THE BOOK--Playing The Percentages In Baseball

Harvey, I moved your comment to the appropriate thread:
http://www.insidethebook.com/ee/index.php/site/comments/hardball_times_annual_2008_starring/#42

Sorry for the confusion Tom.  I will make sure to get this fix for the next round of articles.

I should stress that this is my opinion, which shouldn’t carry more weight than your personal preferences.  But, thanks for clearing anything up!

I’m not 100% sure (though I could be convinced). If someone hurls a standard fastball with no movement other than gravity would a hitter say that the ball had “break”?

I have not read Josh’s article yet, and I have nothing but great respect and admiration for these guys that are doing pitch f/x work…

I was thrilled that John in the THT Annual finally used run values to estimate the value of a pitch.  I agree with Tango that using OPS or something like that (I am still not even sure what “OPS” means for one pitch) is bad.  Every one of these guys needs to use run values.  And they are easy to use.  I think that John found that a ball was worth .065 and a strike -.097 runs, on the average.  Once you know that, figuring the run value of any one pitch is easy.

The only problem with that, as John keenly points out in the THT article, is that the ball and strike values are based on the average distribution of counts.

What I mean by that is that, for example, at a 3-0 count, a fastball is thrown probably like 90% of the time and at an 0-2 ount, an offspeed pitch is probably thrown 50% of the time.  So the ball and strike value of a fastball or offspeed should not be based on the average percentages of all counts, and will end up with different values.

O.K., that didn’t come out too well, but the best way to present the run values of a particular pitch is either to always just use the count at which it is thrown (IOW, at 3-0, the value of the strike or ball is different than if it is 0-2) OR to use different ball and strike values for different pitches, but use those constant values all the time.  IOW, Walsh’s .065 and -.097 really should not apply to all pitches.  Each pitch should have its own values.

Yes, that’s why the Burley values I linked to are so important (if done right).

For example, on a 3-2 count, the gap between a called ball and a called strike is .443 runs.

It looks like Craig used “runs created” run values, rather than linear weights.  The gap must be around .600 runs, since a called pitch in those cases is either a walk or a strikeout.

The actual value of the ball or strike would be based on what the Linear Weights is once you reach a 3-2 count (which you would know from the link I provided).

Maybe I’ll present my “official” numbers on this.  I think Phil Birnbaum also presented his numbers in a BTN a few years ago.  Phil, if you are there…

#4 That’s been my point all along, too.  That’s why I don’t add the gravity part into my movement plots.

BTW, Tango, what is this time component to the break that you mention above?

John/4: if you throw a 85mph curve ball, and a 75mph curve ball with the exact same spin and release point, I would think the batter would say that the 75 curveball would have a bigger break.  One would finish say at the letters and the other would finish at the knees.

John/7: I meant gravity+time together.

And I am not advocating to not show the spin-induced movements.  Not at all, since this tells us exactly what the pitcher is throwing.  But, under no circumstances should we use the word “break” here.  Break is a human term, as the batter sees it.  And he sees it in combination with the gravity+time influence.  You can call it the “additional movement due to spin” or “spin-induced movement” or some such.

As for the presentation of the spin, I prefer the sample chart I recently showed, which Mike Fast implemented.  I do not like seeing it as is currently prevalent, since it is highly confusing to the reader: 90% of readers will look at those standard charts and presume it’s the change in position of the ball, including gravity+time.

Only becuase it is likely that a 75mph curve will have more spin on it that an 85 mph curve .

If you through a 30mph parabolic underarm (maxima at say +15ft) would you say that would have a lot of break?

I’m saying throw the same pitch the exact same way, and the only difference being the speed, and a batter will say that the slower one has a bigger break.

Your other point is comparing a different trajectory, so you are introducing more parameters to consider.  Let’s stick with the easy one: you throw a curveball, at the same release point, with the same spin, one at 85mph and another at 75mph.  One ends up at the letters and the other at the knees.  Isn’t the batter going to say that the slower one had the bigger break?

I highly doubt he’s going to say: “Well, judging but the speed and where the pitch landed, I can infer that both pitches were thrown with the same spin, and therefore since both had the same spin-induced movement, they both had the same break.”

If you threw two pitches with the same spin but different speeds the slower one would have more break because by definition as it is slower the spin has more time to “act”. Therefore the pfx_z displacement, as measured by pitchf/x, would be bigger.

Therefore this statement ““Well, judging but the speed and where the pitch landed, I can infer that both pitches were thrown with the same spin, and therefore since both had the same spin-induced movement, they both had the same break.” isn’t correct because the slower pitch would have more break (but the same spin).

However, I don’t think that is really your point. Your point is that gravity will also make the ball drop more (it is probably more important than the spin in causing the ball to drop) so it will appear to magnify the break even more.

I don’t think there is a right answer here. Throw two balls with no spin, one at 90mph, one at 60mph. Does the one at 60 mph have more “break”. If you define break as including gravity then yes ... however, I think break implies movement imparted on the ball by the pitcher. And you conclude the 60mph was just a slow toss.

However, if you do the same example at 75mph vs 77mph (same spin) then the batter would conclude that the 75mph ball had more break.

I guess it all comes back to context.

Right, that’s my point.  “Break” is a human term, as the batter sees it.  Gravity+time exists.  That’s the context.

There is no context that excludes gravity+time that anyone would use the word “break” for.  In the example I cite (all same parameters, except speed), what ballplayer and fan would say that the two pitches have the same break?  Nobody.

I’m not convinced gravity needs to be added back in. Hitters often talk about rising fastballs, which wouldn’t be the case if they saw gravity as part of break. Our brains calculate an expected path of the ball to home plate, so even though a fastball drops some due to gravity, it’s perceived as being straight. Likewise an overpowering fastball will be perceived as rising, even though we know it is moving downwards.

It seems like the proper context isn’t what the ball is physically doing, but how it looks in the mind’s eye. I happen to like the positive vertical break numbers since it explains why hitters see rising fastballs.

We really need to know what a batter perceives as straight--is it a 90mph fastball, a BP fastball, or does it vary by pitcher (so a batter will see an 82mph fastball from Moyer as straight, but a 95mph fastball from Joba as straight)?

I agree that we need to know what the batter thinks

I said that he won’t say this:

Well, judging by the speed and where the pitch landed, I can infer that both pitches were thrown with the same spin, and therefore since both had the same spin-induced movement, they both had the same break.

But, Anthony’s point is that the batter might very well be thinking this, that he realizes that a slower pitch will automatically drop lower.  Therefore, he might not see it as a big break even if it’s a big drop.

At the very least, the “drop” occurs in the real world.  And, the spin-induced (excluding gravity+time) movement is what it is. 

The question is: “what is break”?  I would think batters would utter the word break synonymously with “drop” as I’m defining it here.

Anyway, it looks like we’re all on the same page as to the issue.  What we need is a MLB player to come here and answer the question.

I’ve thought about it some more, and I think I know what we (or an MLB player) wants: reaction break.

A while ago, I said that I preferred to know the reaction distance of a pitch: how many feet will a ball travel in its last 0.20 (or 0.30) seconds.  I think we want the same thing here: how much will the ball break in the last 0.20 seconds?  This will take care of the issue of a really slowly thrown ball.  It’s not that we really care that a ball will physically drop 3 feet, but rather how “sharp” it drops (i.e., the speed at which it drops so much). 

Reaction break.  That I think is what we want.

Tango, you may be onto something here that will be useful both for pitch classification and for talking about what hitters perceive.

Hitters and scouts talk a lot about certain pitches having “late break” which I always took to mean the same thing as just saying that it had a lot of break.  The acceleration of a pitched ball is very close to being constant (except for a knuckleball, which isn’t what’s being described as having late break here).  The bulk of the movement happens in the last part of the ball’s travel, but that’s simply the nature of a constant acceleration.  So previously I didn’t see a significant difference between total break and late break, probably because I was thinking in terms of distance rather than time.

But what you describe as “reaction break” seems to me to line up with what players are talking about when they talk about a pitch with “late break”.  If it’s thrown hard and it breaks a lot, it has “late break”, which can be measured as you described.

I’d still love to have an answer to the question of what hitters would identify as a pitch with no break.  We don’t necessarily require a major-leaguer to answer that question.  Anyone who has competed at a level where pitchers throw fastballs with a lot of hop (~90 mph or so) could probably give us some insight.

For those who haven’t seen it, I brought up the concept of reaction distance (of which Joe came up with the cool name and work) linked to here:

http://www.insidethebook.com/ee/index.php/site/comments/pitch_trajectory/

Perhaps I am just being dense, but I have been wondering about ‘break’ and how it is handled both by equations and as the batter sees it. I’ve been reading Alan Nathan’s site and Josh Kalk’s site and a few others, and see nothing that seems to answer my question. Probably best to give an example before trying to explain:

LHP to LHB
A) Pitcher throws a slider that to the batter seems to be moving inside for a ball, but curves back over the inside corner of the plate
B) Pitcher throws a slider that to the batter seems to be over the middle of the plate but curves off the plate away from the batter

Now, it seems to me that the exact same rpm, rotational axis, and speed could result in both of these pitches, with only a change in the release point (earlier/later by whatever means) being the difference. They are very different pitches to the batter, but does the data and equations see them as such? Since the equations are comparing the trajectory to a straight line from the begining point to the ending point (just in front of the plate) it would seem to me that they are seeing them as the same pitch; but then the comparison lines between points would be of differing lengths so the time would be different. I hope I am making sense.

Perhaps the straight line from release to homeplate that the trajectory is being compared to should be taken into account, since this plays a large part in what the batter sees?

SirKodiak, you are thinking along the correct lines.  However, it’s a change in the initial velocity vector rather than the release point that results in a difference in the final location crossing home plate.

It helps to keep in mind what is trying to be accomplished by various analysts.  It’s true that almost every major league pitch is unique.  Many people are trying to find ways to group them for better understanding of how they move. 

One powerful principle that has come from the PITCHf/x data is that certain types of pitches move the same due to spin no matter where the pitcher locates them in (or around) the strike zone and that this movement can be quantified.  The acceleration on a particular pitcher’s curveball will be about the same no matter whether he throws it up and in or down and away.

That doesn’t mean that location is irrelevant, just that it is convenient to ignore it for some purposes.

I would think that the initial velocity vector is determined by hand speed and release point, but that was a minor point I was trying to get to.

It is just when I see ‘break’ being put into tables along with results, and location ignored, that I become concerned.  While there are many things that go into the effectiveness of a pitch, it seems that the information is there to quantify trajectory, speed, and location; and that to leave out location when looking at effectiveness (rather the grouping) is not good.  If there was a study that showed that the trajectory was vastly more important to the end result than location, then my concern would wane.

I do have a concern about grouping also, though.  To group all of a pitcher’s curveballs or sliders into one group seems sound since most pitchers throw those a single way, but for fastballs and changeups it is a bit muddier.  There are at least 2 distinct chaneups that behave much differently from each other and an individual pitcher may throw both.  Fastballs are even worse.  For example, I saw f/x numbers for Felix Hernandez, and they listed sinker, curveball, slider, and changeup.  He throws a four seam fastball as well, and those pitches must be in the sinker numbers, and would skew the numbers away from the true sinker numbers.

I applaud the work being done, I am just concerned that conclusions and comparisons are going to be made too soon (before the data is understood well enough).

As part of calculating the average run value for various types of pitches, what’s the best way to calculate run value for pitches that were put in play?

First, is it better to use the run value of the actual event (single, double, triple, home run, ground/fly out, etc.) or to apply DIPS and use the run value of the batted ball type regardless of the actual result of the ball in play?  I’m leaning toward using the actual result rather than the batted ball type at this point since we don’t know whether DIPS applies equally to all pitch types.

Second, I assume, whatever method one chooses in the first question, it makes sense to subtract out the expected run value from the ball-strike count on which the ball was put in play.  A single on a 3-0 count is not as damaging as a single on an 0-2 count.  Is this correct?  If so, I’m not sure how to get the overall run value for a given count from Craig Burley’s data (or from Tango’s, for that matter).

I am very late entering this thread, but I am intrigued by the issue of “late break.” Hopefully someone is still monitoring it for new posts.

I started thinking about late break and what it might depend on.  I will skip all the equations and get to the bottom line:  I have managed to convince myself that the rate of break (i.e., the break divided by the time) is directly proportional to the spin on the ball w and independent of the speed of the ball.  In arriving at this result, I am including only the break due to the spin on the ball. 

If you use the formulas in my Jon Lester analysis
(http://webusers.npl.uiuc.edu/~a-nathan/pob/Analysis.pdf) to derive the spin w, then how does that correlate with what others have found for “late break”?

The “late break” or “reaction break” is the amount of break in the last 0.25 seconds of a pitch, prior to crossing the plate.

Mike Fast has an article linked to from here:
http://www.insidethebook.com/ee/index.php/site/comments/pitch_analysis_of_eric_bedard/