THE BOOK--Playing The Percentages In Baseball

Wow. Very impressive and thorough work as always, Tom.

I’ve known about the ball for a while, and that rubber ring inside of it. I wasn’t aware of just how large of an effect 8.7 feet would have on the flight of the ball. When exactly did MLB move its ball-making site to Costa Rica? I know it was somewhere else before (I think Haiti, but I’m not 100% sure), and they were afraid of Civil War destroying it so they moved.

Thanks.

Can’t help you on the timeline though.

And while my focus was on the ball, it could also very well be the bat too.  This graphic here shows 8.7 extra feet per around 1.5mph of extra bat speed:
http://www2.jsonline.com/sports/brew/allstar/jul02/homer70502.asp

With the new bat designs, that’s easily plausible.  That could also explain 1992-1994 happening over 2 years, as maybe only half the guys getting new bats, or still working their way through the old bat.

Or both bat and ball.

Still, I can’t shake the feeling that there’s got to be something about the players themselves.  It’s my understanding that pitchers throw harder these days than in years past, and that the harder it comes in, the faster it goes out.  I’ve read that leg size/strength has a big impact on HR power, and I’m pretty sure guys these days have bigger legs than our favorites of the 80’s.  Not to mention bigger arms too.  They train more often and more intelligently (due to research, more knowledgeable trainers) than their predecessors.  I can’t believe that using the same ball would put them on par with each other.  I guess my conclusion from this is that pitchers have gotten better at keeping hitters from making great contact.  So perhaps that depresses HRs as much as the hitter improvement would otherwise increase them, and you’re left with a wash where the only difference that has no other counterweight is the ball.

Mike,

It sounds like you have completely bypassed the data in the article.  Put aside your personal feelings here, and deal with the article, and try to poke holes in it.  Otherwise, accept the data.

I’ve always been a proponent of the “it’s the ball, stupid” theory for exactly the reasons Tom outlines here - the change was so dramatic over two seasons that it’s hard to imagine it could be explained by increased workouts and/or steroids (which, if they have any measurable impact, would probably occur over a longer period of time as players gradually begin using).

On the other hand, one thought I’ve had is that while the data (home run rates, or overall scoring) jump in ‘93 and ‘94 and then plateau, there could have been other changes that masked a more gradual increase.  We know that even across a whole season, variations like weather can end up not evening out.  Suppose instead of these rates (from Tom’s article):

1990 2.8%
1991 2.9%
1992 2.5%

1993 3.1%

1994 3.7%
1994 3.7%
1995 3.6%

We saw rates like this:

1990 2.8%
1991 2.9%
1992 3.0%

1993 3.1%

1994 3.4%
1994 3.7%
1995 3.6%

All I did was adjust the numbers in 1992 and 1994, but now the pattern is much more gradual.  What if these altered values are the “true” home run rates for those years?  That is, what if conditions such as weather made it look like a 2-year jump when in fact there was a more gradual rise in home run rates if we could somehow take weather out of the equation?  The fact that adjusting the numbers for two seasons completely changes the look of the pattern makes me wonder if the apparent two-year jump isn’t as significant as looks to the eye.

The weather could certainly be a culprit.  I don’t think I discounted that as a possibility.  I only concluded that expansion, parks, and conditioning of the players were not responsible. 

The climate, the umps, the bats, and the balls are still in play.  The others are not.

You might be able to determine if the climate is in play if you looked at minor league data as well. 

You can also look at each park.  Reasonably speaking, it’s not like every city in the country will be affected to the exact same degree by the temperature and wind.  So, you can look at it from that aspect.

As well, in cities where minor leagues and major league share facilities in close proximity, you would expect similar effects.

So, all I did was move this discussion away from players, parks, expansion, and now, someone else can hopefully pick up the pieces.

Pinto also posts:
http://www.baseballmusings.com/archives/024843.php

I talked to the manufacturer in 1993, and they told me at that time they were producing the most consistent ball ever. What they needed to do, however, was produce the mid-range ball, not the high range ball.

Bill James, behind the wall:
http://www.billjamesonline.net/ArticleContent.aspx?AID=584&Code=James01001

Bats have been evolving in their design since baseball was invented, but there has been more evolution in bat design in the last fifteen years than in the previous hundred years.

Think about it:  When was the last time you heard about a corked bat?  Nobody uses corked bats anymore, because the legal bats now are much better for the hitter than the corked bats of bygone years.  There have been at least three large changes in bat design since the early 1990s:

1) The bat wood is kiln-dried to increase the strength-to-weight ratio.

2) The bat, once made, is compressed—systematically crushed—to make the surface of the wood harder.

3) The bat is covered with multiple layers of lacquer, to make the surface even harder. 

I do not know how much this revolution in bat design has contributed to the hitting explosion that began in the early 1990s.  I don’t think anyone knows. 

By the way, if it’s not apparent, this is another article in the WOWY approach (with or without you).

The difference between this approach and other approaches based on “adjustments”, is that these adjustments are made at a group level, completely ignoring the possible unique interplay of the parameters.  For example, it would be foolish to even presume that Coors would affect Juan Pierre, Dante Bichette, and Larry Walker the same way.  “Adjustments” would force you to do so, unless you add additional parameters like weak-armed, etc.  But, these adjustments, even these extended adjustments, would force you to know even more about the data that you are looking at (Pierre v Bichette is obvious… other data might not be).  For example, maybe a tall 1B might have more of an effect on SS throws if the SS is an erratic thrower.  Maybe a good fielding 1B has more value with a weak-armed SS, etc, etc.

WOWY establishes the specific interplay of all these variables, like Raines v Drabek at Olympic Stadium.  I can’t know, and neither can anyone know, how much Raines is or is not affected against Drabek at Le Stade, specifically.  So, we shouldn’t presume so.  You simply treat that as something unique, and not as 3 parameters, each of which needs some global adjustment.

There is of course the downside with WOWY (you need the sample size).  But, in things that I am studying, and others, in most cases, that’s not an issue.  So, I urge researchers to pay more heed to this.

Great detective story, Dr. Tango.

Does anybody have a cache of “historical” MLB bats available for testing?

On the umpire theory, can you use Retrosheet to test for same pitcher-batter-umpire combinations? Or just for umpire effects (HR/contact) within and across seasons?

Interesting—certainly plausible. Haven’t I read that bats also break far more than they used to because of the thin handles? Personally, I’d like to see a rule that charges the batter with a strikeout—or at the very least a foul ball—if he puts his ball and any part of his bat into fair territory at the same time. Jagged, flying pieces of wood should not be part of the game.

Related article here (with the baseball part at the end):
http://lyflines.blogspot.com/2005/04/steroid-era-of-major-league-baseball.html

As well as here, which used some material from Eric Walker:
http://www.bugsandcranks.com/the-clubhouse/barry-bonds-deserves-an-apology-it-was-the-baseball/

The article by Jay Jaffe of Baseball Prospectus addresses changes in the ball:

http://baseballprospectus.com/article.php?articleid=7167

I have the 2007 and 2008 data.  The HR rates changed quite a bit from 2006 to 2007.  Using the same matched-pair process (same pitchers facing the same batters in the same parks in back-to-back years, using the minimum of the two PA), in 33,290 PA there were 1036 HR in 2006 and 917 HR in 2007 for a 11% drop.  This was preceded in 2005/06 that had a 10% increase.

To recap that time period: from 2005/06, the control group had a 10% increase, while league-wide showed a 6% increase.  From 2006/07, the control group had a 11% drop, while league-wide showed a 9% drop.

Between 2007/08 there were an identical number of HR hit among the control group and a 1% drop league-wide.

***

Among the control group, the biggest jump in HR were:
1968/69: +43% (26%)
1976/77: +57% (48%)
1978/79: +26% (14%)
1981/82: +33% (24%)
1992/93: +20% (22%)
1993/94: +19% (15%)

Remember, this is the same batters, pitchers, parks.  The league-wide changes are in parens.

Typically, the matched pairs comprise 25% of the league total, which is a healthy sample.

From 1976 to 1982, the HR rates among the control group doubled.  League-wide, the HR rate increased by only 36%.

(Note all my numbers exclude pitchers-as-batters.)

So, there was a severe imbalance here toward pitchers.  Expansion, the five-man rotation, the bullpen all happened in this time period.  The result was a severe depression of HR.

The other place I found interesting was strikeouts.  Doing the same thing, K rates dropped around an average of 4% each year with the control group, which is quite substantial.  League-wide though, the K rate increased by 1% each year.  This means there’s been an influx of power pitchers and power hitters.

Indeed in every single year in the last 50 years, the K rate increased year-over-year league-wide more than the control group and it is by far the biggest change in talent distribution in MLB.

Thanks for the update—very interesting data.  Two thoughts:

1) It appears the fluxuation in HR rates is much greater than we’d expect from binomial variation.  We would expect to find changes of +/-5% in 95% of the seasons.  Yet much larger changes are routine.  And the fact that the control group always moves in the same direction, and usually a similar amount, confirms that these changes are non-random.  Weather could explain a bit of this, but it sure looks like the quality control of the balls has been pretty poor.

2) The one factor not held constant by your WOWY method is player age.  For most purposes, with large samples, it shouldn’t matter:  some pre-peak hitters are improving, others declining, and same thing for pitchers.  But when looking at Ks, I think we expect pitchers to basically have a secular decline starting from age 21, whereas the net age impact on hitters is probably zero.  So there should be some decline in your control every year (but 4% does surprise me).

ooops, that last post was mine, addressed to Tango.

Funny.  I’m looking at age as we speak.

I was hoping that the aging would cancel out.  That is, my batters are one year older, but my pitchers are also one year older.  If both gain or lose the same amount in wisdom / physical skills, then it would cancel out.

However, this doesn’t seem to necessarily be the case, at least for strikeouts.  The pitchers’ propensity to drop in K starts earlier than the hitters’ propensity to increase in K.  And, pitchers’ K rates may drop faster in their 30s than hitters’ K rates increase in their 30s.

I’m staring at the data right now, and it’s pretty overwhelming right now.

***

Oh, I also matched on whether the pitcher and batters were starters or not.  So, when I match Raines to Mulholland, I treat Mulholland/starter separate from Mulholland/relievers.

Sample size drops to around 20% of the total population, which is still great.

IIRC, batters decrease their SO% up to about 28-29, then it starts increasing.

I haven’t looked at pitchers, but they may peak earlier.

On my site, I have plenty about the aging patterns of hitters and pitchers.  But, those did not control for who they were facing.  In my case now, I am looking only at hitters who faced the same pitchers year-to-year (in the same parks).

Once I do that, I get some very differnt results.  Indeed, the “aging” that we are talking about traditionally is not so much aging, but adapting to a new set of opponents the following year.

If for example the pitchers in 2009 were the same pitchers as in 2008, and pitched in the same proportion, the older hitters would do much better than we’d have otherwise expected.

Tom - your article was THT puts forth the theory of a juiced ball, one that has continued for 16 seasons. I concur, and last week heard something which caught my attention. Fred Lynn was being interviewed on XM’x MLB channel, and made a comment about how often these days balls are tossed out, replaced with a new one, compared to his days when the same ball would remain in the game much longer.

Could the ‘juiced’ ball actually be a ‘fresh’ ball, that is not left in the game long enough to get scuffed and battered?

It doesn’t seem that that could happen overnight, could it?  It’s possible that there was an edict that says to replace balls under certain circumstances between 1992-1994.  I just think it’s more likely that the ball (or bat) manufacturing process changed subtlety enough over that time period such that 1994-2009 can be considered one era, distinct from the 1992-and-earlier era.