Suggestions on presentation

1) ditch the grey background - it's chart junk. If you need the contrast, use light grey

2) Solid lines

3) As before, numbers bottom and left, rather than at the axis

4) when breaking the curves out by position, use the same colors for each data series. Trust the picture to tell the story.

5) Field landmarks, rather than vectors, as labels

6) Kill the dots that mark the datapoints, they aren't adding anything.

I'm hoping you'll publish these for the rest of MLB as well.

Interesting to see that for the Yankees, the SS weakness is a 3B strength. ARod gets all the balls to his left, which kinda makes sense since it's easier for him to make that play. Plus ARod is a better fielder than Jeter.

David-- forgive me if this is a stupid suggestion. I'm not clear on what data/method you used to figure park effects; but if you used actual events rather than theoretical models, is it possible that _everybody_ positions their fielders to guard the line more often in Anaheim than elsewhere, not making the ground ball there "easier to field," but rather "more often fielded"... and producing your "hump?"

That is, while the "score" of the defenses at the humps is low, the contour of the "actual" line tends toward conformity-- would that happen if the prediction were a glitch? Wouldn't it just be an "empty" peak if this were pure system error?

John,

I do use actual events rather than a theoretical model. It's possible that every one in Anaheim guards the line, but it's also possible that the Angels guard the line everywhere much more than other teams. It may not be a park effect, but a team effect.

Then David-- stretching out to where the theoretical gets even thinner-- is it possible that the Yankees cheat their middle infielders toward second in double play situations, and so run them right into balls hit back throught the middle, creating "extra" outs in a lump there? Or perhaps the Yankee staff fields grounders through the box poorly leaving more chances for Jeter and Cairo or whoever? Or a combination of the two?

The impressive thing to me about your charts is their isomorphism to real players and real actions and reactions. I'm sure there will be tweaks needed, but you are breaking new ground here--it may be easy to mistake information for noise for a while. I'd be very surprised if your stuff doesn't show us all sorts of things we haven't been spotting up to now. Perhaps not here-- but we may be blind men describing an elephant for a while... it might be good to walk around it a few times before deciding which end to feed and which to shovel...