brooksbaseball

Graphing the Future

2009-07-12T19:33:00.001-05:00

The data is hosted in a different place, so a slightly updated version of the PitchFX tool (with Futures Game data) can be found here:

http://brooksbaseball.net/pfx/indexfut.php

Enjoy!

It's true...

2009-07-03T09:33:00.002-05:00

It's true, I'm terrible at updating my blog.

A few quick notes:
1) I will be at the PitchFX summit next week! It ought to be a great time, and I'm looking forward to meeting many people who I've emailed with over the last few years.
2) You can now search brooksbaseball.net by using baseball-reference.com. Simply select a pitcher, select "game logs", and then the links in the "Pit" (pitches) column will link directly to the PitchFX data for that day. Makes searching easier!

Opening Day

2009-04-06T11:20:00.003-05:00

Welcome to the 2009 season!

This is technically the first opening day for brooksbaseball.net. I hope you have fun accessing all the PitchFX data for the 09 season.

If you notice a bug, or have a request for some new feature, don't hesitate to send me an email: dan (at) brooksbaseball dot net.

The King Felix Debate

2009-03-09T01:56:00.002-05:00

If you happened to catch the USA/Venezuela game last night, you were treated to some great play-by-play by one of our Sox radio broadcasters (Dave O'Brien) and some of the worst color commentary ever recorded by Rick Sutcliffe.

But, in a stunning, shining moment of clarity, both Dave and Rick managed to have 5 minutes of un-forced conversation that approached the issue of King Felix Hernandez, who pitched admirably and effectively against the Italian team, coming out of the pen. The point of all of this talk was simply "why?". Why was King Felix, who you'd figure to be one of the real weapons on the Venezuelan squad, coming out of the pen to pitch against Italy?

And, fair enough, good question. Felix pitched 4 innings of 1-hit ball, posting 4Ks. Why waste this kind of pitching on an undermanned Italian squad when you know you've got to face the US if you win? Why burn King Felix if you don't need to?

Well, I'm not sure. But here are two possibilities. The first is that, actually, at the time the Venezuelans brought him into the game, the score was tied, 0-0, on account of Mark DiFelice and his 10,000 cutters (A thought: Does Mark DiFelice extend this strategy into other games as well, throwing only "Scissors" in Rock/Paper/Scissors?). The second is that Felix's effectiveness really masked the fact he isn't really up to MLB speed quite yet.

If you don't know anything about him, let me briefly describe Felix Hernandez. He throws hard. Very hard. A lot hard. When he can locate and he isn't acting like the Mariner version of Julian Tavarez, he has disturbingly good stuff. For example, in a lame attempt to correct for park effects (who knows yet if the cameras are similar this year, we'll just have to guess at it), here's a Felix start from last year at Toronto, on June 11th:

Yeah, that's right. 97.4mph max on that fastball. He can bring the heat, and sustain the heat, the entire game.

Here's the same kind of plot for his four innings in the WBC:

Maybe I'm just reading too much into things, but that certainly looks different. In this one, he's all amped up for the first inning, hitting almost 96, and then he hits a wall. A big wall. Next inning he's down around 94. Then he's down around 93. And by the time the fourth inning rolls around, he's hovering at 92.

This isn't to say anything bad about Felix - he's in a version of extended spring training just like everyone else is, and he isn't tuned up yet. I don't think this is characteristic of the Felix Hernandez we're going to see in the MLB. But I think it certainly might account for why Venezuela wasn't saving him up for the US team, which is laden with bona fide major league sluggers. Felix might just not be ready for that kind of intense competition yet, at least not at the level that Dave and Rick were expecting.

Mark DiFelice: The Italian Mo

2009-03-08T12:51:00.003-05:00

So, the Italian team had a relatively unknown pitcher, Mark DiFelice, start their WBC Game. He pitched four innings, didn't give up a run, and generally did a good job. So why a blog post about him?

Well, Mark DiFelice managed to go 4 innings, throw 45 pitches, and throw only 4 fastballs and 2 curveballs. It wouldn't be so odd if I told you that the 39 other pitches were something like knuckleballs - maybe he's just the Italian Tim Wakefield. Or, maybe the 39 other pitches were a heavy sinker, which a sinkerballer could conceivably throw that often in a short start.

No, Mark DiFelice, starting pitcher for Italy, managed to throw 45 pitches, 4 fastballs, 2 curveballs, and 39 cutters/sliders. 39 Cutters!

The graph above shows all of Mark's pitches, with the horizontal movement of the pitch due to the spin of the ball on the X (horizontal) axis, and the speed of the pitch on the Y (vertical) axis.

Despite the fact that Mark apparently throws at three decidedly non-fast speeds ("slow", "pretty slow", and "very slow"), the "Green" is clearly his "fastball". It's faster than the rest, and it moves like a fastball from a RHP would move. The "Blue" is clearly his curveball. It moves like an RHP's curveball and is thrown decidedly slower than the rest.

And then... Purple. An entire mess of one pitch: the cutter. Never mind the fact that a lot of these cutters don't really even move that much, he continued to throw them. A lot. Ignore the few mislabeled pitches on the next plot (Gameday algorithm continues to frustrate...), but look, in the fourth inning, Mark DiFelice manages to throw ALL cutters. Not a single other pitch.

Impressive.

And, just in case you were wondering: "didn't team Venezeula figure this out?" The answer, almost certainly, is yes. Despite throwing 4IP of 0ER ball, he gave up two hits and a few very loud outs (including a sensational diving play and a runner gunned down trying to take second).

How can we tell Venezuela was looking cutter? Well, only one non-cutter pitch was even swung at all game, and every at-bat ended on a cutter:

World Baseball Classic PitchFX!

2009-02-25T22:59:00.006-06:00

I shot off a quick email to the man in charge at MLB Stats, Cory Schwartz, and got the word that MLB will be tracking the WBC with PitchFX!

So, I have added a new tool at the right that is specific for the 2009 World Baseball Classic.

Only games in MLB Venues will be tracked by the system (because that's where they've got the cameras installed), but this should give us all the first look (and most detailed picture) of some of the hottest international pitching prospects and stars. It will also effectively give us spring training data on a number of MLB pitchers that are taking part in the Classic.

I will be closely monitoring this new page/system for possible bugs. If you notice one, please contact me immediately.

PitchFX Data for Rays/Phillies

2008-10-29T20:00:00.002-05:00

The PitchFX Data for the Rays/Phillies Endless Game 5 is still posted under 10/27/08.

The strikezone maps include all of the calls from Monday.

Just a heads up for anyone planning on using any of it.

Someday We'll Watch It, The DiceK Correction

2008-10-11T00:45:00.004-05:00

The scariest part about Daisuke is that he consistently pitches this way. We're talking about a guy who pitched 7 shutout innings, striking out 9, while walking 4. Walking the bases loaded in the first inning and managing to escape. Stranding a runner on third with 0 outs. It shouldn't be possible to walk the bases loaded in the first and pitch a no-hitter through 6. But for whatever it's worth: it is.

For those of you still wondering what the dots mean, here is a plot of all the pitches Daisuke threw with annotations:

Green: Changeup
Blue: Slurve (Slider/Curveball)
Orange: 2-Seam Fastball
Red: 4-Seam Fastball
Purple: Cutter

While that plot is nice for looking at all of the pitches, this one might be a bit nicer for breaking down what actually worked for DiceK over the course of the game:

Here we can see his 9(!) strikeouts in black: 2 changeup, 5 fastball, 2 cutter (although, one of those is a very slow cutter, more like a true slider). You can also see that a bunch of his trouble came from the Slurve - 3 walks and 2 singles off of it.

That's the initial breakdown, but I'll have more on this start tomorrow.

A link to Daisuke's performance in Game 1 of the ALCS can be found here.

Establishing the Fastball

2008-09-18T19:11:00.010-05:00

Today's post concerns an interesting trend in pitch distribution: "establishing the fastball". You often hear about this strategy from broadcasters, but essentially the idea is to establish that you will throw your fastball early in the game (perhaps to set hitters up in terms of timing or trajectory), and then work off of that set up to effectively mix in your other pitches.

My disclaimer here is first that I'm just looking at a few pitchers on one pitching staff that's been caught by one catcher the majority of the season. Can we capture this trend and gain some insight into pitch distribution?

Data for this project has been gathered for from every game in the 2008 season. Pitch identification has been done by the MLB Gameday Algorithm, which isn't great, but we're also looking for something that's supposed to be a very large effect, so it's possible that this won't prevent us from finding it.

To begin, I'll look at Daisuke Matsuzaka, who's fun to write about and sometimes aggrevating to watch pitch.

We'll start by dividing up "Pitch Counts" into a series of 20-pitch groupings and then looking at the relative distribution of pitch types within those groups.

You can pretty clearly see that the proportion of fastballs (plotted here in green) decreases steadily through the first 40 or so pitches, and the proportion of sliders generally increases; both of these values then level off into a relatively stable proportion for the rest of the start. Meanwhile, the relative distribution of other pitches stays essentially unchanged.

I've labeled these things here:

So, it's fairly easy to see that the pitch distribution for Daisuke, averaged across all of the games in the season, is at first biased heavily (over 60%) in favor of throwing a fastball, which steadily levels off to around 50%, and the proportion of sliders increases from less than 20% to greater than 30%.

These effects are even more dramatic when looking at what Daisuke throws on an 0-0 count:

Early in the game you have an over 80% likelihood of seeing a first pitch fastball, which decreases to (and levels off at) just above 50%. Likewise, seeing a slider on the first pitch starts at around 5% and increases to (and stablizes at) 25-30% likelihood.

Some pitch distributions, however, remain relatively unchanged. For example, if you get into a full count, in most game situations the pitch distribution is relatively equivalent (although, caveat emptor, there's a lot of variability due to small sample size):

On the other hand - and I was quite surprised by this - if you're in any 2-strike count, the pitch distribution still follows the game plan. So, early in the game he's not trying to put you away with the slider.

What does all this mean? Well, I think one thing it shows is that although a pitcher (like Daisuke) might have a relatively stable pitch distribution after some reasonable point in the game, using the aggregate pitch distribution numbers that are shown on some websites might not be such a good idea. In order to get a good idea of the real proportion of pitches, it's best to first adjust for establishing the fastball.

Let's take a look at other Red Sox pitchers to see if we can get a better handle on this. Here's Jon Lester, showing a similar trend, but with a quicker "fastball establishment":

But, the trend becomes more like Daisuke's trend if you look at only 0-0 counts:

So, again, over the first 40 pitches, the fastball is "established" and continues to lose likelihood of being thrown, while the curve and slider/cutter increase. Of course, Jon Lester throws a much larger proportion of fastballs, but the trend is still relatively apparent.

For Josh Beckett, on the other hand, we see a much more gradual decline in the relative proportion of fastballs thrown over the course of the game:

But again, this trend (40 pitches of fastball establishment) is recaptured quite nicely if we look at only 0-0 counts:

So, what have we learned? Well, it's fairly easy to see "Establishing the Fastball" as a gameplan strategy by the Red Sox staff as a whole (and not just Daisuke), and it appears to take about 40 pitches before Sox pitchers stablize into their gameplan. After that, the relative proportion of pitches remains relatively unchanged.

Psychadelic Charts for the Common Man

2008-09-18T11:21:00.004-05:00

There is a thread at Sons of Sam Horn entirely devoted to talking about Jon Lester.

This is odd, because it is against their policy of megathreads, but if you're looking for some retrospective analysis of Jon's stuff over the course of the season, you could just compile that and post it on the web. A friend and I are planning to do that, just after I manage to crawl out from the mound of papers currently stacked up on my desk (note: if my advisor is reading this, it's my first blog post in a month, really).

Anyway, a recent discussion was about whether or not Jon was getting stronger as the season went on. Fair question. Here's my post on the issue:

The first figure is simply a graph showing his average fastball velocity and average curveball velocity over the course of the season, for all the games except Philly (their system is pretty clearly broken).

There's a steady rise in both Fastball and Curveball velocity, suggesting that the rise in fastball velocity isn't due to a change in fastball type distribution.

The other graph is a bit more psychadelic, so here is the explanation. I realize that averaging across pitch types can be bad because of the known issues in trying to identify pitches, so I took one of his first starts from each month of the season and plotted each pitch on the same Horizontal Movement X Speed axis. To account for the fact that different systems may have slightly different speed readings I used his first start at Fenway for each month.

Warmer colors (yellows and oranges) are early in the season, cooler colors (blues and purples) are later in the season.

I think it's pretty clear that the velocity increase is not due to pitch misidentification or something else, as the velocity increase is really across the board. Cooler colors are clearly above warmer colors in each cluster in this figure.

For those struggling to identify these pitches I have labeled them.

1. Bottom Left (Ugly Dark Purple) - Curveball
2. Middle Left (Light Blue) - Cutter (a few sliders towards the curveball side)
3. Top Middle (Green) - 4Seam Fastball [a few changeups are thrown with this kind of movement and erroneously grouped in the circle, they are the very slow 4seam fastballs, sorry]
4. Top Right (Light Purple) - 2Seam Fastball
5. Bottom Right (Orange) - Changeup

Around the Web II: Updates

2008-08-11T14:18:00.003-05:00

Some updates from around the internet:

Oliver Perez seems like quite the popular figure, as he is the focus of an update on New York Mets Daily devoted to a PitchFX analysis of what's worked since his mechanics change (http://newyorkmetsdaily.com/?p=209) by Brian Joura. Perez is also one of the players examined in a Newspaper article (http://www.nysun.com/sports/is-the-metamorphosis-of-perez-pelfrey-for-real/82691/) by Tim Marchman.

Jeff Samardzija has recently been the subject of an article (http://bleacherreport.com/articles/41242-jeff-samardzija-pitchfx) by Kanka, who looks at some of the pitches he threw while making his major league debut.

Jon Lester is pretty damn good this year, and has been inspiring all sorts of media. Peter Bendix takes a look at Lester's progression in an article (http://www.beyondtheboxscore.com/2008/8/4/585966/jon-lester-the-best-pitche) on his blog.

As always, if you're planning on writing an article about pitching or a pitcher and either have questions about using PitchFX data or need specific graphs or plots made, send me an email at dan@brooksbaseball.net.

One Great Pitch, One Terrible Pitch

2008-08-10T17:54:00.005-05:00

Clay Buchholz is a bit of an enigma. He's a guy who pitches a no-hitter last year and has, supposedly, incredible stuff... at the same time he can't seem to put it together to get out Major League hitters.

Over his last few starts, he's been shelled. In the 5 starts before the one this afternoon, he gave up 5 HR and was good for a 6.59 ERA, with 15BB and only 24K. Today, he gave up 3 more HR, 5 more runs, and 7 hits - all in 3 innings pitched.

What's working for Clay, and what's not? To answer this question, we'll turn to a specific PitchFX plot from his most recent start.

An interesting feature of Clay's changeup is that it doesn't move the same as the change that's normally seen in the majors. Most changeups fade away from LHH (or in to RHH). This makes them fairly easy to identify on a Horizontal Movement X Speed plot because they usually lie opposite of the Curveball. Clay's pitches are a bit more difficult to identify using only horizontal movement because his changeup fades in the other direction. This also leads to the Gameday system (and, somewhat amusingly, the Fenway Park scoreboard) to call most of his changeups as sliders:

You can see it's difficult to tell apart his offspeed pitches which all cluster in the righthand side of the graph.

On the other hand, it's easy to tell them apart if we plot Speed x Spin Direction Angle. We can calculate the spin direction by using a formula from Alan Nathan to recreate the way the ball was spinning on its way to the plate (which subsequently results in most of the movement on the ball):

There. That's easier. On the plot above, there are 4 distinct pitches: the fastball (green halo), the changeup (red), the slider (blue), and the curveball (orange). You can see by looking at the individual dots that most of the hits (light blue) come on the fastball and the curveball. If you watched the game, it was pretty clear that his fastball command was bad and hitters were really teeing off on it.

On the other hand, look at the changeup (red). 13 changeups. 12 strikes, 7 swinging. And only 1 ball. That's basically as effective as an offspeed pitch is possibly going to be in a single game. Hitters had no idea what to do with it.

It's pretty hard to argue with the results when your pitcher is getting lit up and he can't command his fastball. But he still has that dominant pitch.

The strange thing about this dominant pitch is that while it was leaned on very heavily early in the season, he has stopped throwing it as much as he once was. At his start in Tampa on 4/26, where he went 9 innings (eventually taking the loss on a 2-run bomb), he threw 40 changeups in 113 pitches.

It's just as dominant then as it was today, but then it was 35% of his pitches:

Today, he threw 13 changeups in 74 pitches, which is 17.5% - or exactly half of what he used in the dominant start at Tampa. Admittedly, part of this is probably dictated by the fact that he was pulled early in the game, but through his first 74 pitches at Tampa he had 33 changeups - a 45% clip. On 8/4 he threw 15 changeups in 93 pitches. At LAA he threw 18 in 90. At Tampa he threw 41 through 113.

Every time you don't call for the changeup, you're replacing a great pitch with a mediocre fastball. And since his return, he's been much more fastball-heavy, even though his changeup hasn't lost any of it's effectiveness.

Is the Clay we saw today the same pitcher as early in the season at Tampa or last year during the No-no? Perhaps not. The Tampa start might have been a fluke. Or, we could be seeing the result of strange game planning, pitch selection, and bad execution from Varitek and Buchholz. For whatever reason, he's gone away from the dominant changeup, and it isn't working.

A Defensive Oddity

2008-08-10T16:10:00.004-05:00

If you happened to be watching the Rays @ Mariners play into the 10th inning on Saturday, you saw one of the strangest defensive alignments in Baseball: their center fielder was playing on second base.

It wasn't just that the center fielder was playing second base. He could have been, because their center fielder is BJ Upton, who played second for them up to June 7th of last year. He was playing literally on the base, and there were only two outfielders.

Here's an unlabeled screengrab from the game:

And here's a labeled one:

The arrows point to Rays fielders. Note the 5 in the infield.

Here's the situation: it was the bottom of the 10th, bases loaded, and only one out. Rays manager Joe Maddon knew that a fly ball to the outfield was easily going to score the runner from third, and so he started to pull Upton in. He motioned in. And then in. And then in some more. And finally, Upton was playing on second.

It should be noted that at the time, Chad Bradford (a submariner / groundball specialist) was pitching, and so the extra infielder may have helped. It didn't turn out to figure in the play, as Bradford got a sharp grounder to third which resulted in a 5-2-3 double play. But still, as odd a defensive alignment as you might see.

In this shot, you can pretty clearly see Upton holding the runner on second:

Just thought this was one oddity in a season filled with strange plays that was worth mentioning.

Site Feed

2008-08-10T12:13:00.002-05:00

BrooksBaseball.net now has a site feed where you can subscribe to all the latest updates, articles, and random bloggings:

http://feeds.feedburner.com/Brooksbaseball

Enjoy!

Quick Pitch: Getting a Rise out of Good Pitching

2008-08-09T21:20:00.010-05:00

Besides watching Daisuke's best outing of the season, if you were paying close attention, you noticed something strange: his velocity steadily and fairly dramatically improved through the first few innings for both his fastball and his slider/curve.

I thought I'd point it out because it is a fairly drastic example of a pitcher getting stronger as the game went on (at least, up until the 6th).

Note
Hat tip to Alex for also noticing this during the game.

You can view all of Daisuke's data from his 8/9/08 start against the White Sox using the PitchFX tool by clicking here.

Quick Pitch: Kennedy Struggles Again

2008-08-09T12:42:00.008-05:00

First, let's dispatch of the obvious elephant in the room: I am a Red Sox fan.

Now that we've got that settled, let's take a look at Ian Kennedy's latest stint in the bigs. If you haven't heard of him, Ian Kennedy is one of the Yankees' blue-chip pitching prospects that was flagged as "untouchable" in the Johan Santana trade talks last year.

He struggled mightily in the majors early this year, posting a relatively painful 7.41 ERA in 37.2 innings pitched over 9 appearances at the beginning of the season. He was then sent back to the minors, and recently recalled with the devestating news that Joba Chamberlain does not, in fact, have a titanium shoulder that can withstand 60+ pitches of 98+ mph every start.

Kennedy started 8/8 at Anaheim, where he couldn't get more than 2+ innings and gave up 5 ER and 9 hits. He had a single strikeout and one walk over 61 pitches.

Kennedy is essentially a four-pitch pitcher that throws a fastball, a changeup, a slider, and a fancy curveball. They can all be seen fairly easily by plotting horizontal movement x speed:

In the above plot, the different pitch types are outlined to make it easier to explain. The red group is his fastball, which clocked at 88-90 mph (topping at 90.6). He actually gets fairly decent movement on the pitch, but he still gave up 3 hits on it (the light blue squares). His lone strikeout came on a fastball, but so did his only walk.

The dark blue group is the changeup (fades away to LHH and thrown slower than his fastball). This was actually a moderately effective pitch compared to his other breaking pitches, but again he gives up 2 hits from it. The yellow group, his slider, was woefully ineffective: of 8 sliders thrown, 1 is taken for a called strike and the rest are either balls or hits. His curve was equally ineffective, with one pitch of three taken for a ball and the others going for hits.

Obviously it's easy to find flaws in a guy that only goes for 2+IP and gives up 5 runs. But I think it's clear that at this point, his breaking pitches aren't effective enough to get through a good lineup. Granted, the Angels are one of the best teams in the game, so this is perhaps unfair.

If you look at the location of the hits, you see that he's leaving far too many balls out over the plate. Of course, he also gives up two singles on two very low pitches, so part of this was likely some bad luck or lack of deception as well:

Overall, an ineffective start, but one we can explain relatively easily. Bad location on his pitches combined with an ineffective slider and curve simply isn't going to get it done against the Angels.

Notes:
You can view all of the data for Ian Kennedy's start at Anaheim by using the PitchFX tool. Quick link: Kennedy @ Anaheim 8/8/08

Looking Knuckleball, Part 1

2008-08-03T01:13:00.021-05:00

The Last Surviving Knuckleballer
A few days ago I posted about what to expect from Daisuke. Next, I'd like to take a more in-depth look at Tim Wakefield, the last surviving knuckleballer (I know, R.A. Dickey pitches for the Mariners, and Haeger is still floating around the White Sox organization, and our very own Charlie Zink is dominating the minors, but doesn't it sound more impressive to be the last of one's kind?).

Tim Wakefield, the last surviving Knuckleballer.

The Repertoire
Everyone knows Wakefield throws the Knuckler - he also throws a fastball (at a blazing 73 mph) and one of the slowest curves still thrown (57-59 mph). Depending on the count, your chances of seeing a curve and fastball vary quite a bit: if you manage to get ahead of Wakefield, you're likely to see the fastball; if you manage to fall behind, you may see the curve. This isn't rocket science: the fastball is thrown when Tim needs a strike, and the curve is thrown when Tim has the hitter set up and expecting knuckleball, only to see the slow hammer dive sharply into the strikezone. But, for the most part, Tim isn't up there dazzling with excellent fastball command or flashing his knee-buckling bender. He's up there to do one thing, and do one thing well: throw a 66-68mph butterfly, pitch after pitch.

Breaking Down a Single AB
Before looking at a large plot of dozens of knuckleballs it helps to acquaint oneself with the actual movement on the pitch. Most people watch it on TV and laugh when guys swing and miss because it doesn't look that hard to hit. The color commentator, especially if he's an ex-player, is trying to convince you (probably with little success) that it really is hard to hit a knuckleball, all while Wakefield appears to effortlessly get up there and deliver the same 68mph junk ball. So, rather than focus on the aggregate (that will be Part 2), we're going to focus on a single At-Bat - a strikeout of Jack Cust on 8/1/08.

To try to give you a sense of exactly what the ball is doing, here's an overhead view during a single at-bat during which Wakefield threw only Knuckleballs:

That's right. All four of those pitches (resulting in a strikeout) were knuckleballs, even though one of them has a really heavy outside tailing movement, two curve to the right, and one sails far inside (Cust is a LHH).

The Psychology of Hitting the Knuckleball
Let's take a minute to break down the psychology of actually hitting the Knuckleball. Remember that for the most part, although pitchers try to be deceptive and generally throw things that are tough for hitters to tell apart, their pitches, if discriminable, follow relatively predictable trajectories. If you could tell with 100% fidelity that a four-seam fastball was a four-seam fastball and a curve was a curve, you would have a pretty good idea of when and where to swing the bat. Second, most hitters get up there and "guess". They get in a particular count, have received extensive scouting information about a pitcher, have seen a number of pitches, and sometimes have good information about what they're likely to see (sometimes not, if the batter is badly fooled).

Right away, one of the large advantages of throwing a knuckleball is that neither of these strategies work. For the most part, except in rare situations, hitters know they're going to get the knuckleball, so "guessing right" doesn't help. Second, when they first recognize the pitch as a knuckleball, this gives them no particular advantage in hitting it, because they have no good information about where the thing is going to go.

Another way of saying it: hitting the knuckleball is hard because it's random. Suppose your swing needs to start about .25 seconds before the pitch reaches the plate, close to the ~225-235ms time estimated by Robert Adair in The Physics of Baseball (though this probably needs to be updated). Even though it might look relatively slow on TV, the average time-to-plate of a Tim Wakefield Knuckler is about 500 ms. That means that all the information you get about a relatively random trajectory is in the first half of it's flight to the plate, and you've just got to make your best educated guess at what the second half is going to be.

Let's take a look at that information. I've broken down the knucklers from this at-bat into two groups: the last 250ms of each pitch's trajectory, which we know that the hitter cannot see before deciding to swing, and the the time before that, which we approximate the hitter can see. For example, here is a look at those same knuckleballs in a similar format:

This is straight down on top of the pitch trajectory, just like the plot a few paragraphs ago. Okay, so we might be able to connect those trajectories to their eventual locations, but especially with the two on the left (both of which were swinging strikes), we've got very little information to work with. Remember, up to this point, the hitter has no idea how much more each of those pitches is going to move. Here's how much they do move after the hitter's last look at the pitch before deciding to swing:

From this view alone, we can see that a large majority of the knuckleball's movement actually occurs AFTER the hitter has decided to swing (or not) at the pitch. Essentially, if you're swinging at a knuckleball with a lot of movement, you're really just making a guess as to where it's going to end up.

The same is true of vertical movement. Here's the "before" shot of Knuckleballs from a side view:

Which one of these will eventually be the highest ball? It might not be the one you think:

These knucklers actually have fairly uniform movement in the vertical axis, although one sails up. Still, compare the two plots and you see that by far the most movement occurs after the hitter's decision point.

But still, you might think: "this really isn't that hard." Maybe you're good at matching trajectories and it isn't all that difficult to tell these ones apart. How does Jack Cust fail so badly at this task?

Remember that Jack is looking at the pitch from the Batter's Box, and so part of his ability to see trajectory information is further obscured by his angle relative to the release. Even though Wakefield is a RHP and Cust is LHH, there's still a relative distortion on the trajectory information available to the hitter compared to looking at the pitch straight along one of the axes of movement.

And the post-perceptual trajectory of these pitches:

I hope I've convinced you that through looking at only a single at-bat, it's really not so easy to hit the Knuckleball. This is mainly due to two factors: the fact that the movement of the knuckleball is highly random, and the fact that most of this movement occurs after the batter has made his decision to swing. In Part 2, we'll look at the aggregate movement of the Knuckleball. For a preview, here are the aggregate plots from the 8/1 start:

Notes
You can comb through all the data from Tim Wakefield's start on 8/1 by using my PitchFX Tool. Click here to go directly to this game's data.

Dan Fox wrote an article about aggregate data on the Knuckleball for an issue of Baseball Prospectus.

Matchups

2008-08-01T14:14:00.003-05:00

I added a "Matchups" selector to the PitchFX tool. If you select a particular at-bat, all of the other plots will be restricted to pitches that occured in that particular at-bat.

ABs are noted by Inning, Player, and Result.

Enjoy!

Tracking the Knuckleball

2008-07-31T13:44:00.004-05:00

First off, I want to acknowledge the very fine people at Sportvision and at MLBAM. On the very short list of companies/organizations involved in professional sports that I have ever had contact with, they are very clearly at the top.

A few weeks ago, a few emails began circulating among some of us who do PitchFX tracking and analysis that something about Tim Wakefield's knuckleball was creating some odd problems in the data. We noticed this because it appeared as though pitches his pitches were being dropped and others were being erroneously attributed to him. Several times a game, it would appear as if Tim had suddenly decided to start throwing 95 or pitching sidearm lefty.

After a few emails had been circulated, the fine folks at MLBAM and Sportvision got involved. They were pointed towards games and pitches where the problem was occuring, and they started to work on the problem. Only a few days later, we received a reply that said they believed they had identified the source of the issue, which had nothing to do with actually tracking the knuckleball. Instead, it had to do with the odd peculiarity that switch hitters who would normally bat lefty against a RHP were batting righty against Wakefield, and this was causing some issues in the data output. They are currently investigating whether this bug also occurs on inning-ending baserunning plays.

I've just gotten another email from MLB a few days ago (and permission to share it with everyone) that said they were in the process of rebuilding the data from their tracking system and everything should be corrected in a short amount of time.

So, by way of MLB and Sportvision, the "knuckleball problem" (which was really never a knuckleball problem anyway) is essentially solved. The diligence with which they listened to the problem in the data, identified the cause, and quickly rolled out a solution - while keeping us all involved - is really admirable.

If you think you might know of an instance where a batter who normally bats RH against a LHP or normally bats LH against a RHP and has for some reason switched to the other side of the plate, let me know and I will pass it along.

Note
I left out the names of specific individuals involved in all of this because I didn't want to step on anyone's toes, just let everyone know that the problem is solved and the guys who run the system are good people.

Leveraging the Count

2008-07-29T09:43:00.018-05:00

Easily Testable Things Make Me Happy
"In a hitter's count... very seldom will he throw him a fastball and give in". - Jerry Remy, 7/28/08, on Daisuke Matsuzaka.

Every once in a while I'll be watching a game and the announcer says something so immediately testable I'll start checking mid-inning. Today's version concerns Jerry Remy's apparent belief that, if the count is a hitters count, Matsuzaka refuses to back down and will continue to throw his breaking pitches, presumably to keep challenging the hitter or to keep him off balance.

How to test it? Pretty simple, right? Just find the hitter's counts, and check to see what he throws.

Identifying Hitters Counts
If we wanted to, we could simply guess at which counts favor hitters. 3-0, of course; 2-1, and 3-1 are the most favorable hitters counts. But how favorable are they? Enter Tom Tango and Mike Fast, who compute values called Linear Weights (LWTS) for each particular count. Linear Weights are like relative values for each of the counts - counts with positive Linear Weights have a greater likelihood of leading to runs, and counts with negative Linear Weights have a lower likelihood of leading to runs. So, hitters counts are those in which the LWTS are greater than 0; pitcher's counts are those in which the LWTS are less than 0. LWTS values used are at the bottom of this post.

Testing the Theory
Now that we've got LWTS for various counts, we can test Remy's Theory.

We'll start with last night's (7/28) game:

This graph shows all of the pitches (different labels) he threw in various counts. FA is Fastballs, FC is cut fastball, the rest are breaking pitches. To make sure none are mislabeled, the Y axis here is the speed of the pitch (since sometimes cutters can be mislabeled as sliders and vice versa). The X axis here is the LWTS of the count - so, farther to the right are "hitter's counts", and farther to the left are "pitcher's counts".

What do we find? Remy's wrong! For this game, if the hitter is in a hitter's count, he sees a fastball from Daisuke, without fail, every time.

Well, maybe we're just cherry picking and it happened to be the one start that didn't match up with the theory. After all, he said this early in the game, so maybe it was a true observation that just didn't hold true for one particular game, right?

Let's go back a start to 7/22 @ Seattle:

Hey! That looks like the same picture! No breaking balls in hitter's counts again. Okay, maybe that start is just a bad example.

Back in time again, this time to 7/13 vs. Baltimore:

Still no luck. Back in time again, to 7/7, vs. Minnesota:

There still hasn't been a single breaking pitch thrown in a hitter's count!

Back AGAIN, to 7/2 vs. Tampa:

Hey! Finally! One slider thrown in a 2-0 count (and one mislabeled cutter thrown at 3-1)!

Conclusions
For the last month, Daisuke's pitch selection has been like clockwork: if you get into a hitters count, he throws you a fastball. We should probably revise the quote to say something like:
"In a hitter's count... look fastball. It's all you're evah gonna see.". - Jerry Remy, 7/28/08, on Daisuke Matsuzaka.

Notes
If you want to generate the above plots for any pitcher, just go to my PitchFX tool and select "Leverage Based" as the plot type.

Hat tip to Jon for pointing out the typo: CU is Curveball, not Cut Fastball. FC is Cut Fastball.

LWTS Values for Various Counts
LWTS values, along with heat plots by Jon Hale showing what pitchers throw in various counts, can be found here: http://tangotiger.net/halejon/allcounts.html

PitchFX Tool Updates and Comments

2008-07-28T22:34:00.008-05:00

So, after watching DiceK pitch pretty well for a few innings and then get absolutely lit up in another, I've taken the time to write a short post on updates to my PitchFX tool (which has undergone some extensive updating in the last few days).

First, you'll notice the color scheme has changed, so it no longer looks like it was written in 1994 to be viewed on Lynx. Nothing substantive should have changed when I updated the colors - if you notice something broken, please contact me.

Beyond that, "new" features:

Four selections are now available for plot styles. Normal plots will simply show balls, strikes, and balls in play. Pitch Type plots will mark which pitches are which (Fastball, curveball, slider, etc), as identified by the Gameday algorithm. Extra detailed plots will show each pitch broken down into game events (balls, foul balls, ball in play, fly out, etc). Result plots will plot only pitches that end at-bats and show the results of the at-bat. A fifth selector for Experimental Sabermetrics (such as Linear Weights for each count, "Nibble" quotient, "Wildness", etc.) is the way and will be uploaded shortly.

New fancy visualizations can be obtained by clicking on the "Pitch Types" plot type. These will plot the average pitch trajectories for each pitch type as well as all pitch trajectories when viewed from the right or left batter's box.

Strikezone maps will now open in their own separate page and give complete, LHH, and RHH maps for the game. These are all normalized to the height of the batter and will only include calls made by the Home Plate Ump to give you a feel for where the strikezone is. These can be very useful after a game to determine if your pitcher was getting squeezed (or getting generous edges) and may have had an impact on his performance; it can also be nice to check during a game to view the strikezones as they "emerge".

An Info Table now appears above all of the charts that gives up to date information on average and maximum velocity for each pitch type, as well as velocity histograms for the selected pitches, which can give you a good estimate of the average of different types of fastballs for guys that throw both 2 and 4seamers.

Anyway, there are lots of other little updates that have been added along the way. Nevertheless, if you have any questions or comments, please feel free to email at daniel-brooks@uiowa.edu or leave a comment on the blog and I'll take up any suggestions you might have. Feedback is always welcome!

When Les is More

2008-07-27T22:37:00.008-05:00

Jon Lester has suddenly - and by suddenly, I mean after a period of long Major League development - become a very good pitcher.

I think the hardest hit ball all night was a double by A-Rod that was barely fair, and that's it. Otherwise, and especially early in the game, his fastball was pure dominance, clocking in the mid 90s with a top of 95.7, which is harder than he's thrown all season.

Again, his cutter - circled above in blue - was very strong tonight. A number of swinging and called strikes and weak outs are quickly making this look like one of Jon Lester's better pitches after he really struggled to control it early in the season. His 4 Seamer - circled above in green - was also an incredibly effective pitch. The curve was sharp, and he picked up two Ks with it, but it also got hit for two singles and a double.

After being his go-to pitch in his 7/3 CGSO at the Stadium, his 2seamer - circled above in purple next to the 4seamer - wasn't well controlled and didn't seem to do much for him tonight. Two ground outs off of it, but lots of balls. And even though Orel Hershiser lauded Lester for "using all 5 pitches", he threw a grand total of ONE changeup the entire game (circled in pink).

Last but not least, since I would feel guilty about not using my newest toy, here's what it looks like to face Jon Lester as either a RHH or a LHH hitter.

As a RHH, you get a very apparent, sweeping breaking pitch and the tailing/cutting action of his different fastballs. All three of the hits on the curve were by RHH.

As a LHH, the curveball really vanishes into the rest of the pitch trajectories and it becomes much more difficult for hitters to quickly identify, especially when thrown low in the zone.

On average, I think this top-down view really gives you a good idea of how much his pitches are moving on their way to the plate, and how subtle and late the movement on the cutter (in orange) really is.

Lester has developed into a "suddenly" developed into a dominating pitcher that's very fun to watch, especially compared to early in the season where he couldn't buy a strike.

Visualizing the Platoon

2008-07-27T01:47:00.017-05:00

The Platoon Split
The "Platoon" split is one of the most interesting statistics in baseball. Simply put, it is the statistical split that documents the usual advantage a pitcher gains while pitching against hitters that hit from the same side as he throws.

The Platoon split has been discussed heavily is Sabermetrics; a section in The Book deals with its analysis and documentation, and an article by Dan Fox at Baseball Prospectus reveals the following:

"Platoon advantages exist and are significant, and that the difference is larger for left-handed hitters than for right-handers...The average left-handed hitter hit 34 points better against right-handers, but that the difference from the mean spans a range from 12 to 45 points (34 plus and minus the standard deviation of 22 points). For right-handed hitters the mean value for batting average is 18 points with an average spread of between 1 and 35 points. As you can see, there is a fairly large amount of variability for hitters from both sides."

But what causes the platoon split? Color commentators often describe hitters as being "uncomfortable" against pitchers with substantial platoon splits - because they "hide the ball well" or have "deceptive deliveries". I think focusing on the pitcher is important, but not nearly as important as focusing on the pitch.

A Matter of Perspective
Suppose we were to break down hitting into two parts: part perceptual and part motor. The perceptual part deals with the visual recognition of the pitch, and is thus affected by things such as the similarity of the various pitches thrown, the ability to quickly guess at the final location of a pitch on a particular trajectory, and the ability to judge the speed of the pitch. The motor part deals with the physical actions required to hit the ball, such as the action of swinging the bat, adjusting to the ball mid-swing, etc.

Presumably, the advantage for a Left Handed Hitter (LHH) against a Right Handed Pitcher (RHP) - or vice versa - is not a motor one. LHH do not get any advantages in terms of the amount of time it takes for the ball to reach the plate, nor would we say that they get any extra boost in power or quickness as a result of being on the opposite side of the pitcher. If anything, the ball is hidden from their view for a longer amount of time while most pitchers complete their delivery. How is that that LHH are better against RHP, and RHH are better against LHP?

Hitting a baseball starts with a complex visual discrimination task that literally takes place in the blink of an eye. Before he has time to think, the hitter must react to the pitch at hand - and do so with very little information about the pitch. The hitter sees only part of the trajectory of the ball clearly before "deciding" to swing.

Most of the time, when you see the trajectory of the ball, you view it from a very different angle than the hitter. Usually you see an offset pitcher's perspective of the pitch, over one of his shoulders; sometimes you get the "behind the plate" view that shows you essentially the Umpire's perspective. In recent years, some ballparks (especially Fenway Park) have adopted a much better viewpoint, from directly behind and over the head of the pitcher, which gives us a much clearer view of how pitches move as they approach the plate. But as viewers, we're fundamentally crippled by the fact we never get to see pitches as hitters do, which makes it very difficult to understand why pitches are so deceptive.

Enter PitchFX, which allows us to precisely capture the trajectory of each pitch. The PitchFX system accurately represents the pitch at every point along it's violent flight to the plate, and so, if we wanted, we should be able to do some tricky math to figure out what the trajectory looks like from various points on the field. This technique was first examined by Mike Fast, who, in addition to doing many other very clever things with PitchFX data, alerted us to the idea that views from the LHH batter's box might be different than from the RHH box, and also published a technique for applying the transformation. Specifically, concerning one pitcher-hitter matchup, he says, "The fastballs look almost straight to the right-handed hitter, and the breaking pitches have a noticeable hump; however, all the pitches appear to bend more sharply to the left-handed hitter, and the trajectories overlap more. If all left-handed pitchers' repertoires look somewhat like this, no wonder left-handed batters have a harder time with them than right-handed batters. It makes me eager to apply this technique to other pitchers."

The Data
So, now that we know how we got here, let's take a look at what we're dealing with. As a sample case, we're going to use Justin Masterson, an RHP for the Red Sox with a sizeable Platoon split and a strong sidearm delivery that primarily throws three pitches: a 2seam fastball, a 4seam fastball, and a slider (we will not deal with the occasional Masterson changeup since I will not graph plots with any thrown).

People like me always get accused of not actually showing the real behavior before we delve into the statistics, so let's take a look at Masterson on the mound, delivering a pitch. His sidearm release is very clearly shown.

If we abstract that back into data recorded by the PitchFX system, you will see that we get a relatively faithful and consistent representation of his release point.

So, Masterson is a heavy side-armer. PitchFX can see that. Things are working so far. Let's next take a look at one of the most traditional PitchFX graphs, which shows the Horizontal and Vertical movement on his Pitches. The Horizontal Axis here is horizontal movement from the Catcher's Perspective, due to the amount of spin that Masterson puts on the ball. In other words, this shows the amount of action that Masterson has put on the pitch. The same is true for the Vertical axis, which shows that most of Masterson's pitches have relatively low vertical movement (compared to other fastballs, this is why he is termed a "sinkerballer"). In this graph, we can easily identify three groups of dots: the cluster on the right is his slider, the cluster on the left can be split into his sinker and 4seam fastballs.

This gives us an idea of the kind of movement that Masterson is putting on his pitches: lots of horizontal movement with a very large difference between his sinker (which cuts in very heavily to RHH) and his slider (which cuts away very heavily), both of the pitches thrown with relative sink. We should also note the relative similarity in speed in these two pitches: the 2seamer (sinker) is thrown at about 90mph, the slider at about 84mph.

Many PitchFX breakdowns would stop here - we've documented the slider and the sinker, and how each break. But this PitchFX breakdown is intended to explain why Righties have so much harder a time against these two pitches than Lefties. Let's take a look at another view, which attempts to plot the trajectories of the pitches as they're seen from the Umpire's perspective. This plot shows every pitch thrown in this appearance, with the trajectories colored to represent faster (more red) and slower (more blue) pitches.

Okay, so we see how disturbing those trajectories appear from his release point. I think from this angle we can get a good idea of just how much movement is on each of these pitches, but it's even better from up above, looking down on the pitch:

We can also average those trajectories, to try to get an even better view of what's happening when he throws the ball:

Now the two very different horizontal trajectories of the slider and two-seamer are incredibly apparent. One pitch breaks very heavily inside to RHH, and one breaks outside to RHH. But does this really explain the Platoon?

I would argue - no, not quite. The Platoon split isn't really about stuff, in the sense that LHH have essentially the same dilemma to face with Masterson, they have one pitch that breaks strongly away and one that breaks strongly inside. So what is it about Masterson that's so disturbing to RHH?

For this we turn to Mike Fast's Batter's Box plots, which can show us that hidden hitter's perception, as the pitch is viewed from the Batter's Box. Let's take the LHH case first, remembering that they do relatively well against Masterson:

The big, sweeping slider is easily identified among the 2-seam fastball, as we can clearly see the large humps in the trajectory that identify it as the slider. In other words, the perceptual problem for LHH is pretty easy, because the slider and the 2seam fastball have such different trajectories when viewed from the LH side of the plate.

But then, what does a RHH see when they see Masterson's stuff? What creates the Platoon split?

The slider is gone! Well... not gone, but not visible, either. Of the eleven pitches on that plot, FIVE are sliders. Try and pick them out based on just the trajectory information: it's a mess. For RHH, the sinker/slider combo is pure deception, because there's no more simple heuristic to identify the pitch. This, and not discomfort, delivery, or magic, looks to be the source of the platoon split. When viewed from the batter's box, these two pitches appear wildly different. To LHH, they are easily identifiable and non-remarkable; to RHH, they are nearly indistinguishable.

Final Thoughts
Are there other things that contribute to the Platoon split? Absolutely. Stuff matters - the extra 7mph on the inside part of the plate, combined with the heavy sinking motion of the 2seamer, absolutely creates problems for RHH when they come up against a guy like Masterson. But this visualization shows us that sometimes, these well documented, traditional splits in baseball, that influence everything from relief pitcher usage, pinch hitter usage, and the advent of the LOOGY, might really be due to a simple difference in perception.

Note for Readers
All of the above plots can be generated using my freely available PitchFX tool. To generate viewpoint plots, select "Pitch Types" as the plot type and they will automatically appear.

Thanks
Thanks to the countless articles by Mike Fast, Josh Kalk, John Walsh, Alan Nathan, and many others, for making PitchFX so accessible and fun.

Last Strikezone Post Ever

2008-07-26T18:28:00.005-05:00

So, after mentioning the poor job yesterday, I feel obliged to mention the great job today. For the most part (and dealing with much more difficult pitches to call, Knucklers aren't easy), Darryl Cousins called a fantastic game.

There was a lot of griping both from the booth (Buck and McCarver) and from certain message boards that the Strikezone was unnecessarily tight and that Pettitte was getting squeezed.

To this, I say, look at this graph:

Honestly, with a few bad calls up in favor of the Yankees and one call inside in favor of Boston, the Strikezone was textbook all game, with pitches called as they should have been called, according to the locations tracked by PitchFX. Basically the only bad spot in the zone was the bottom right corner, and it was a consistent non-call. Other than that, the edges look very clean and sharp.

I'm also going to let you all in on a little secret: I do not have magic abilities to generate these maps, you can do it too! Select a game and pitcher with my PitchFX Tool and then click "Create a Normalized Strikezone Map". You will get a page like this one, which will show you all calls made on each team's pitching staff. Really, next time before you complain about the strike zone, check out the PitchFX data first, you might be surprised. The camera angle put on by most broadcasts really distorts your view of where the pitch ended up.

I would also be remiss not to mention that Justin Masterson got absolutely shelled today. Why, you ask?

Well, let me clear that up:

You can't be a sinkerballer looking for a ground ball and throw belt or higher pitches to Major League hitters and expect good results. It just doesn't work to leave pitches out over the plate.

Strikezone Theatrics

2008-07-25T21:36:00.000-05:00

Part of me wants to make this post just about how good Josh and Joba were. Because they were good. Very, very good.

Beckett was dominant throughout the game, and had one of his best curveballs of the season.

The Curveball is the pitch in the bottom right of the graph, with lots of horizontal and downward vertical movement. Lots of swinging strikes, and a few beauties that were called.

When I first did an analysis of Josh Beckett for SoSH a few months ago, a few people disputed the existence of the cutter.

Well, it's there pretty clearly in this start, but unfortunately get its hammered. Of 6 thrown, 3 are hit for singles.

But let's not kid ourselves, Joba was the better pitcher tonight. His fastball was overpowering, and his slider was incredible.

The slider is the cluster of pitches in the middle of the graph peppered with orange (swinging strikes), red (called strikes) and dark red (foul balls). Red Sox hitters only muscled 3 offspeed pitches onto the field all night, and all three were outs.

It's pretty easy to see the dominance of the slider here:

It's littered with black marks indicating strikeouts.

Color me impressed with his ability to sustain velocity.

He's topping at nearly 100mph very late into the game, and averaged 95.5mph on his fastball. That's bringing the heat.

But the big PitchFX story, unfortunately, wasn't the really fantastic pitching that we were treated to, but the really terrible umpiring. I know that it's taboo to use this information to critique the umps, but this game was so poorly called I can't imagine that anyone could have watched it and not thrown up their hands in disgust.

Specifically, the left side of the strikezone (from the Umpire's PoV) was an absolute mess. These plots show all CALLS made by the Umpire over the course of the game - so, called balls and strikes (not swinging, foul, hit, etc).

It's like he was going out of his way to call pitches far inside to RHH and far outside to LHH, which factored heavily into the 9th inning of the game, when one of the most awful calls I've ever seen went against Mike Lowell:

That pitch that was called Strike 3 to Mike Lowell after an incredibly tough AB was a hard pill to swallow, because it was over 1.5 feet from the center of the Strike Zone. I have no idea what Marty Foster was thinking there, but damn did that sting.