02 (638 inn) followed by Daubach (456 inn) and Jose Offerman (331 inn).

Jason, thanks for looking up who played 1B in 02-03. Clark looks decent in these rankings, I don't know much about Daubach and Offerman, and Ortiz isn't the most agile man alive.

I haven't seen a picture of the slices but I don't think that 16 is straight away 1B. The three highest Predicted DERs for both of these guys are slices 17, 18, and 19. So my guess is that slice 18 is straight away 1B.

Since the common perception is that Mientkiewicz moves better than Millar my best guess on how Millar outranks him is that he plays further from the line. It looks like Millar's Red Sox rating still outranks Mientkiewicz's so the 02-03 bias can't be the complete difference (although my guess is that it's a big part).

Hey Nick, thanks for the reply. I didn't realize that both Millar's "Difference" *and* DER beat Minky's respective scores. Maybe Minky just had a bad year? I wonder if we could get David to post 1B ratings for 02 and 03... Oh, and maybe we can get him to post a chart of the directions, or at least written descriptions of the major positions. I saw in comments from another post that 8/9 is straight-away center.

I'll try to draw a chart at some point.

Thanks David, that would be cool.