THE BOOK--Playing The Percentages In Baseball

What I don’t get is the break.  Are they measuring how far it drops or how it moves side to side?

I’m surprised to see the pitches with the highest values for break are actually the fastballs.

Two things:  One, those are awfully large differences between release and home plate.  Do all “guns” measure at around the same location?  There needs to be a standard as to where pitches are measured if the numbers are to mean anything.

Two, these “break” measurements ar ridiculous unless “break” is defined.  Fox has recently had some kind of “movement” in their fox trax.  It seems to have nothing to do with what kind of pitch is thrown.  I’m not sure what it means.  Something like where the pitch ends up compared to where it started, regardless of the “break”.  For example, a low fastball may have a “movement” listed of 10” and a curveball that lands over the plate may be 7”.  That is ridiculous of course, but not surprising from Fox.  The decision makers at Fox appear to never actually watch TV or sports.

MGL,

The physics of baseball book (by Adair) says you’d expect a 7mph slowing of the pitch between release and the plate; which is consistent with the zumaya data.

Guns should record the release velocity (if they do then it is impossible to juice them)

John

Total guess here… but if it was me doing the “break” stat, I’d draw a straight line from the release point to the point at which the ball crossed the plate.  And I’d draw a line that followed the exact path of the ball.  Then put up some “ribs” between the two lines, which are perpendicular to the straight line.  Longest rib = break.  I’d guess Zito’s curve might have the biggest “break” stat in the league.

John:  Why should guns record release velocity?  I would think we care about the velocity when it reaches the hitter.  (Although consistency probably matters more than this choice).

I think the procedure for radar guns was to take two readings, the release and when it reaches the batter, and average them.

What FOX does for the playoffs is just give us release velocity.

"I would think we care about the velocity when it reaches the hitter.”

The important number is the time from release to when the ball crosses the plate.  But unfortunately we have been programmed to think in terms of pitch speed.

I can’t wait to see what the ball path mapping technology does with a knuckleball.

Break, as experienced by the batter, is a compared perception of what a pitch does in reference to what a normal fastball is expected to do.  So a rising fastball doesn’t really rise, it just doesn’t fall as much as a normal fastball does from the influence of gravity during its trip to the plate.  The numbers would make more sense if gravity’s influence was factored out which would be an easy thing to do.

WE might care about the velocity when the ball reaches the batter, or the time it takes to go from release point to the batter.

But the other 99% of people who watch MLB.com gameday broadcasts want to know how fast Zumaya can physically throw a ball, and they want that number to be as large as possible.  So that’s why I think they include the muzzle velocity - for the “oooh, 103 miles per hour!” factor.

Would be interesting to see break versus fastball break, I like that idea.

I think Peter’s idea (time in air) makes the most sense, and if you extend it more, I suppose you really want to know “distance from home plate when ball is 0.20 seconds away”, or whatever the average MLB player reaction time is.

John:  Why should guns record release velocity?  I would think we care about the velocity when it reaches the hitter.  (Although consistency probably matters more than this choice).

Does it really matter as the rate of slowdown should be similar for all pitchers .... I’d imagine

As I was thinking about this new technology I realized that it was already perfectly set up to calculate speed of hit balls off the bat and angle of flight in both the horizontal and vertical planes.  We should be lobbying to have that information included on Gameday as well as it would be very useful for defensive assessment.  Bat speed should also be within the technology’s capabilities.  All very useful information.

HEY MGL -

IN SOME WAY, YOU’RE TO THANK FOR LAST NIGHT AND THE NEXT WEEK OR SO....

SO THANK YOU!!!!  THANK YOU VERY MUCH!!!!!

Theres a little button on GameDay that leads to a FAQ with a number of questions, including the definition of “break.” They claim it incorperates both horizontal and vertical movement.

The #1 question should be: how accurate are the measurements? The reported speed in tenths of MPH suggests a high level of accuracy… Is that really possible?

Its the holy grail for pitching data, if it really is that accurate.. oh, and once they figure out how to gather data on balls that were swung on.

To be accurate to a tenth of a MPH their triangulation program would have to be able to discriminate to about 5/8”.

In John Beamer’s primer on Guns:
http://www.hardballtimes.com/main/printarticle/zoooomaya-and-speed-guns/

he says:

Wow! Zumaya’s gas is over 2 mph faster than Billy Wagner’s. Is that a statistically robust result? Running a simulation of pitch speeds for a typical reliever shows that the standard deviation of a fastball is around 2 mph. A first response may be to conclude that the result isn’t statistically significant. This would be slightly disingenuous, because it ignores the fact that the incremental difference in average pitch speed between Zumaya and Wagner (numbers one and two on the list) is 10 times what it is between Wagner and Bobby Jenks (numbers two and three).

Actually, John is wrong.  The standard deviation he is calculating is on the wrong population.  He is taking all fastballs thrown by all pitchers, and figuring that the standard deviation is 2 MPH.  But, that’s not the correct population!

What if I calculated the standard deviation of times on base (which would be 34% a “1” and 66% a “0")?  That answer would be almost .50.  And if we look at Bonds’ .500 OBP and compare that to the league average of .340, then we can’t conclude it’s not statistically significant!

Let’s take an extreme case, and assume Zumaya always throws 98-99MPH on every single pitch (all 1000 of them) and Wagner always throws 96-97 on every single pitch (all 1000 of them).  Is the difference between the two statistically significant?  You’re darn right it is. 

What if Zumaya throws between 90-102 MPH, in an even distribution, with only 30 pitches, while Wagner throws 88-100 MPB on an even distribution, with only 30 pitches.  Now is it significant?  Hardly.

***

The only reason I’m so picky here is that this stuck out like a sore thumb in an otherwise excellent article.

On reflection (in this one minute), he may be taking the standard deviation on the average for each pitcher.  While that’s better (and renders my Bonds example moot), it’s still wrong.  Unlike OBP, which follows a binomial, what we care about here is the SD for each pitcher.  And the SD for a fastball for a pitcher cannot be 2 MPH.  That would imply that 95% of a pitcher’s fastball would be +/- 4 MPH (say 86-94).  That’s not true.  I’d guess the SD would be 1 (e.g., 95% between 88-92), if not less.

The only reason I’m so picky here is that this stuck out like a sore thumb in an otherwise excellent article.

Tango, you are right. Schoolboy error on my part which John Walsh bought to my attention. Hopefully a correction should be posted shortly.

My only excuse is .... heck, I don’t have an excuse.

On reflection (in this one minute), he may be taking the standard deviation on the average for each pitcher.  While that’s better (and renders my Bonds example moot), it’s still wrong.  Unlike OBP, which follows a binomial, what we care about here is the SD for each pitcher.  And the SD for a fastball for a pitcher cannot be 2 MPH.  That would imply that 95% of a pitcher’s fastball would be +/- 4 MPH (say 86-94).  That’s not true.  I’d guess the SD would be 1 (e.g., 95% between 88-92), if not less.

I actually did neither. What I did was to build a relatively crude simulation of fastballs based on a bunch of data I got from guns on mlb.com (of the same pitcher), I then took the standard deviation of that specific population and used that. I think my original simulation was probably a bit wonky and gave a slight wide range. I agree with you std dev closer to 1 mph, maybe a tad higher—Zumaya probably clocks 95% of his pitches between 97 and 102 mph. Looking back at my results I actually got 1.7mph, which is a little high. I rounded up to 2mph.

What I should have done then was to divide by sqrt(n) to get the sample standard deviation. This was the error.

John:
Nice article.  I couldn’t tell if your figures are velocity when leaving the hand, crossing the plate, or somewhere in between.  Which is it?

John, I don’t know why Nolan Ryan’s 100.9 is considered the record.  Last summer Baseball Analysts did an article on that game.  There was a lot of hoopla and speculation about how fast Ryan would be, because people hid little fram of reference before that.  They had scoreboard photos from the game, and the fastball results were all over the place.  The highest pitch was 100.9, and that’s all that people remember now.  I’ll try and find a link to that.

Here it is:

http://baseballanalysts.com/archives/2006/09/remembering_the.php

Tope speed by inning was only 87.6 in the first.  Later in the game:

6th: 91.3
7th: 96.7
8th: 94.1
9th: 100.8

Seems like there was a lot more error in those measurements than what we see today.

Guy ... the readings are supposed to be as the ball leaves the hand. Shoddy gun positioning and different release points mean that in reality most pitched ball are read 5ft or so from the point of release.

Rally ... I think the Ryan pitch is the record because the gun was specially set up calibrated and verified for the occassion. You top speed by inning data is interesting—if true, that makes the 100.9 look questionable perhaps.

The game has Ryan listed at 100.8, and its early September 1974.  Almanac has his record listed at 100.9 on August 20th.  I couldn’t tell from the linked article if there was any difference in method used.

In the article there’s a link to a photo of the stadium scoreboard, which shows top pitch for each inning.

I’ve always wondered why we can’t use video and film to determine how fast earlier generations of pitchers (Ryan, Feller, Koufax) really threw.  We know the distance from home to the mound hasn’t changed.  If there’s a way to determine with some precision the images/second, and thus time elapsed from release to bat or catcher’s mitt, then we should be able to determine pitch speed.

That should be easy enough, no?  Play your VCR in “slow motion” and count how many “ticks” from release to contact in a 2006 World Series game with the fastball recorded by Gameday.  So, now you know how to translate ticks to MPH.  Make sure you do it for 3 or 4 pitchers, preferably 2 or 3 per fastballs per pitcher.

Then, pop in your old-time World Series tapes, and count how many VCR slo-mo ticks each pitcher takes.  (Doesn’t apply to Sid Fernandez who had the strangest release point.)

You can also try to get that Bob Feller v Motorcycle and do the same thing.

You can’t get any precision that way.

Assuming film with 24 frames per second (not sure about time frame here), here’s the MPH for a given # of frames to hit the mit:

12 = 75
11 = 82
10 = 90
9 = 100

The best we could do with Koufax is guess he threw 95+ if his fastball takes 9 frames more often than 10.  I don’t think this tells us anything we don’t know.

Assuming 24 frames per second, then if you’ve got a large enough sample, then you can get your precision that way.  So, my Feller example won’t work, but it should work for a World Series game pitched by Koufax.