THE BOOK--Playing The Percentages In Baseball

The more interesting question is whether this suggests a “real” change (for the worse of course) in true talent or ALL players having a banner good or bad year also have a large increase or decrease in air ball distance?  I suspect that it is a little of both (for example, if a player is having an unlucky year in the HR department of course his average air ball distance is going to be down).  What about Giles?  There are two players in baseball that keep playing every day, that have been absolutely terrible, like single A terrible, and who are both suspected of being “done” for various reasons, some of which are not too flattering.

Good question, MGL.

Garrett Atkins is another guy who has “fallen off the cliff” and is seen as having the big fork in the back.  Granted, he didn’t have as far to fall as Giles and Ortiz.

Someone (maybe me) should look at the speed off the bat numbers for Ortiz. I think that would be the definitive answer.

Of course we don’t have historical data from previous seasons, but you could compare him to other players. If he’s hitting the ball like Rajai Davis then you know he’s in trouble.

I mean: from HITf/x (got ahead of myself there)

Someone (maybe me) should look at the speed off the bat numbers for Ortiz.

Harry did here (in the comments):
http://www.beyondtheboxscore.com/2009/6/6/900817/graph-of-the-day-average-batted#comments

Well, that is d-mn fascinating.  I need to put BtB in my Google Reader view immediately.

How is the speed off the bat (SOB) affected by the angle off the bat?  Say that you get “all of it”, or you get “half of it” (popup), what is the different SOB, on an otherwise same pitch / same swing?

How is the speed off the bat (SOB) affected by the angle off the bat?  Say that you get “all of it”, or you get “half of it” (popup), what is the different SOB, on an otherwise same pitch / same swing?

I assume you’re asking about the vertical angle of the batted ball.  If you define getting “all of it” as the maximum possible SOB that occurs when the center of the bat and center of the ball are aligned, then…

...when the offset between the center of the bat and center of the ball is 0.5 inches, a line drive with 99.5% of the max possible SOB results.

...when the offset between the center of the bat and center of the ball is 1 inch, a fly ball with 98% of the max possible SOB results.

...when the offset between the center of the bat and center of the ball is 1.5 inches, a popup with 94% of the max possible SOB results (the batted ball is now going almost twice as fast vertically as it is horizontally).

...when the offset between the center of the bat and center of the ball is 2 inches, a popup with 88% of the max possible SOB results, and it becomes very difficult to keep the ball fair.

(All of this assumes the ball is hit on the barrel of the bat.)

With Giles and Ortiz, I am trying to separate a player who is simply having an unlucky season (as well as a change in true talent) from a player who is truly “done.”

In Papi, we have a player who many people say had “old player” skills to start with, and at least one scout, according to an article I read the other day, who says that, “There is no doubt that Papi is older than we think - that his birth certificate is a fake.” I also don’t think anyone would be even remotely surprised if we found out that Papi was juicing.  Before I am accused of unfairly arousing suspicion about him, what I mean is that I don’t think we would be remotely surprised if we found out that ANYONE was juicing, let alone a slugger, and one who al of a sudden became a slugger late or relatively late in their careers.

In Giles, we have, according to many San Diegans, a player that was heavily juicing, as was his brother who fell off a cliff and never came back (their mother was a body builder for whatever that is worth).

Atkins, I have no idea.

Plus, like David Gassko and his 29 underrated or overrated players, I feel like I can watch a player and figure out to some extent if he is just having a fluke bad year or not.  (Keep in mind two things:  One, when a player has a fluke bad year, his true talent estimate STILL goes down a lot - IOW, if before the season started, we thought it was .800 and he then performs at .550 for a season and it is a “fluke,” our new projections is a lot lower than .800 - just not any lower than any other estimate that includes recent performance with the appropriate weighting.  Two, when a player is having a bad, random fluctuation (or good one), he is going to “look” worse (or better) than his true talent really is.)

Given that, when Andrew Jones was swinging and missing at every pitch under the sun and simply looking like he had no clue at the plate, TWO YEARS AGO, I think it was pretty obvious to any knowledgeable baseball who watched him that something was wrong.

Similarly, when I watch Brian Giles, it looks to me like he has no bat speed or pitch recognition skills anymore. I would bet a lot of money that he is going to under-perform his normal projection for as long as he is allowed to play from here on in. I have not watched enough of Papi to be able to say anything similar, but according to “scouts”, he appears to indeed be “done.”

Again, Atkins, I have no idea and I have not read anything about him one way or another. There are of course going to be X percentage of players who have terrible spates of performance by chance (little or no change in true talent), no matter what.  The trick is to find clues to help you one way or the other, although identifying those clues can be tricky.  A player can be having a random, bad fluctuation in many aspects of his performance.  Even if a player is just being “unlucky”, if that bad luck is 2 or 3 SD away from his mean, it is unlikely that the only thing that is happening is hard hit balls are being caught, or he happened to face tough pitching or just tough pitches in general.  It is likely that he happened to have been “swinging” really badly along the way for whatever reasons.

I assume you’re asking about the vertical angle of the batted ball.  If you define getting “all of it” as the maximum possible SOB that occurs when the center of the bat and center of the ball are aligned, then…

That’s why the first thing that I did when I got the data was calculate vXY, the component of the speed off the bat that is in the XY plane.  How fast your pop up or high fly ball is going up toward the sky is not particularly important; how fast its getting to the area where it is going to either be caught or a base hit is aand that’s what vXY gives you.

The other interesting thing from my first pass through the data is seeing how high average hitters have an average vertical angle that is less than the high home run hitters.  There is definitely a trade off being made, either consciously or unconsciously, that some batters are sacrificing some chance for a home run for a greater chance at a line drive hit.

Peter: I think they are very conscious.

SoB: so, SoB by itself it not important.  We also want to know the angle off the bat (AoB).  The combination of SoB and AoB will give you the horizontal SoB (HSoB).  And HSoB is the interesting number we really care about.

Average HSoB (in mph) for batters with >=20 batted balls in Apr-09:

85.4 Ry.Howard
84.7 Mi.Cabrera
83.2 Fe.Lopez
82.4 Ja.Saltamacchia
82.1 Ar.Ramirez
81.9 Mi.Napoli
81.3 Fr.Lewis
81.0 Jo.Votto
80.6 Ro.Paulino
80.5 Cr.Monroe
80.0 Al.Pujols
79.9 Jo.Morales
79.8 Da.Wright
79.5 La.Berkman
79.3 Ra.Ibanez
...

67.2 Ic.Suzuki
67.1 Ke.Kouzmanoff
67.0 Ro.Johnson
67.0 Ka.Matsui
66.9 Da.Ortiz
66.7 Al.Casilla
66.6 Er.Hinske
66.6 En.Chavez
66.6 Vl.Guerrero
66.6 Ra.Furcal
...
64.1 Ga.Matthews
64.1 Di.Navarro
63.4 Da.Murphy
63.4 Ni.Punto
63.3 Jo.Anderson
62.1 Ke.Johjima
61.4 Ch.Ianneta
61.1 Em.Burriss
60.6 Ed.Encarnacion
60.6 Ch.Snyder
60.5 Ca.Maybin
60.1 Er.Chavez
59.7 Er.Byrnes
57.6 Wi.Taveras

Average HSoB for the 435 batted balls hit by pitchers in April was 53.5 mph.

Mike,

Thanks for posting this. I really need to find some time to did into the data. There’s no May data yet?

Mike: fantastic!

So, Ortiz hits like Endy Chavez and Kaz Matsui.  And see how low Vlad has fallen as well.

Peter:

I agree with you reasoning about the importance of HSoB, but you can’t take it too far. Say you had two hitters who had the same SoB and one averaged ASoB of 0 deg and the other 20 deg. The 20 deg hitter would be far more valuable than the 0 deg hitter even though his average HSoB would be about 6% lower (assuming I’m doing the calc correctly). He would hit for more power, and I would expect also a higher BABIP.

The reason being that once the ball hits the ground it loses a lot of velocity. Plus the low balls can be fielded by the infielder for one-hoppers or line drive outs. Line drives have the best BABIP and hit value and sorry to state the obvious. Although, straight up HSoB should be decently correlated to BABIP - so it’s certainly a reasonable approximation.

I would bet there’s a maximally efficient launch angle (~15 deg I would guess) and that you could predict BABIP based on the average SoB in that plane.

You are of course right on that any pop flies (I would guess over 45 deg) it doesn’t matter except in the case of a bloop hit in which case lower HSoB speeds are better.

Regards,

Matt

Check out Hardball Times tomorrow, I dove into the hitf/x a little more and some of the topics in this thread will be touched on.  I use SLGCON and mix in the stringer’s classifications for some context and a jumping off point.

Matt,

The optimum vertical launch angle for a well-struck* ball appears to be around 11 degrees where batting average peaks around .900.  Batting average is >.400 for vertical angles between -3 and 30 degrees.

The optimum vertical launch angle for a poorly-hit* ball appears to be around 17 degrees where batting average peaks around .750.  Batting average is >.400 for vertical angles between 10 and 25 degrees.

*I defined well-struck balls as anything with >80 mph initial speed in the plane 11 degrees above vertical, and anything with a lower initial speed in that plane as poorly hit.

To refine my previous post #16 a bit, I divide the poorly-hit balls into two categories, medium-hit (55-80mph), and poorly-hit (less than 55mph).

The optimum vertical launch angle for a medium-hit ball is around 17 degrees where batting average peaks around .900.  Batting average is >.400 for vertical angles between 9 and 25 degrees.

The optimum vertical launch angle for a poorly-hit ball has two peaks.  One is beating the ball into the ground at around -45 degrees where batting average peaks around .250 or maybe .300.  We don’t have a lot of data from HITf/x for this peak, so its exact shape is a bit murky, but batting average hits a minimum around .050 near -15 degrees and starts to rise above .100 as you go below -30 degrees. 

The other peak for poorly-hit balls occurs around a vertical angle of 25 degrees where batting average peaks around .250.  These are the very soft flies or bloops.  Batting average is >.100 for vertical angles roughly between 5 and 45 degrees.  For vertical angles greater than 55 degrees on a poorly hit ball, the batting average is below .020.

Here are a couple graphs to illustrate my points under the theory that a picture is worth 1000 words.

Here’s a look at the April 2009 batted ball initial speed distribution broken down by batted ball type for a few selected players.

The second image link in Post #18 should point here:

I published a list of average batted ball speed (in the plane 11 degrees above horizontal) for all batters (20+ BIP) here:

http://www.hardballtimes.com/main/blog_article/which-batters-hit-the-hardest-balls-in-april/

I removed bunts from consideration for that list, which probably presents a more accurate picture.  As I mention there, without bunts the average SoB(11deg) is 75.1 mph for hitters and 66.1 mph for pitchers hitting.  David Ortiz checks in at 72.9 mph.

Dave Allen shows that Ortiz’s distance numbers are now back up:

http://www.fangraphs.com/blogs/index.php/is-ortiz-powering-back-up/

What that shows me is that this newer analysis is subject to the same random variation as traditional stats.

Show that the distance of Ortiz’s hits is not all that far for two months, and you’ve shown he’s in a slump, but have not told us anything about a change in talent level.  It doesn’t mean anything more than the fact he had 1 HR in 150~ at bats.

Sometimes I think people get too excited by new analysis that wasn’t available until recently and read too much into it.  Ortiz was slumping.  We all knew that.  His distance of flyballs did not tell us anything special about his sudden loss of talent.  It was just additional data telling us what we already knew: He was underperforming his ability.

Exactly.  In the terms of Hit F/X, with the small sample (April) data, its hard to infer anything from the results.

Rally: I agree with your basic premise.  The question is how much of it is noise.

For example, if you just look at a pitcher component ERA you need say about 300 PA to regress his performance 50% toward the league mean.

But, if look at this K/PA rate, you need say 100-150 PA to regress 50%.

So, for distance, we need to figure out how much noise there is.  I would suspect that the less “moving parts” there is, the more real and persistent is what we are seeing.

#25, exactly.  The more granular you get with the data, the less noise you have, generally.  That does not mean no noise, but less noise is important in terms of inferring changes in talent level.  If a player is “slumping” but his air ball distance is the same, we can probably infer more noise than if the same player is in the same slump but his air ball distance has decreased.

The danger, and I have commented about this before, is concluding that talent had changed significantly just because the more granular data supports that notion.

Basically, we need to know what we are working with. For example, how much noise is there in average air ball distance as compared to, say, HR per PA or per fly ball?  Before we start drawing conclusions about player talent or a change in that talent, we need to determine things like that.

And there were reports of scouts saying that, “Ortiz was done.” How much weight do we put on that kind of “data?”

Oh, and one more thing:  Of course (likely that is), Ortiz fly ball distance was going to “go back up.” It was going to regress towards his historical norm and the league average for his type of player just like EVERYTHING and ANYTHING else…

How much does selection bias play in the expected regression of any player?  Meaning if a player is not expected to regress to his historical norm or even league average (say management thinks he’s completely done and a much worse player now, worse than replacement level), then management would “select” not to play him any longer, and this particular player’s stats would cease to exist in the database of players which researchers draw from to test whether players regressed to the norm or not.  On the other hand, if management did expect the player to regress, then his stats would continue.  So, in other words, is it possible that selection bias makes the “regression to the mean” argument look better?  Don’t get me wrong, I do like the regression to the mean argument and agree with it, just saying that on the downside of a player’s career, the regression to the mean argument may look stronger than it really is.

King, sure.  You would be a little surprised, I think, how small the percentage of players who drop out and don’t continue to play are.  A player who used to be good gets a lot of rope to hang himself if he starts performing badly, as you can see with players like Ortiz and Giles. So, I agree that there is an effect, but I don’t think it is a large one overall.  It is one of the problems we face in computing aging curves, but there we are looking for subtle effects, so the selective sampling process you describe is problematic.