# Into the next dimension with HIT f/x

Note: Last week I wrote an article on hit f/x, and within 15 minutes a number of astute and informed readers pointed out that I was wrong about where the point of contact was being measured from, which led to some incorrect conclusions. Mea culpa, but it did start an interesting discussion which you can still find in the comments below, although the numbers and conclusions in the following article have been corrected:

Warren Spahn once said: “Hitting is timing. Pitching is upsetting timing.” And for all that PITCHf/x has told us about who has the nastiest pitches and what locations hitters feast off of, it has had very little to say about such a critical aspect of the hitter-pitcher duel as timing. As a result, while PITCHf/x can easily tell **where** in the strike zone batters are succeeding or struggling to make solid contact, for the most part **why** that is the case is just guesswork.

Further complicating saying anything useful about hitters is that the normal PITCHf/x data gives the location of each pitch as it crosses the very front of the plate. That’s fine for pitchers, since that’s where balls and strikes are decided. But with breaking balls moving wildly as they pass the plate and a full six feet of batter’s box in which the hitter could theoretically be standing as he reaches for the ball, where the pitch crossed the front of the plate and where the batter actually made contact with it could be far apart.

Enter HITf/x to save the day on both counts. Along with the horizontal and vertical location at which the bat struck the ball, it measures how far along its flight path the ball was when it was struck. This will be especially useful once we have enough data to compare a hitter’s normal contact point to where it is when he is in a slump or on a streak to see if anything has changed. But for now, here are the average travel distances for each type of pitch using the available April data. I have converted the values hit f/x provides to where the ball is in relation to the front of the plate (as pitch f/x measures things), with positive numbers how far out front of the plate the ball was struck, and negative numbers for when a ball was hit after crossing the front of the plate.

Pitch Type Average Distance Average Velocity (mph) in front of Home Plate (ft) Fastball 1.06 91.2 Slider 0.84 83.1 Changeup 0.79 82.4 Curve 0.71 76.5 Total 0.95 87.0

So on average, batters make contact with the ball about a foot in front of the plate, with fastballs hit sooner than offspeed pitches by about four inches. Here’s a look at the distribution of contact for every batted balls in April (hat tip to Alan Nathan in the comments):

There isn’t much variation between hitters, with batters at the extremes (of earliest and latest contact) in the league taking their cuts at balls within a foot of each other. Here are the top five of each through April—for the full list, you can download this Excel file .

Earliest contact (at least 25 balls in play)

Player Name Average Distance In front of plate (ft) Alexei Ramirez 1.44 Mike Lowell 1.42 Hank Blalock 1.35 Justin Morneau 1.32 Garrett Atkins 1.30

Latest contact (at least 25 balls in play)

Player Name Average Distance In front of plate (ft) Anderson Hernandez 0.40 Emmanuel Burriss 0.46 Nick Johnson 0.51 Kazuo Matsui 0.53 Chipper Jones 0.55

One player on this list who demands attention is Alexei Ramirez: he hit .214 in April but has hit .288 since, suggesting that he may have been jumping at pitches too soon in April. But it’s really too early to know if these are typical numbers for these hitters, let alone what this might all mean about their swings and approaches at the plate.

But in general, *should* different pitches be hit at different depths? We saw above that offspeed pitches on average travel about an extra four inches before they are hit. But should batters actually try to let certain pitches travel further, and contact others aggressively before they cross the plate, or is that just a result of the type of pitch? The following two graphs show how batting averages are affected on balls hit into play by how far in front of the plate contact is made. Zero is the very front of the plate, which extends 1.417 feet past that (picture a pitcher throwing from the right hand side of the following graphs, with 0 being the front edge of the plate).

**Fastball distance**

**Offspeed Distance**

As both groups of pitches (I lumped all the offspeed pitches together because they were essentially identical) peak at about the same point, there seems to be no difference in where hitters should try to make contact with the ball for maximum results. However, while offspeed pitches are equally hittable where they are contacted, batting average against fastballs rises steadily the later they are contacted, to a difference of almost 200 points between hitters when they make relatively early to late contact.

Once all the data are available, that will open up a new area of study—not just for hitters, in terms of little tidbits like who has the best timing on each pitch, or who can still get a hit when they’re fooled, but for the whole art of pitching that is disrupting timing, and how they do it as well. For now, here’s a rundown we’ve learned during our preview of the third dimension:

— Hitters tend to hit the ball about a foot before it crosses the front edge of the plate.

— Batters hit fastballs earlier, and let offspeed pitches travel further before making contact.

— However, if hitters are able to stay back and contact fastballs later, they get better results.

— On the other hand, it doesn’t matter when a player makes contact with offspeed pitches.

— Fastballs and offspeed pitches are equally bad with which to make very late contact.

That second strikezone chart doesn’t look right. How could the pitch rise a foot after it crosses home plate?

Great point about needing to know how much a pitch moves from the front of the strike zone to when it’s contacted. I’m having trouble believing some of the data in the charts, however.

For example, the change-up to Chipper dropped a foot vertically while only traveling 1.6 feet horizontally? That seems really steep. And the curve to Josh Johnson traveled 1.5 feet sideways over 1.3 feet horizontally? That’s a ridiculous angle.

I’m open to having my world flipped upside down, but those seem unlikely. Or I’m reading the graphs wrong.

>Almost all balls are hit BEFORE they reach home plate, not after.

Bravo Mr. Jensen!

Some of those saberanalysts musta never seen real baseball.

Hey Jon.

This is great work and certainly provides an early look into the possibilities of HIT/fx.

One thing that I must challenge, though – and this, in a way, reflects Peter’s sentiments – is your measurement and reporting of “travel”. Home plate is roughly about 17” in depth from the front edge to the back corner. So I’m finding it difficult to believe that most hitters, on average, are making contact with a fastball 2.5’ beyond the front edge of home plate, or over one foot BEHIND home plate. In a recent article at Viva El Birdos (http://www.vivaelbirdos.com/2009/8/19/992832/albert-pujols-anatomy-of-the-swing), you can see that Pujols is making contact just about at the front edge of the plate. While Pujols is certainly not representative of the average hitter, I imagine that this contact point is a better estimate across the board.

While this article is very exciting and interesting, I can’t buy into the information that most big league hitters are making contact with pitches in foul territory.

Otherwise, the author makes a great point:

“… that will open up a new area of study … for the whole art of pitching that is disrupting timing”

And that’s why I wanna see not just Pitch f/x and Hit f/x,

but BAT F/X!

~~~

PS Mr. Hale, why doesn’t your article mention MGL? This study was all based on his original ideas, no?

Peter: Thanks, what a boondoggle. I should have known by where all the bunted balls were contacted. So yes, there are a lot of opposite conclusions to be made here:

– Hitters do manage to contact fastballs earlier in the zone and wait on changeups.

– But the more they’re ahead of fastballs the worse they make contact with them.

– Alexei Ramirez was probably standing forward in the zone or lunging at balls in April.

That’s really too bad that it might not be accurate at all, though – it makes a lot of these excited ramblings less rambling. Maybe game f/x can incorporate something about where the ball lands as you did.

Nick: Bad graph! It should be the other way around.

Thanks Jon, I probably should have figured that out.

Also, I hate to be a bother, but do you think that you or Peter could give me a quick explanation of some of the fields in Hit f/x? I’ve just started playing around with my data, and I’d be nice to know what I’m dealing with.

Re Peter Jensen’s comments:

1. The hit_y values in hitf/x are the distance from the point of home plate (with a positive number meaning toward the pitcher). They are NOT measured from the front edge of home plate. For example, if hit_y=1.417, the contact is made right at the front of home plate. If hit_y is larger than 1.417, contact is made in front of home plate. etc. etc.

2. Peter’s 2nd comment is worth expanding on a bit. The hitf/x data do not extend right to the contact point. Instead, they start some number of feet from the front of home plate (I forget the exact number and in any case it is ball-park dependent, and even batter-dependent). But in any case, the trajectory is measured some distance from the contact point and is only measured over a short distance. The basic idea is that the measured trajectory is extrapolated backwards to find where it intersects the pitch trajectory. Now, as Peter has said, the extrapolation assumes a linear trajectory (i.e., constant velocity), since there is not enough information to do a constant-acceleration fit. This was discussed briefly at the summit. Rand Pendleton (from Sportvision) and I were working on a technique to use the reasonably-well-known drag coefficients and gravity (which is very well known) to calculate the acceleration, then use that to extrapolate back to the contact point. It is still a work in progress. I doubt it will have a great deal of effect on the actual contact point (it has more of an effect on the contact velocity—the so-called speed off bat—which is actually a bit higher than hitf/x tells you because of drag).

Sky: I know, they messed me up too. But I checked them again and against others and that’s what we’re looking at in terms of amplitude at least (although as I mentioned that sideways curve to Johnson was kind of unusual) of the distance between the two.

Maybe something is just completely wacky with the data, but I managed to convince myself that a curveball travelling down and away at a 45 degree angle (i.e. down and sideways as much as it’s going forward) right before it got to the catcher’s glove wasn’t a completely ridiculous idea.

Alan – Thanks Alan for correcting me. All Pitch f/x and Hit f/x Y values are measured from the back of home plate, not the front. I new that, of course, but in my haste to correct Jonathan’s error I made one of my own.

Jon, but if your initial assumption was wrong, and contact is being made in front of the plate, I don’t see how your graphs can be right. Again, that would assume that the ball is rising.

Sergei: If MGL has written an similar article, please share. Otherwise these are my original ideas (and mistakes).

Scratch my last comment.

With all due respect to everybody else, I have to confess that Mr. Alan Nathan is my king!

Keep up the good work.

Alan, I’d only ask you to maybe forward the Bat f/x idea – I’m sure you understand what I’m talking about – to people who might be actually able to make it possible someday.

Or maybe even tomorrow.

– Sergei from Moscow

Sergei: I saw your bat f/x comment but actually don’t know exactly what you refer to. Please expand on this a bit. And while you’re at it, tell us how someone from Moscow got interested in baseball (privately, if you prefer, at

).

Here is a link to a hastily prepared plot of the distribution of hit_y values for the 15k hits in the hitf/x data base. Note that y=0 is the point of home plate, y=1.417 is the front of home plate. The data are binned in 0.25 ft buckets.

http://webusers.npl.illinois.edu/~a-nathan/pob/hit_y.GIF

(I just posted this also at The Book)

Note that the distribution is peaked just in front of the front edge of home plate (1.5-1.75 ft)

Jonathan: Despite your misinterpretation of the meaning of hit_y, your idea for an analysis is a good one. I hope you are able to continue it with your redefined values. Let me know if I can be of any help.

Nick: you wanted a definition of the various fields in the hitf/x data. To my knowledge, no one has written that down yet. Here is a quickee (assuming you are already familiar with the pitchf/x coordinate system):

hit_initial_speed: batted ball speed at contact, in mph

hit_horizontal_angle: spray angle, in degrees, with the convention that 1B line is 45 deg, 2B 90 deg, 3B line is 135 deg.

hit_vertical_angle: vertical launch angle, in degrees

hit_vx0, hit_vy0, hit_vz0: initial x,y,z components of batted ball speed, in ft/s (you can check that this is consistent with hit_initial_speed)

hit_x0,hit_y0,hit_z0: location of ball x,y,z at contact, in feet. hit_y0 (called hit_y in several posts) is the quantity currently under discussion.

hit_ave_lop: you can ignore this. It is some measure of how well the constant-vel curve fits the data (smaller is better).

Most of the remainder of the fields are pitchf/x quantities, which should already be familiar to you. Given the 9P fit to the pitchf/x trajectory and the initial hitf/x trajectory, you should be able to produce a plot showing the full pitch trajectory and the initial part of the batted ball trajectory.

Alan, e-mailed you with some thoughts on bat f/x.

Don’t see a real reason to post it here – briefly, it’s just about capturing the bat itself – because you’re the only one to reply on it anyway.

But I’ll answer you second question right here as well – just for fun

> And while you’re at it, tell us how someone from Moscow got interested in baseball.

Mostly a typical case of “seeing the greener grass where we ain’t” – it was back in the 80s.

But now, having watched enough of Russian baseball, turning to cricket instead.

While it was commented on that the FB avg dropped sooner than the offspeed avg, the offspeed avg drops suddenly off a cliff at 4’ while FB is still solid to 4.5.

If there is not a measurement problem it may mean that extra late contact on an offspeed pitch is hit softly and the late break makes it harder to hit solidly, while late FB contact still produces some off-field liners. They both have a bump at 4.5 possibly because so few are hit there that the fielders are not in a position for it. After that, outs in foul territory without any credit for fouls that drop probably cause the steep drop.

Great stuff, Jonathan!

Jonathan – I am afraid that you are misinterpreting the Hit f/x data. The Hit_Y values are the distance toward the pitcher from the front of the plate where contact was made. Almost all balls are hit BEFORE they reach home plate, not after.

I also don’t think that Sportvision has been able to resolve the point of contact very well. I mentioned this to Marv White at the recent summit. Since they don’t have the full 9 parameters (they don’t have enough information to estimate accelerations) on the hit ball it is almost impossible to extrapolate back toward the point of contact and get an accurate measurement. When I worked with the actual Pitch f/x-Hit f/x camera images last year I was not able to locate a precise point of contact until I estimated the balls landing location from the commercial video feed and calculated the acceleration values for the hit ball. Then I was able to resolve the pitch path with the hit ball path, usually to within 1/2” and .001 second in time. I did this for about 10 hit balls and every one was hit before it reached the front of the plate.

Just want to thank Professor Nathan for representing the University well.

Go Illini

This analysis is getting very interesting. I suspect the key to how early a ball is struck is not the pitch type but the location. Offspeed pitches are more likely to be outside, so they must be struck later, otherwise the batter will miss or hit the pitch off the end of the bat. At he same time the hitter has to get the bat around more quickly and hit the ball further in front to get around on an inside fastball.

Doing this same analysis but breaking the hitting zone into 9 quadrants might be very informative.