THE BOOK--Playing The Percentages In Baseball

The color coding for some colors was in the article.

Red - Slider
Changeups - Yellow/Orange

I’m guessing Fastballs are green and Curve is purple, but that wasn’t labeled.

I’d be curious to see this on a dead pull hitter and see how it looks.

Yeah sorry about the lack of pitch type labels.  They correspond to the same colors in the FanGraphs Pitchf/x section.  Anyway they are:

green--fastballs (two-seam, four-seam and cut)
red--sliders
yellow/orange--changeups
purple--curves

Troy here is Aaron Hill.  He is about as dead pull hitter as they come. 

This is awesome Dave.  One thing that I would like to see is the bottom part of the chart be narrowing.  Like a trapezoid.  That would, in my opinion, make it more intuitive - almost like a baseball diamond.

I made the same comment on fangraphs, but to me the graph would be a little more intuitive if instead of using two points to create a line use a point and an angle.  The point being where the ball crossed the plate and the angle being the actual departure angle.

I like what Dave did, because it shows the “clumping” of the angles of a pull hitter like Aaron Hill. 

Otherwise, you have to look at each point, and think about the spray angle for each point.

As I said, it takes a little getting used to (that the line only exists to link the two data points, and not to visually show the angle), but I prefer it.

Dave always comes up with some nice graphical representations of data.  I like this one very much.  Like Nick’s comment in post 3, I have also wondered whether these graphs might be more powerful as a trapazoid and that is something Dave can experiment with. 

I am more concerned that they are somewhat misrepresenting what actually occurs.  Dave is using the Pitch f/x x_value where the pitch would have crossed the front edge of the plate, not the x_value where the ball is when it is struck, which is estimated by Hit f/x.  Hit balls that are pulled are usually struck before the ball reaches the plate, so the actual x_values at contact are further away from the center of the plate for pitches from a same handed pitcher.  Until we have full Hit f/x data Dave is stuck using the Pitch f/x front of the plate x_values, but it would help me visualize the actual struck ball location if he differentiated the handedness of the pitcher by using a dashed or dotted line for opposite hand and solid line for same hand.

Peter - Is it possible to estimate when the ball was struck based off velocity and hit angle?

Nick - Depends on how accurate you want your estimate to be.  To be anything close to the accuracy that we have come to expect from Pitch f/x pitch data, i.e. within 1 inch of actual position, you would need to know the full 9 parameters of the hit ball.  So the short answer to your question is no, you need more information than velocity and hit angle.

This took far far longer than it should have, but combining the data with hit tracker makes for some interesting viewing (this is far from 100% accurate—it’s more like a preview of what could be done):

Doesn’t seem to have posted, we’ll try with a link:
http://www.imagedump.com/index.cgi?pick=get&tp=554098

I could do a trapazoid and indicate the handedness of the pitcher with different line types.  I will look into that for future graphs. 

Peter, I had not thought about the issue of where bat and ball make contact and how it will be different for pulled and opposite field hits.  That is key observation, thanks for bringing it to my attention.

Sal/10 That is fantastic.  Dave, you should definitely look into combining this with the hit tracker data as Sal has.

Awesome stuff Sal.

Sal, I really like your way of doing it.  I like how it better represents the true angle of the HR in play, while still giving the horizontal location of the pitch. Very nice.

Thanks, but I just hope that someone can determine a better way to automate the process, because that was a tedious tedious job (I did it in an illustration program).

It could probably be done easily with Excel and some trig, but the data is from two different sources, and may be a bit difficult to match up (the trajectories on hit tracker and those on the original graph are very different).

Sal #15, that looks easy to automate, if I had the pitch data.  Send me pitch data for a couple players and I’ll send you back a file for it that you can just drop data into and hit a button.

Greg, it’s not a problem of automating the graph, but rather of automating the data collection for pitch locations combined with your distance and direction data.

Ideally we’d have data for all batted balls (not just homeruns) so that we could view patterns by pitch type and location in a snazzy interactive way. For instance, we could view all curveballs on the outside part of the plate and where they went and if they were outs, etc.

That’s probbly a job for hitFX when it comes out though.