THE BOOK--Playing The Percentages In Baseball

There’s something missing to this story because the pieces don’t add up.

K-Zone uses PITCHf/x data, so it should be highly accurate.  Jonathan’s article doesn’t actually show anything from K-Zone.  I don’t know what Sportsnet uses for their strike zone graphic.  I’m not aware of anyone other than Sportvision providing that kind of data, but maybe Sportsnet has a different way of doing it.

Anyway, I’m skeptical that anything has been proven here.

This sounds like you are claiming that your eye is more accurate than the digital cameras around the stadium which the TV stations (some of them at least) use.

There are pitches in every game that look like obvious strikes, and then they show the over-head camera (behind the catcher), and it is clear that it was more than a couple inches off the black. If you want to claim that your eye is better than mine at seeing this, I’m completely fine with that-- you’re older than me and have presumably seen more games than I have. But I’m sure there are plenty of readers here who have felt this way before.

It looks like he’s saying that the pitch location is correct but the TV strike zone overlay is not.

That reminds me, why did Gameday remove their strike zone?  The answer is probably obvious but was it every addressed?

Dan, are you talking about me?  I am not claiming anything.  I am simply taking the article and the couple of examples it provides, at its word.

Mike, unless the author is cherry-picking some unusual examples, or he screwed something else up, it is obvious from the captures and from the pitch f/x data that they are NOT the same.

And yes, it is true that virtually every pitch that appears to be on the corner that the umpire calls a strike is NOT represented in the strike zone according to K-Zone.  So either K-Zone is consistently wrong or the umpires are.

Again, I am not claiming anything.  The only thing I am claiming in my last paragraph, which has nothing to do with K-Zone, is that I can generally tell what is typically called a strike and what is not, regardless of the “offset” camera angle.  Anyone who watches hundreds of games a year can.  That is not the same thing as saying that I can tell whether a ball crossed the strike zone better than a camera and computer system.  All I can do, and anyone who has watched thousands of games can do, is to see a pitch in a particular location and then search my mental inventory for hundreds or thousands of those same pitches.  When I find it, I can recall if it is usually called a strike or a ball by a typical umpire or whether it is a coin flip.  Obviously I don’t do that on a conscious level.

There’s a layer of human interaction in PitchFx people may not be aware of: in data handling, occasionally a pitch gets mixed up with an adjacent pitch, and Alan Nathan found one case in his analysis of Bonds #756 where data for a return throw from the catcher ended up in the PitchFx database.  So, we may be looking at some special cause variation here, rather than excessive common cause variation (i.e. the selected pitches may not be representative)

I think any analysis of PitchFx accuracy, if it intends to arrive at a conclusion rather than just discuss, should cover a wider popultion of pitches, and should attempt to investigate the apparent discrepancies.

Also, I don’t know if the author is focusing on the same plane as PitchFx - I can’t recall if PitchFx shows the location when the pitch reaches the plane through the front of the plate, the middle, or the back of the plate.  Do you recall, Mike?

On some pitches this will make a significant difference…

Sportvision doesn’t sell PITCHf/x to Sportsnet so whatever they are using to show the strike zone obviously isn’t as good as PITCHf/x.  But ESPN buys PITCHf/x so when you see that used in their broadcasts it should match up exactly with the data on MLBs website (or else ESPN should complain).

It looks like he’s saying that the pitch location is correct but the TV strike zone overlay is not.

That’s a good point.  So the pitch trajectory could be correct but the overlay graphic inserted incorrectly into the video.  Still, I was under the impression that Sportvision did the video production for K-Zone from what I saw when I was at Sportvision.  Maybe the networks do their own video production for that effect, but I don’t think so.  And, I suppose Sportvision could be doing it and still screwing it up, although I was very impressed with their technical chops.

That reminds me, why did Gameday remove their strike zone?  The answer is probably obvious but was it ever addressed?

The folks at MLBAM said that they didn’t want people questioning the umpires on every borderline call that happened to fall one inch outside the strike zone.  In the old days when the pitch location data came from the stringer clicking on the screen where he thought the pitch was located, if the ump called it a strike, the stringer would put it in the zone.  Now that the location data comes directly from the PITCHf/x system, the umps and the official rulebook zone may disagree.

Mike, unless the author is cherry-picking some unusual examples, or he screwed something else up, it is obvious from the captures and from the pitch f/x data that they are NOT the same.

Jonathan didn’t show any screen captures from K-Zone, yet he acted as if he had proved they had bad data, too, and you seemed to accept his data from Sportsnet as proof about K-Zone.  I don’t believe anything has been shown about K-Zone.

K-Zone is a Sportvision service just like PITCHf/x is a Sportvision service.  They use the same trajectory data.  Now it’s possible that the Sportvision people are screwing up the video production, but I’d like to see some evidence of that before I take your word or Jonathan’s word that something is wrong.  Data can be acted upon.  An assertion without evidence doesn’t do any good.

If someone can show evidence that K-Zone and PITCHf/x disagree, I can go talk to the baseball producer at Sportvision and find out what’s wrong.

Sportvision doesn’t sell PITCHf/x to Sportsnet so whatever they are using to show the strike zone obviously isn’t as good as PITCHf/x.

Ah, thanks, Josh.  That would make sense then.

Also, I don’t know if the author is focusing on the same plane as PitchFx - I can’t recall if PitchFx shows the location when the pitch reaches the plane through the front of the plate, the middle, or the back of the plate.  Do you recall, Mike?

Greg, the PITCHf/x location is at the plane at the front of home plate.

I’m not sure about K-Zone, but the graphical overlay being wrong would surprise me about as much as the Mariners selecting a college closer in the first round (Not one #@*@&#!’ bit).

As a frequent watcher of FSN games, they have their own version ( Tracer) of K-Zone that is also supposedly built off of Pitch F/X and have noticed numerous discrepancies between where the Tracer thinks a pitch was and where MLB’s gameday plots it. I don’t what, but something is off between the two.

Matthew,
Josh says Fox Sports Net gets their strike zone data from something other than PITCHf/x.  I’ve looked to see if I can find if FSN says where they get their data, but I haven’t found it yet.

I tend to believe Josh’s source since Marv White told me that they would like to sell K-Zone to other broadcasters but ESPN has an exclusive contract.

Totally unrelated to this post (strange coincidence I think!), I’ve just written a quick page that allows you to take all of the pitches thrown in a single game and plot them to get a feel for what the Umpire’s Strike Zone was. I realize that this changes from batter to batter (and thus needs some normalization), but in point of fact, it looks like Umps are fairly consistent and pretty good at their jobs.

Here’s a link to one zone plot taken from this game:
http://www.brooksbaseball.net/pfx/index.php?month=6&day=5&game=gid_2008_06_05_tbamlb_bosmlb_1/&pitchSel=452657.xml&prevGame=gid_2008_06_05_tbamlb_bosmlb_1/&prevDate=65
http://www.brooksbaseball.net/pfx/zoneplot.php?xml=http://gd2.mlb.com/components/game/mlb/year_2008/month_06/day_05/gid_2008_06_05_tbamlb_bosmlb_1//pbp/pitchers/452657.xml&innings=yyyyyyyyy&s_type=1&sp_type=1&h_size=700&v_size=500

I’m not trying to shamelessly self promote, I just thought it was pertinent to the conversation.

People often watch games and yell at the screen because “that’s a strike!”, but for the most part, Umps are actually pretty consistent in how they call the corners of the plate. I think it’s also fairly apparent from that graphic that at least Jeff Kellogg calls the bottom of the plate in a fairly consistent manner really regardless of who’s up.

I think you can look at that chart and find obvious examples where the PitchFX data disagrees and finds places where Balls are called even though Strikes had been called farther away from the plate (and vice versa). When you look at an entire game, I think this is captured in the PitchFX data pretty well.

Note that these charts only include *called* strikes and balls - swings or fouls or whatever else might happen aren’t included. I have also included the name of the Home Plate Umpire in the title of these charts, since for what it’s worth, he’s responsible for basically everything on this chart. =)

Jonathan didn’t show any screen captures from K-Zone, yet he acted as if he had proved they had bad data, too, and you seemed to accept his data from Sportsnet as proof about K-Zone.  I don’t believe anything has been shown about K-Zone.

I have no idea what “K-Zone” is.  I was using that term as a generic description of the screen shots that the author showed in the article.  If that is a different system than “K-Zone” then I apologize for being misleading.

I should have said that whatever “system” that the screen shots are using is clearly different from pitch f/x.  I also stand by my comments that when I watch games (I don’t recall what networks and I have no idea what system) the screen graphics seem to show pitches more on the outside than the umpire thinks they are and what I think they are.  Again, if what I am watching is not “K-Zone” then I apologize.

Just substitute “whatever the author was showing in the article” and “whatever I am typically watching on TV” for “K-Zone” in all of my posts.

In the article, Jonathan first refers to, “the little strike zone box” on TV.

Then he specifically says “Sportnet.” I don’t know if that refers to the network, the strike zone system, or essentially both.  Then he gives two examples (one has one pitch and the other has several pitches) where the “Sportnet thingy” on TV is significantly different than the pitch f/x strike zone with respect to those pitches.  So unless he cherry-picked those examples and they are not representative of an overall comparison of the two systems, then he DID document a big difference, and one that comports with my thoughts (that the box on TV is too small) after watching hundreds of games with the “little box thingy” on TV.

Later on in the article he says that he thinks (he makes the assertion) that ESPN does the same thing (has an incorrect strike zone on TV), but he does not back that up with any data/screen shots/comparisons.

Does ESPN use a different system than Sportnet? What is “Sportnet?” How many systems are there?  What do the systems share, in terms of data, machinery, etc?

You guys that work with pitch f/x data and go to all the conferences throw around these companies and terminology and it confuses lay people like myself, although I can only speak for myself of course.

Fantastic summary.  We need one of those for any long thread!  I’m going to highlight Mike’s post so that it stands out going forward.

"You and Jonathan and many others have expressed doubts about the accuracy of K-Zone, but I have found that people’s observation of the strike zone from TV is often inaccurate.  You claim to be a much better TV observer than the average fan, but in absence of some way to evaluate your claim, I am skeptical of it.”

Mike:

And how do you know that people’s observation of the strike zone is bad? I’m not trying to nitpick, but I suppose this is one of those things where an experiment is in order. Even still, we’d need some good way to objectify where the pitch actually went, because the argument here is that (whatever system they are using) is less accurate than the human eye.

I do want to point out that the over-the-shoulder look where you get a rotated view of the plate is being moved away from by a few broadcasts - some ESPN broadcasts, the Orioles and the Red Sox are both using Centerfield cameras that essentially look just over the Pitcher’s head, and the ability to see the movement / location of pitches is really phenomenally improved. I’m not saying you’re good or bad at it, but if you have MLB.tv or something like it, check out archived footage from one of their broadcasts and tell us if you’re seeing the pitch clearer - if you are, I’d think it was good evidence that the “mental rotation” isn’t as seamless as you think it is.

Here is a screencap of the camera:
http://i27.tinypic.com/96g55z.gif

Mike:

And how do you know that people’s observation of the strike zone is bad? I’m not trying to nitpick, but I suppose this is one of those things where an experiment is in order. Even still, we’d need some good way to objectify where the pitch actually went, because the argument here is that (whatever system they are using) is less accurate than the human eye.

Sportvision did a very impressive (to me at least) check on the accuracy of their PITCHf/x system.  They put a foam board at the front plane of home plate and had a pitching machine throw pitches into the foam board.  They tracked these pitches with their camera system and compared that measurement to the circular impression that the ball left on the foam board.  PITCHf/x was within half an inch.

Josh says Fox Sports Net gets their strike zone data from something other than PITCHf/x.  I’ve looked to see if I can find if FSN says where they get their data, but I haven’t found it yet.

I tend to believe Josh’s source since Marv White told me that they would like to sell K-Zone to other broadcasters but ESPN has an exclusive contract.

Mike, I’m not sure if it’s pF/X or not, that was just my impression from talking with a few people in SanFran when I brought it up at the Giants game how the FSN Tracer seemed horribly messed up. I’d love to get official word on this one way or another though.

"Jonathan didn’t show any screen captures from K-Zone, yet he acted as if he had proved they had bad data, too,”

Huh? I didn’t even mention K-Zone. In fact, until now I didn’t even know what it was. I thought I was pretty clear I was talking about the Sportsnet broadcasts that I watch, and took pictures of.

Since the data is fine and it’s just the box that is dumb, every broadcast could be different (although I doubt it because in passing I’ve noticed they’re all the same shape). If someone wants to send me some pictures of K-zone showing the zone, we can do the same thing for them, too.

I’ll try to capture something from ESPN to follow up on this but just mentioned it in passing because whatever they were using for playoff broadcasts last year was way off, too (although so was my zone).

http://bjays.wordpress.com/2007/10/17/fox-pitchtracker-is-set-to-make-umpires-look-bad/

These are not special cases- I just took the first couple of that game that had a pitch on the corner. It’s ALWAYS the same problem, equally to both sides. It’s not a problem with looking at the wrong plane either because curves and fastballs come out the same. As you can see in my article, the pitches are showing up in the right place, it’s just the box is way, way, too narrow.

Yes, great summary!

Again, I am saying NOTHING (sorry about screaming) about K-Zone.  I was merely commenting on the “Strike Zone” graphic that Jonathan was using in his article.

I accidentally called it “K-Zone”.  I also used K-Zone as a generic term for all of the TV graphic strike zones.  I did not know that there were 2 of them.

So we are in agreement about 1 thing:  Fox’ “Strike Zone” is apparently quite different from pitch f/x, and probably pitch f/x is the better one.

That was the whole point of Jonathan’s article which I thought was interesting. I had also thought for some time that some TV strike zone was “off” after watching many games over the last few years.

Jonathan also mentioned that “K-Zone” was probably off, based on his subjective recollection, but gave no evidence one way or another.

You say that “K-Zone” uses pitch f/x data/technology so there is no reason for it to be “off” unless somebody is doing something “wrong” in between getting the data and creating the TV graphic.

I have no opinion on that.  I watch many more games on the FSN channels than ESPN, so my recollection that the “TV thingy” was off was probably based on “Strike Zone” and not “K-Zone.”

I have no idea whether “K-Zone” on TV is off or not, but if I now watched a game or two that presents a lot of K-Zone, I can probably tell from the umpires calls and NOT from my great pitch recognition abilities!.

My interpretation of pitches has nothing to do with “Strike Zone” or “K-Zone”, so please stop bringing that up!  In an aside to my comments on the article in question, I merely mentioned that my friend thought that a person cannot tell a strike from a ball when it is close because of the camera angle, and I replied that a person (including myself) can make the mental adjustment to the camera angle, because they may have watched thousands of games and they get help from the umpire.  IOW, you can give me a 63 angle from a tree house 100 feet in the air, and as long as I have seen lots of pitches from that angle and have seen whether umpires call them balls or strikes, I can tell whether a pitch is going to be called a ball or strike by an average umpire,
whether it looks like the ball is in the strike zone or not, and regardless of the angle. This is not a unique gift I have.  It is something that anyone can do if they have watched enough games.

And again, my comments about that have nothing (absolutely nothing) to do with the discussion about how accurate K-Zone or Strike Zone is or isn’t.  Why do people keep bringing it up?

How did I “know” then that Strike Zone has always been off, as Jonathan has shown, at least for those couple of examples?  It has nothing to do with my strike zone recognition skills or the camera angle.  It is because when a pitch appears to be close to the edge of the strike zone and an umpire calls it a strike, the Strike Zone camera always has it outside of their strike zone.  That happens 10 times a game.  Regardless of what I think, there are only two conclusions:  One, the umpires are “wrong” on those pitches, or the TV strike zone is wrong.  Well, we have already seen studies that show that the umpires are more or less in line with the pitch f/x strike zone, so the ONLY conclusion left is that the TV strike zone (Strike Zone) is different than pitch f/x, rather than the umpires are wrong and the TV zone is right.

Sorry that this is so far into the interchange, but I just read the article, and am puzzled by the supposed “correct” strike zone as noted by the yellow box.  This box is a square, not a rectangle.  The width of the square is 20” (17 plus 2X1.5") as he notes.  Thus the vertical sides must be 20” as well.  So, I’m having trouble visualizing a player with a 20” vertical strike zone from the top of his knees to his waist.  Even on me (5’9"), it’s over 22” w/o adding the 3” for the ball widths.

Or am I missing something?

I don’t know if it is drawn to scale.  Plus, I don’t know if K-Zone, Strike Zone, or pitch f/x uses the rule book zone or the “de facto” zone.  In todays game, the rule book zone is still higher than the “de facto” zone.  The average umpire does not call a pitch at the top of the rule book zone a strike.  They also tend to call balls on pitches at the hollow of the knees, which I think is a rule book strike.  And of course the actual dimensions of the strike zone high and low varies with the player.

But, you would have to ask the “experts” what they mean when they draw a strike zone or compare one zone to another.  I assume that when they put the K-Zone or Stike Zone rectangle on TV, that it is custom made for the batter.  Other than that, whether it is the rule book zone or the de facto zone for that batter, I don’t know.  If it is me, and I am drawing a strike zone, I am using a de facto zone, not the rule book zone, unless I want to show people the differenced between the rule book zone and the de facto zone.  If I want to do any kind of analysis or even present data for people’s interest, of what use is the rule book zone if it is not called that way?  So I can say, “Look 98.3% of the umpires in baseball call a high strike a ball?” No, if 98.3% of the umpires call it a ball, then let’s ALL call it a ball, regardless of what the rule book says.  If we want to analyze pitchers and batters in terms of how many pitches are in or out of the zone, how many in or out of the zone they swing at, what happens on pitches that are struck in and out of the zone, etc., we obviously want to use the de facto zone to define “in and out of the zone.” Basically, I see no use for the rule book zone whatsoever, except, as I said, to occasionally show the difference between the rule book zone and the de facto one, for people’s amusement.  As to why the strike zone changed at some point (I should say during some time period, as things like that do not change overnight), I don’t know. Someone must have decided that more offense was needed, since high fastballs that are strikes are really hard to hit.  That assumes that at one point a high pitch WAS a strike.  I think that is the case from watching videos of games from the 60’s and even 70’s, but I am not sure.  I think that if baseball were invented right now and the umpires were given the rule book strike zone, but somehow they knew that it was OK if they tweaked it a little, they would probably lower the zone on their own.  When you are umpiring a game, it is natural to call a strike zone in which you think it is reasonably possible for a hitter to hit the ball and get a hit.  A 95 mph fastball thrown at the letters is really hard to hit and if you were an umpire, I don’t think you would be too happy with calling it a strike.  Most umpires, or the average umpire, that is. Obviously there are some umpires that call whatever they like, regardless of how hittable they think pitches might be.  I mean, why the heck does Daryl Cousins call virtually every close pitch a ball?  Almost no one else does that.  What is in it for him?  Maybe he thinks he is being paid by the hour.

MGL:

Have you considered selection bias - that the TV producer only calls for the strike zone graphic when the umpire call and the output of the “system” disagree?  Certainly they don’t show the strike zone on every pitch, or every called strike, so you’re not seeing the full population… This is TV, after all

"Jonathan didn’t show any screen captures from K-Zone, yet he acted as if he had proved they had bad data, too.”

Huh? I didn’t even know what K-Zone was until now, and I think it’s pretty clear that I was referring to the Canadian (Sportsnet) broadcasts that I watch and screen captured. If someone has a picture of the K-Zone in action, send it to me and I’ll check it out.

I mentioned ESPN in passing because last playoffs they had the same problem with the zone (although I made a mistake with mine as well).

http://bjays.wordpress.com/2007/10/17/fox-pitchtracker-is-set-to-make-umpires-look-bad/

And on the few ESPN feeds I’ve seen this year it looked exactly the same shape and seemed to be just as stingy on the outside pitch.

There is no selection bias at all here. I just took the first two times it was used in the game I recorded, because they happened to show pitches on the corners. Sometimes there’s no conflict with the ump, they just show an entire at bat and it’s the exact same and the strike zone is 100% consistently squashed on every call.

Let me be clear I’m not ripping whatever system is tracking these pitches. The pitches are all in the right spots (and I assume it’s using pitch f/x because they’re identical). And it’s not taking pitches from the wrong plane because breaking pitches and fastballs are the same. The only problem is the box representing the strike zone is drawn as if home plate was about a foot wide, not 17 inches.

MGL:

First, fun discussion.

Two points:
First-
According to John Walsh, it turns out that the zone is really around 20 inches wide because it’s 17 inches + the ball (1.66 feet), while the vertical zone is about 1.95 feet rulebook (but about 1.75 feet when you actually look at the “de facto” zone). That’s a lot more square than a lot of people realize (even the rulebook zone). The zone *is* rectangular, but it’s not nearly as rectangular as commonly portrayed, I think.

The article I am talking about is here:
http://www.hardballtimes.com/main/article/strike-zone-fact-vs-fiction/

I think it was discussed on a post by Tango on this very blog last year. =)

Second-
It would actually be a pretty interesting question to ask the Sportvision people how they determined where sz_top and sz_bot are located (these are the parameters in the file that corresponds to the top and bottom of the zone, but WHAT zone exactly, no one really knows).

From Mike: “sz_top: the distance in feet from the ground to the top of the current batter’s rulebook strike zone as measured from the video by the PITCHf/x operator.”

So, it seems as though Mike knows the zone given by PitchFX is the “rulebook” zone and not the “defacto” zone - Mike, is this true?

I’m not sure it is. I checked out the Strikezone for a few games tonight and the zone given by sz_top and sz_bot, when normalized for the different batters, actually does a pretty damn good job predicting whether or not a ball is going to be a strike. Sure, things get fuzzy around the edges, but there are plenty of strikes called just at the edge of the top of the zone (as given by Gameday) and just at the bottom. According to MGL, if they were using the Rulebook zone, Umps would be very likely to call those pitches balls.

So, some questions:
1) Are we sure that the Strikezone is appreciably different from what the rulebook says? (this: probably yes).
2) Are we sure that the Gameday people are using the Rulebook Strikezone and not trying to make their own judgment about where the zone probably starts and stops? (I would guess they are actually not using the rulebook strikezone, because the parameters now seem to fit the defacto strikezone too well)
3) If they aren’t using the rulebook strikezone, how exactly are they setting those parameters?

One interesting thing, looking back at Walsh’s article, is that the top of the zone is actually accurate - it’s the BOTTOM of the zone that gets shaved by the defacto zone.

-Dan

Good questions…

1) Sure. But it gets really cloudy in a hurry because it’s different for R/LHB, umpires call it significantly differently, and the corners of the zone are not called.

http://bjays.wordpress.com/archives/missing-corners/

I usually just define the edge of the plate with the caveat than on average umps call it a little larger than that. I think there’s too much going on to say a lot about what the “real” zone is in two dimensions.

2/3) I’d like no know that as well. I’ve noticed the same thing, that their definition of top and bottom seems to be calling pitches pretty well, which would make me think it’s not going by the letter of the law- but as that Walsh article points out maybe the strike zone as the umpires call it isn’t really that far off from the rulebook vertically, anyway.

The sz_bot and sz_top in the Gameday data are set by the PITCHf/x operator at the beginning of each at bat.  He puts one line at the bottom of the hollow of the front knee and the other line at the belt.  The PITCHf/x system then automatically adds four inches above the line at the belt to determine the top of the zone.

It’s a bit of judgment call by the PITCHf/x operator because he tries to do this as the batter settles into his stance right as the pitch is released but before he starts moving his body for the swing.

I definitely agree with Jonathan’s point in #26.  Trying to draw the real zone would be too complex.  The corners are shaved off, as Jonathan said, so the zone, as called, is not a rectangle.  But I usually just draw the rulebook rectangle and explain, if I need to, that the umpires call a few extra inches outside for lefties and don’t call the corners.

"The sz_bot and sz_top in the Gameday data are set by the PITCHf/x operator at the beginning of each at bat.  He puts one line at the bottom of the hollow of the front knee and the other line at the belt.  The PITCHf/x system then automatically adds four inches above the line at the belt to determine the top of the zone.

It’s a bit of judgment call by the PITCHf/x operator because he tries to do this as the batter settles into his stance right as the pitch is released but before he starts moving his body for the swing. “

You should update your guide to say this, if it’s not already somewhere else on the web. Thanks for answering the question. =)

So, if they’re only adding four inches above the belt, that seems like a good indication that actually the PitchFX zone is not the rulebook zone and instead maps to the defacto zone, since, maybe I’m wrong here, but the letters are a bit higher than that above the belt. This makes sense, given the data.

Put another way, the actual rule-given zone is:
“The STRIKE ZONE is that area over home plate the upper limit of which is a horizontal line at the midpoint between the top of the shoulders and the top of the uniform pants, and the lower level is a line at the hollow beneath the knee cap.”

I would wager that most people have larger than an 8-inch torso.

I’d also be interested to see splits for LHP and RHP, and not just LHH and RHH. In fact, we have so many pitches that it probably makes sense to just plot all four graphs in a 2x2. Anyone have those handy?

Well, there’s this cool blog I read…

http://tangotiger.net/halejon/

It’s just raw strikes, but if you squint in the right light at the .pdf’s at the bottom, you can see how the strike zone changes for all the L/R matchups. I haven’t seen the 50% of balls being called strike = edge of strike zone trick done for all the matchups, but pretty easy to do.

Great posts above!  I stood in front of a mirror and measured the distance from the top of my pants to the top of my shoulders and it was about 20 inches, give or take.  I am 5’10” so I imagine that for some of these taller guys the distance could be 24 or 25 inches.  Anyway, using 20”, that puts the rule book zone 10” above the belt, a far cry from 4 inches above the belt. I think we can safely say that the de facto zone, or at least the one that Sportvision uses, is nowhere near the rule book zone at least at the top.

And of course, one of the biggest things that varies among umps is the height of the zone, both at the top and bottom.

The Sportvision/pitch f/x people should be setting the strike zone to correspond to the average umpire, regardless of what the rule book says.  It sounds like they are trying to do that. The way to do that, I would think, would be to take all the pitch data for all the umps and see exactly which zone gives the lowest mistake percentage among all umps combined, using some mathematical algorithm.  For example, if 3.5 inches above the belt results in fewer mistakes than 4 inches above the belt, than that should be used.  It would be silly and wrong for pitch f/x to set the strike zone at exactly 4 inches above the belt and the hollow of the knees if that isn’t the strike zone that the “average” umpire uses.  I am not saying that it is or it isn’t, just that the Sportvision guys should carefully determine the best strike zone based on the average umpire.  As I said, when we analyze the data it is important to know what we mean when we say that a certain pitch was in or out of the zone.  If, for example, the real strike zone for the average umpire is only 3 inches above the belt, then in an analysis we are going to mis-classify all those pitches that are between 3 and 4 inches above the belt as “in the zone” when in fact they are “out of the zone,” again, for the average umpire.

Personally, I find measuring x” from the belt to be a little odd.  I had a roommate in college who was 77” tall with a 40” inseam, while I was 73” tall with a 30” inseam.  Maybe we are both freaks, but even though he was 4” taller than me, from the top of the inseam to top of the head, I was 6” longer, and his neck was much longer.

Perhaps the umpires don’t care and do just go x” above the belt, but it seems to me this does a disservice to different body types.

Umps need the reference point, and x-inches above the belt is perfectly acceptable in practice.

Since John Walsh has already provided the data in an article last year, we know “exactly” what the defacto strike zone is.

And if the Sportvision guys, bless their heart, are setting the top as 4+belt, then we know exactly where the belt-line is for all hitters, and we can see if there is an umpire-bias for guys with high-belt line and being tall/short, and guys with low-belt line and being tall/short.

If we can show that umpires are indeed calling pitches at x+belt for all these combination of height/inseam players, then we just need to come up with the x (either as a straight additive number, or as a percentage of height minus belt).

SportsNet uses QuesTec PitchTrax as its source for strike zone information according to the QuesTec web site. Also on the site is an announcement that they will soon offer HitTrax to track hit balls.

The QuesTec website has some outdated information still being shown, so it may not be a certainty that it is still providing information to Fox for SportNet, but it is the probable source.  But HitTrax may have been something they were planning back in 2003 and never implemented.

A 2003 article by Nathan Fox for Baseball Prospectus interviewed Ivan Santucci, a project manager for Questec’s UIS.  Santucci says that the strike zone upper and lower limits are set by the operator for each pitch almost identically to Sporvision’s Pitch f/x.  The operator sets lines to the batters belt and knee levels and the computer automatically raises the upper limit 2.5 ball diameters above the belt line. 

Another article at Baseball Prospectus on 8/10/2004 was written by Jason Karegeannes, a former Questec UIS operator.  That article, titled “Confessions of a QuesTec Operator” goes into much more detail about the system.  It also states that as of that time QuesTec had installed cameras in 23 of the 30 ballparks.  Karegeannes also talks about possible sources of error in the QuesTec system.

Hale/20 and Hale/25 may be a duplicate post.  One was queued for moderation, which I have just unqueued.

After watching the ESPN game and its Pitchtrax intently last night, the only thing I can conclude with any certainty is that Joe Morgan’s brain resides on a planet other than Earth.  I am referring to his conversation with Jon Miller about who has hit the hardest line drives “on a consistent basis” in the history of baseball.  Joe said, “Here is a trivia question for you.” Miller replied, “That’s a trivia question?  Where would I look up the answer?” The rest was hilarious.

While we are on that note, I am not sure that this was Eric Young, but I was listening to Baseball Tonight last night, which I rarely do.  Someone sounded like a cross between Bugs Bunny and Elmer Fudd.  I think it was EY.  When he finished his interview, did someone from ESPN say, “Now that’s a voice we need on the show guys!  You’re hired!”

Here is an outtake from Marv White’s presentation on the PITCHf/x system at the summit.  It shows how the strike zone is set.  The two red lines are the ones set by the operator. 

http://fastballs.files.wordpress.com/2008/06/marv_white_pfx_system_overview_slide14.jpg

This slide wasn’t made to demonstrate the strike zone.  The video frame was probably snapped after the operator set the strike zone since the pitch has almost reached the plate.  So it’s after the moement that the batter’s his strike zone was measured, although he’s not swinging at the pitch, so he probably hasn’t moved much. 

I’m not sure from the looking at the image that the four-inches-above-the-belt measurement is that far off from the rulebook strike zone.  (The top yellow box is the strike zone.) Both the batter’s torso and his belt are at an angle such that a 4-inch vertical adjustment seems pretty close to rulebook to me.

Ok, I’m pretty sure the K-Zone is messed up, too (though not as much and it’s a better shape). Someone just sent me a picture of (among others) a Ronny Cedeno single against Brad Penny. On TV the pitch is inside and maybe the edge of the ball just grazes the strike zone - but pitch f/x says it was 8.184 inches from the middle of the plate. Unless I’m missing something, that should be a clear strike, especially considering the width of the ball. I’ll put up some screenshot comparisons soon.