THE BOOK--Playing The Percentages In Baseball

Whatever you want to call it, I think the “trend” (as much as I typically hate that word) is clear.

However, I think he is jumping to conclusions, when he states with some implied certainty that fielders get weary as the season wears on.

My explanation would be that the warmer weather causes faster (which it does of course), and thus harder to field ground balls.  For example, before the humidor, Colorado had by far and away the lowest ground ball out percentage in baseball because the air was thinner and ground balls travel faster in thinner air.  Warmer air is thinner and that is why all batted balls travel faster in warmer weather.

So why the drop in September?  It is likely that it is because of all the September call-ups who are younger and faster and are beating out some of those infield ground balls.

He really needs to look at the percentage of ground balls that go through the infield to test that theory.

He could be right about fielders getting weary, but it is by no means a given, I don’t think.

I agree, it’s difficult to separate the two. 

He really needs to look at the percentage of ground balls that go through the infield to test that theory.

MGL, what do you mean by that?  How is that different than what I did?

Also, I figured that if the summer weather was the cause of harder ground balls, we’d see it reverse course at the end of the season.  It’s actually a pretty steady decline for the most part; I didn’t see any evidence of reversal.

Why not use the playoffs to test MGL’s callup theory?

Bill: excellent idea.

You can also just add temperature as a variable and include that in a regression.

I apologize for trying to comment on an article I’m not paying to read, but it appears that the out rate on fly balls increases slightly as the year goes on (I am just breaking the data into calendar month buckets.) It’s not clear why fatigue should affect only infielders fielding ground balls. Nor does there seem to be a decline in line outs caught by infielders over the course of the year. If their fielding was affected by fatigue, I would expect that too.

I doubt that callups are a significant part of the explanation. I don’t think there are enough AB there to move the rates as much as Ben observes, given that only a fraction of the callup pool actually fits that younger/faster profile. The decreased out rate on ground balls over the year looks like it is associated with increases split fairly evenly between hits to the infield and hits to the outfield. Errors show a different pattern (higher at beginning and end of year with a dip in June/July/August) but compared to hit rates, error rate is still is only a minor contributor to overall out rates on ground balls.

I looked at a fairly random selection of 15 full season players who played each year in Ben’s interval (3000+ GB in each monthly bucket), and this group of batters actually showed the pattern he observes in even stronger fashion. September callups, at least on offense, don’t help explain that. Also 2003 doesn’t fit the seasonal pattern of 2004-2009 very well, so Ben’s sample might be overstating whatever effect there is.

This is not the kind of question which can be convincingly answered by competing aggregations of the data. It will be necessary to find the roots of the pattern by looking at individual players, then posing hypotheses.

And it may not have been expressly pointed out elsewhere, but this apparent seasonal pattern on ground ball rates explains the observation in the other “John Dewan speaks” thread: why a rolling 1 year average for plus/minus, mostly comparing out rates in late 2009 (when rates are “naturally” low, to early 2010, when rates are “naturally” high) makes a lot of 2010 to date infielders look above average…

Good stuff Joe.  If the infielders are getting .005 more outs per BIP because of the seasonality, that’s .004 runs times say 13 GB per game, or .05 runs per game.  After say 2000 games for the 30 teams, that’s +100 runs.

***

Joe: can you send me an email.  I have a question regarding the ID mappings.  tom~tangotiger~net

A handful of questions:

1) What does a similar chart for line drive rates look like?

2) What do GB per BIP rates look like month-to-month?

3) How about GB hits per all BIP hits?

Average game-time temperatures usually peak some time in early August, but they don’t really start to drop sharply again until to mid-to-late September.

Avg game-time temps 2007-2009:
Apr 64 F
May 69 F
Jun 76 F
Jul 77 F
Aug 78 F
Sep 73 F
Oct 67 F

Ben, I used BIS data to look at “hard hit” groundballs as a percent of all groundballs and found the following pattern (looking only at 2004-2009 as you did in your column):

19.5% = First 60 Days of Season
20.0% = Second 60 Days of Season
20.9% = Third 60 Days of Season

Given that the BABIP on hard hit groundballs is .553 (compared to the BABIP on all other groundballs of .189), this could definitely be a contributing factor to the pattern that you illustrated.

-Ed

Well, if weather is a factor, one could look at Tampa and the retractable roof stadiums (Minute Maid, Miller Park, Safeco, Chase Field, and Rogers Centre).  In my experience, Minute Maid is almost always closed in the summer.

"Given that the BABIP on hard hit groundballs is .553 (compared to the BABIP on all other groundballs of .189), this could definitely be a contributing factor to the pattern that you illustrated.”

It may be hard to distinguish chicken and egg here.  I suspect balls that get through the IF are more likely to be coded as hard hit.  Does BIS ever compare the scorer ratings with an objective measure of ball speed?

Does BIS ever compare the scorer ratings with an objective measure of ball speed?

I thought that BIS was now collecting hang time and time through infield on GBs.  Isn’t that supposed to supercede observational estimates of hit ball speed?

My first guess is that batted ball speed increases slightly through the season. In 1993, coinciding with an increased HR rate, BABIP increased too. I think this is a sure sign of a more resilient baseball. Warmer temperatures will make baseballs more lively, but I wouldn’t discount other factors. One possibility is that hitters become more familiar with pitchers as the season wears on and are actually making better contact. If so, this would show up in other ways, such as a shift in strikeout and walk rates.

Right, error rates are higher in the colder weather, so that balances out somewhat the temperature effects, if you are considering errors as “non-outs.” Are you Ben?

Ben, I mentioned the balls through the IF thing in order to exclude infield singles, which will adjust for the effect that I postulated - that in September, there are more IF hits due to the faster and younger call-ups.

My first guess is that batted ball speed increases slightly through the season.

There’s not any evidence of that in the April 2009 HITf/x data, fwiw.

Dividing that data set into four pieces of about 4000 batted balls each, here’s what I found.

April 6-11, avg temp 60 F, avg batted ball speed 81.3 mph
April 12-17, temp 60, bb spd 80.8
April 18-24, temp 65, bb spd 80.3
April 25-30, temp 71, bb spd 80.6

Re #16, I realized there may be two or even three competing effects there.  (1) Temperature should increase the COR of the ball and thus should lead to higher batted ball speeds with increased temperature.  (2) The pitchers may also be still ramping up their game coming out of spring training and may be behind the hitters in this respect.  I’m not aware of any empirical evidence about this.  (3) Pitchers throw harder in warmer temperatures.  It’s unclear if this has any significant effect on batted ball speeds.  (My initial look says no.)

RE: callups, note that players who debut on or after September 1st in a given year only account for about 2% of all Sept/Oct balls in play, so I think that they would have to bat 50 points (.050) higher than the league average on ALL of their groundballs to move the overall league number up by just 1 point (.001).

As a group, debut rookies in Sept/Oct do seem to have a higher BABIP (.275) on ground balls, but it doesn’t really seem to have a significant impact on league aggregate BABIP.

I do recognize that “callups” would include more than just rookies making their debut in a given season after Sept 1st, but those players could also have some PAs in the earlier part of the season that would theoretically cancel out the late season BABIP effect (I didn’t bother to do the actual pre/post on those types of players).

I would guess that the ground is drier, and therefore, harder and faster, late in the summer.  I’m not sure that it is testable, though examining Toronto, the Metrodome and Tampa may give a clue.

Good points, all.  Not sure where everyone is getting their BIS data.  The percentage of soft-hit grounders goes up during the course of the season as well, though not by as much. 

I assigned values to all grounders (soft = 0, medium = 1, hard = 2) and took the average by month:

March 0.995
April 1.034
May 1.052
June 1.047
July 1.050
August 1.059
September 1.055
October 1.047

So you’re right- there is a slight upswing in ground ball velocity that corresponds with the decline in GB out rate (MGL, I’m counting errors as non-outs). 

However, how would you explain the last chart in the article that shows Outs/GB decreasing the longer a team goes without a day off?  It could be that Day 20 is later in the season, thus there are more hard-hit balls on Day 20 than on Day 1.  I don’t think the full magnitude is explained, though. 

Thoughts?

Sorry, I can’t access the article.  But when you look at teams without a day off, are you controlling for the fact they will be forced to use inferior pitchers somewhat more often (who give up harder to field GBs)?  And are you controlling for the fielders, as these teams also probably use bench players in the IF more often (although that may or may not reduce fielding quality).

Mike/16 -

“April 6-11, avg temp 60 F, avg batted ball speed 81.3 mph
April 12-17, temp 60, bb spd 80.8”

In a data set that tight by date, you’ll have sample size issues of course.  April 6-11 2009 were days 1..6 of the season, in which teams would get 2 starts from their #1 pitcher.  If that pitcher is throwing harder fastballs, that may explain the higher velocity in the first week.

These are groundballs, could the ground itself be creating that effect? Infields, especially in really hot weather regions, get harder as the season goes along, perhaps slowing the ball less.

This got me thinking - say a pitcher like Jiminez has an incredible GIDP rate due to high GB rates but the infield defense behind him is also above average.  How does Total Zone treat the extra double plays that he induces?  Does his team defense get all of the credit for the extra double plays turned? Does Jiminez’ WAR see the benefits or is it somewhere in-between?