THE BOOK--Playing The Percentages In Baseball

Can you estimate how what the average variance is by player?

If I get a chance…

Weird that you get the RF playing closer to the line for righties than for lefties… and quite a bit closer…

How is UZR affected by the size of the park? Does it hurt a player who has to cover a large centerfield area?

I have alerted Matt Thomas to this thread.  Matt is an alumnus of both of the PITCHf/x summits.  He sits in the press box at Busch and uses still photography with a carefully calibrated camera to locate the position of batted balls.  Presumably he can also come up with precise positions for the outfielders.  I hope he is able to comment on this topic.

Alan/5, good idea.  Matt’s presentation at the 2008 summit (Slide 24) contained a table with the information about the average position of the fielders at Busch Stadium.

The presentations can be downloaded here:
http://baseball.sportvision.com/sites/default/files/2008presentations.zip

He had the LF at 298+/-11 ft from home plate and 73+/-2 degrees from the right field line.  CF at 314+/-10 ft and 46+/-3 degrees.  RF at 297+/-12 ft and 18+/-2 degrees.

"He had the LF at 298+/-11 ft from home plate and 73+/-2 degrees from the right field line.  CF at 314+/-10 ft and 46+/-3 degrees.  RF at 297+/-12 ft and 18+/-2 degrees.”

The positioning is really meaningless unless you at least split it between RH and LH batters.  Otherwise, it is like the the definition of average: If your head is in an oven and your feet are in a block of ice, on the average, you are at a comfortable temperature.

Anyway, one of the problems with my method is that the balls that are not caught are marked at the position they hit the ground and those that are caught are marked by the position they touch the fielder, which is probably 5-10 feet closer.  So I really need to adjust for that.

Anyway, I thought the exercise was interesting.  Next I want to see the average position for all the parks, or at least the quirky ones.  I will use the visiting players only (and 1/15 or so of the home players) so as not to bias each park with the home fielder who may like to play particularly shallow or deep.  Then maybe I can figure out which players like to play shallow or deep although the sample sizes will start to get small…

MGL, why do you think the handedness matters?  After all, your own data shows such little difference.  In either case, the CF plays straight away.  The corner OF are playing at 17 degrees from the line, give or take.

What this tells me is that we don’t care about handedness but about sprayability.  Because it seems that there’s alot of pull and opposite-side hitters and they average out to spray. 

So, forget about handedness, and instead focus on the identity of each hitter.  Which is what WOWY says to do.

Group by spray angles of each hitter.

If batter handedness matters, then definitely pitcher handedness matters.  Righty-righty, the fielders will play more to the opposite field, while for rightly-lefty, more to pull.

"Righty-righty, the fielders will play more to the opposite field, while for rightly-lefty, more to pull. “

Not true I don’t think. Just the opposite. When a LH pitcher faces a RH batter (or vice versa) they pitch more to the opposite side of the plate (where the ball is pitched determines whether the batters tends to pull or not) which produces more opp field hits.

Tango, right.  At first I was getting bi-modal results at the corners (e.g. 90% outs 15 degrees from LF line, 85% outs at 17 deg. and again 90% outs at 19 deg.) when I didn’t split by batter handedness, but I had a bug in the code, which I fixed. 

I have to play around with it some more.

What might be interesting is this:  Given that coaches and managers use the spray charts of hitters to position the fielders, and given that they probably do not regress those spray charts before they position the fielders, are they over-correcting when doing doing the positioning?

I was discussing this with Jeff Z. the other day.  He is trying to figure out optimal positioning for fielders given the batter’s spray chart.  I said that first you have to regress the spray chart some amount (you probably have to regress 3 elements - the average (or median) distance of all balls, the average angle, and the average spread or SD of angles).

Anyway, you can use out frequency as a function of batted ball location, as we discussed on the UZR thread, and the batted ball locations for those regressed spray charts, to determine optimal positioning for all batters (by working backwards).

I mean it can’t be long (if not now) where teams are using computer models to generate optimal positioning against batters, can it?  Plus, if batters can and do try and vary their hitting according to the positioning of the fielders (which I don’t think they do much), then you have to use some game theory to determine the positioning as well - which gets really messy!

I am making good progress on the who hit more to one side than another.

Switch hitters are kicking my butt and fouls balls

As MGL says, there will need to be some regression to get the true ability (I am looking at the change from 2008 to 2009 - Switch hitters causing issues).

I plan on posting some past data and am working on getting to a more true talent level as I find time.

"The positioning is really meaningless unless you at least split it between RH and LH batters.”

MGL, check Matt’s presentation in 2009:
http://baseball.sportvision.com/sites/default/files/20090711Thomas.ppt

It has splits for handedness, and for batting slot.

Max, thanks. Interesting presentation. According to Matt Thomas (the guy who did the presentation, the angular (theta) position of the fielders doesn’t change much if at all with the batter handedness, but that the LF and RF move in or out (the opp field plays a little shallower than the pull field).

He has basically 301.5 and 295 feet for the corners and 314 for the CF.  The angles are straight away for the CF and 18 degrees off the line for the corners (17 for the LF versus LHB).

He was observing fielders at Busch (III) Stadium only.  I don’t know if it was home, away, or both teams’ fielders.  If it was home or both, it would be biased by the home fielder’s preferences.  It is also obviously biased by the park, and the home pitchers and road batters.

My new numbers (adjusting 5 feet in distance for the height of the catch) using my brute force, trial and error system are:

RHB

LF 298 feet and 18 degrees from LF line.
CF 324 feet and 1 degree toward LC.
RF 296 feet and 14.5 degrees from the RF line.

LHB

LF 297 feet and 18 degrees from LF line.
CF 312 feet and 1 degree toward LC.
RF 305 feet and 17 degrees from the RF line.

So I have some significant differences between LHB and RHB.  I have the CF playing quite a bit deeper v. RHB and much closer to the RF line for some reason.  For LHB, the depth of the corners are more staggered.  Again, mine are rough estimates based on trial and error and out rates for each section of the outfield.

Looks like LF plays at around 265 feet in Fenway, 30 feet closer than at all parks combined!

"and much closer to the RF line for some reason”

I’d guess that RHH hit more opposite-field “slices” where the ball is curving away from the fielder and toward the foul line.  That kind of hit would be most likely when facing a same-handed pitcher throwing to the outside part of the plate, and RHH face same-handed pitchers far more often.  Assuming this is true, then RF should play closer to the line against RHH (while LF would shift toward the line much less frequently).

I’ll check the different combination of pitcher and batter handedness.  That should be interesting.  Remember, opposite-handed pitchers are the ones who keep the ball away from the batters…

Matt told me he posted something here but it didn’t show up.  I will take the liberty of posting it for him.
***************FROM MATT THOMAS*********************

Hi all,

At Alan/#5’s behest, the following is data from the 2009 “Where Fielders Field” presentation at the 2nd PitchF/X Summit. Mike/#6’s link leads to the 2008 presentation, but slide 6 of the 2009 presentation contains not only an update of the 2008 slide 24 overall averages but also the left-handed/right-handed batter breakout called for by MGL/#7. Here are the relevant highlights from that breakout:

891 left-handed batter plate appearances
LF: r(mean)=295’,r(stdev)=15’
LF: theta(mean)=73deg,theta(stdev)=2deg
CF: r(mean)=314’,r(stdev)=11’
CF: theta(mean)=46deg,theta(stdev)=3deg
RF: r(mean)=301’,r(stdev)=9’
RF: theta(mean)=18deg,theta(stdev)=2deg

1377 right-handed batter plate appearances
LF: r(mean)=302’,r(stdev)=9’
LF: theta(mean)=72deg,theta(stdev)=2deg
CF: r(mean)=314’,r(stdev)=11’
CF: theta(mean)=45deg,theta(stdev)=3deg
RF: r(mean)=295’,r(stdev)=13’
RF: theta(mean)=18deg,theta(stdev)=2deg

All of the above is from Busch Stadium III. Variable “r” is the distance (feet) from the apex of home plate to the front of the right foot of the fielder; variable “theta” is the angle the right foot of the fielder makes with the apex of home plate and the right-field foul line.

So it is that we have left-fielders retreating a few feet from right-handed batters, and right-fielders doing likewise against left-handed batters. On average, though, outfielders don’t move nearly as much angle-wise with generic batter-handedness as left-side infielders do.

Hope the above helps.
matthew -DOT m -DOT thomas -AT earthlink DOT net
*************************************************

MGL/16:  You may be right.  But it’s my sense that the trajectory of the outside pitch from same-handed pitchers is more likely to create a slicing ball that tails away from the outfielder. 

Other factors that could play a role:

1) With RHB, FBs to right will be more weakly hit (compared to LHH).  Given 1Bmen’s defensive weakness and need to cover the bag (plus 2Bman shifted toward 2B), RFs need to play more role handling shallow flies to right. So RF play more shallow.  The reverse might not apply for LF with LHH, because 3B and SS have more range to handle shallow flies. (And do LHHs pull more than RHH?)

2) If the 1B plays further from the bag with a RHH up (and no runner on 1B) because GBs tend to be pulled, the RF might be positioned closer to the line in part to handle GBs down the line, to prevent XBH and runner advancement.

d guess that RHH hit more opposite-field “slices” where the ball is curving away from the fielder and toward the foul line.  That kind of hit would be most likely when facing a same-handed pitcher throwing to the outside part of the plate, and RHH face same-handed pitchers far more often.  Assuming this is true, then RF should play closer to the line against RHH (while LF would shift toward the line much less frequently).

This is exactly why.

Except that pitchers who throw from the OPPOSITE side throw more to the outside part of the plate.

So you keep saying.  But there’s another factor, which is that when a same-armed pitcher does throw outside, the ball is on a different trajectory, moving more sharply away from the batter.  That produces the kind of slice that easily becomes an XBH down the opposite field line.

2007-2010 average horizontal pitch location in inches, from center of plate, first base side = positive and third base side = negative

RHB vs RHP: +2.5 inches
LHB vs LHP: -4.3 inches

RHB vs LHP: +1.3 inches (1.2 inches less outside than for RHP)
LHB vs RHP: -4.5 inches (0.2 inches more outside than for LHP)

Guy/21, have you seen this presentation that Alan Nathan did on HITf/x data at the 2009 PITCHf/x Summit?

http://baseball.sportvision.com/sites/default/files/AlanHitfx-v2.ppt (9MB Powerpoint file)

Look at Slide 11, specifically.

MF#23 beat me to the punch, as I had planned to post a comment related to my analysis of combined hitf/x and hittracker data from 2009.  I used hitf/x to determine the initial spray angle of the batted ball and hittracker to determine the spray angle of the landing point (for home runs).  I did the analysis separately for RHH and LHH.  The plot (slide 11) shows (final-initial) spray angle (or the “break") vs. initial spray angle, with 0=2B, +45=1B, -45=3B.  A positive value of (final-initial) is a break toward RF.  What you see in the plot is that for all hitters, balls hit down either line curve toward the foul pole (we have all seen such things).  The larger the absolute spray angle, the more the break.  For a RHH hitting to left, it is a hook.  For a LHH hitting to left, it is a slice.  It can all be easily understood from the expected direction of sidespin on a ball, along with a simple ball-bat collision model.  What surprised me when I first plotted the data was what is going on for balls hit in the general direction of CF.  These balls seem to slice, whether from RHH or LHH.  This trend would be expected if Matt Lentzner’s model of the tilted bat is correct (see THT article from about 3-4 yrs. ago).  So, if the bat is tilted down at contact (as often is the case), then balls hit toward CF would slice.  I would love to have more extensive data (e.g., from TrackMan) to do a more complete analysis on balls other then home runs.

Matt’s article:
http://www.hardballtimes.com/main/article/why-flies-go-one-way-and-grounders-go-the-other/

Thanks for the info Mike.