Prospectus Hit and Run: Running Afoul

Last time around, after discussing how the baseball itself may have changed in a manner that helped to boost home run rates over the past two decades, I took a look at the myth of the shrinking ballpark. To recap, the notion that the stadium construction boom that’s taken place over the past 20 years has left us with a game full of bandboxes is actually a false one, at least when it comes to fence distances:

MLB   1990    2007  Change
LF   329.6   332.0   2.4
LCF  375.5   376.6   1.2
CF   404.9   404.9  -0.1
RCF  376.0   377.6   1.6
RF   329.1   329.3   0.2

With the exception of straightaway center field, those distances have increased by a hair through the 21 new ballparks that have come into play since 1990. However, it’s true that park capacities have dropped about 10 percent in this time, a trend that will continue as the Yankees, Mets, Athletics, and Twins move into new parks able to seat fewer paying customers in the next several years.

These findings drew a considerable amount of response from our readers, who made a couple of good points that are worth exploring. The first one relates to the 1995 introduction of Coors Field, which has deeper fences to help offset the extra distance the ball carries at high altitude. A couple of readers worried that the addition of Coors distorts the average fence distance changes too much, so I’ve recomputed the averages to show its impact:

      2007  Coors  2007'  90-07  90-07'
LF   332.0   347   331.4   2.4    1.9
LCF  376.6   390   376.2   1.2    0.7
CF   404.9   415   404.9  -0.1   -0.4
RCF  377.6   375   377.7   1.6    1.7
RF   329.1   350   328.6   0.2   -0.5

The 2007′ numbers are the averages sans Coors, and the 90-07′ figure is the change from 1990 to 2007 excluding Coors. Note that its removal swings right field from a slight gain to a slight decrease, and that four out of the five measurements decrease by at least some fraction. (Also note that there are some rounded decimals in there which make the deltas look less straightforward.)

The take-home point here is that the balance is still in favor of increased fence distances, particularly on the left side (LF and LCF). That’s important, because about half of all homers are hit to those two areas, thanks to the lefty/righty distribution of hitters (58.8 percent of all plate appearances were from the right side in 2007) and the natural tendency to pull the ball.

In fact, a quick-and-dirty look using the Baseball-Reference.com Play Index shows that the increased distances may have an impact on home run distribution. In 1990, 54.4 percent of homers were hit to left field and left-center field. Last year it was 50.8 percent. The intermediate data is a little squirrely, likely a result of being integrated from multiple sources:

• In the years immediately prior to 1991, the location of more than 100 home runs per year is unaccounted for in the Retrosheet logs, but from 1991 on, that drops down to about 10-15 homers per year.
• The 2006 data at B-R doesn’t distinguish left-center field or right-center field; I corrected this by using harder-to-obtain data from BP’s internal database instead.
• The 1999 data from B-R shows 1,024 homers to center field; that’s up from 571 the previous year, and well above the 683 scored to dead center the following year. BP’s data is of no help here, lacking the left-center and right-center distinctions for that year.

As imperfect as the data may be, the trend is apparent. The percentage of left-side homers hasn’t topped 51.0 percent since Coors opened in 1995:

Ladies and gentlemen, my first-ever graph for a Baseball Prospectus article, and it’s ugly enough to put this graphic designer to shame; we’ll spruce things up next time. Anyway, the trend meshes with the finding that fence distances have in fact increased over the past two decades.

The second major point made by my readers is that the new ballparks may have resulted in a decrease of foul territory, which could be helping to drive up offensive levels, including home run rates, by creating more opportunities for balls to be hit into fair territory. Alas, this isn’t an easy avenue to explore, nor is it an exciting one, far closer to a methodical trudge through the cardboard box factory than a debauchery-laden weekend in Las Vegas. But metaphorically speaking, if we chase down some foul balls, we may catch a thing or two.

Start with the problem that foul territory measurements aren’t systematically reported; at best, the distance from home plate to the backstop can sometimes be found. Even the most definitive source of ballpark data available on the web, the KJOK Park Database (found via membership to this group), has major gaps in its data; the most recent season available, 2004, contains backstop distances for only 20 out of 30 teams, while for 1990 it’s just 12 out of 26 teams–virtually useless for our purposes.

Luckily, some time spent drilling in Google helped me find Clem’s Baseball page, which contains all kinds of information on stadia, including backstop distance figures or estimates for 66 current or former ballparks. With the caveat that there are significant differences between the existing KJOK data and Clem’s estimates, if we use the latter we find that backstop distances have indeed decreased over the years. Back in 1985 (a year I chose so as to encompass all of the modern stadium changes as well as the 1987 home run spike), the average backstop distance was 60.6 feet. In 2007, it was 56.3 feet.

Even if that means an overall decrease in foul territory–and no, I don’t know of anyone who’s bothered to systematically quantify that particular area–is that really having an effect on the matter in question by resulting in fewer foul outs and thus more balls in fair territory? To answer this, I looked at play-by-play data to see whether the wave of new stadia has changed the frequency of foul-ball outs. Because strikeout rates have gradually risen over the time frame in question–from 5.37 per nine innings in 1985 to 6.67 last year, or from 14.0 percent of all plate appearances in 1985 to 17.1 percent last year–I used foul outs per contact out. When I did, I found something odd:

Foul-out rates have fallen slightly in the grand scheme, from 4.57 percent of all contact outs in 1985 to 4.25 percent last year. That supports the notion that foul territory has decreased in recent years, thus creating more opportunities for home runs. But there’s a major dip in the data as seen in both leagues from 1993 to 1998. It’s unclear whether this is simply another squirrel running amok in our database, a result of the union of differing sources of play-by-play data, or a legitimate–if curious–trend. It’s worth noting there’s considerable stability over the three neatly-divided time periods of this data:

 Period    PctFoul  StDev
1985-1992   4.60     0.10
1993-1998   3.34     0.08
1999-2007   4.36     0.08
Total       4.19     0.51
Non-dip     4.47     0.15

To review, 1993 saw the introduction of Coors Field’s precursor, Mile High Stadium (no backstop distance recorded), as well as Your Ad Here Stadium in Miami (Joe Robbie until 1996, Pro Player from 1996 to 2005, Dolphin Safe Tuna more or less since then). Cleveland’s Jacobs Field and the Rangers‘ Ballpark in Arlington came along in 1994, Coors in 1995, Atlanta’s Turner Field in 1997, Arizona’s Bank One Ballpark and the then-Devil Rays‘ Tropicana Field in 1998, and Seattle’s Safeco Field in 1999. That’s a fair bit of turnover, and the trend has continued in recent years, but I’m not sure it explains the temporary dip. Nor, I’ll wager, do the truncated 1994 and 1995 seasons.

Looking at the data on a more granular level–every team, every year since 1985–doesn’t offer tremendously greater clarity. Consider the following:

• Of the 47 times foul-out rates dipped more than 1.0 percent in the same ballpark from one year to the next, 15 come in 1993, the year rates systematically fell. Similarly, of the 43 times foul-out rates jumped more than 1.0 percent, 15 of them come in 1999, the year rates rose again.
• Within a sample size of 601 contiguous park-seasons from 1985-2006, we find a moderate amount of year-over-year correlation (r = .61) in foul-out rates. That correlation, for some reason, suffers considerably if we excise the 1992-1999 data (the questionable 1993-1998 period, plus the contiguous year on each side), dropping to r = .41.
• Within the much smaller sample size of 17 instances where an existing team (not an expansion one) changed park from one year to the next, the correlation between foul-out rates is higher than the other two samples, r = .68. This opens up the possibility that foul-out frequency may be more a product of personnel–pitchers, hitters, and/or fielders who are more or less prone to produce such events–than of environment.
• There’s actually very little correlation between the estimated backstop distances and foul-out rates, and what correlation there is is negative. Using raw percentages for the 1985-2007 data, r = -0.1. Indexing the data in a manner similar to ERA+ (where 100 is league average, every increment away from 100 is the percentage above or below average; we’ll call the result FO+), the correlation is a bit stronger, r = -.15.

Even if raw backstop distance has nothing to do with the foul-out frequency, it’s quite reasonable to assume that the frequency is something that requires multiple years of data to show anything meaningful, particularly given that such events happen roughly one or two times per game. We can start to get a hint of the effect of the building boom on foul ball rates once we classify the parks into four groups:

Group 1: Ballparks of existing teams that have been constant from 1985-2007
Teams: ANA, BOS, CHN, KCA, LAN, MIN, NYA, NYN, OAK

Group 2: Ballparks of expansion teams
Teams: ARI, COL (two parks), FLO, TBA

Group 3: Ballparks that have been replaced during the time in question
Teams: ATL, BAL, CHA, CIN, CLE, DET, HOU, MIL, MON, PHI, PIT, SDN, SEA, SFN, SLN, TEX, TOR

Group 4: Ballparks that have replaced other parks during that time
Teams: ATL, BAL, CHA, CIN, CLE, DET, HOU, MIL, PHI, PIT, SDN, SEA, SFN, SLN, TEX, TOR, WAS (was MON)

Here’s the breakdown:

Grp  Bkstp   Pct    FO+   Share
1     56.6   4.15   98.2   32.3
2     55.0   4.16  103.7    7.5
3     62.8   4.17  100.0   36.3
4     56.8   4.27  101.3   23.9

Reading across, we have the group (Grp), the backstop distance (Bkstp), the foul-out percentage (Pct), the percentage relative to the league (FO+), and the group’s relative share of our sample size (Share). Foul outs in the old-guard stadia, the ones unchanged since 1985, occur at a frequency that’s nearly two percent lower than the league average despite relatively short backstop distances, a counterintuitive finding. The expansion stadia, on the other hand, yield foul outs at a rate 3.7 percent above the league average, this despite the fact that this relatively small share of the sample includes a pair of home run havens in Arizona and Denver, another counterintuitive finding. Meanwhile, the new parks that have replaced older ones have resulted in a slight increase in terms of foul-out frequency, despite the fact that the changes have been as extreme as a 9.8 percent increase in Texas, and a 16.3 percent decrease in Atlanta.

Further grouping these to illustrate the impact of the newer parks:

Grp        Pct    FO+   Share
Old (1+3)  4.16   99.2   68.6
New (2+4)  4.21  101.9   31.4

We close with the exciting conclusion that a field trip to the box factory is pretty lame foul outs in general have decreased slightly in frequency since 1985, but the newer stadia aren’t to blame for this. On the contrary, they’ve produced relatively higher foul-out frequencies. Even if we’ve identified a way in which ballparks have shrunk over the years, the decrease in foul territory–or at least a perceived decrease, since there’s no systematic measurement of foul territories–doesn’t appear to have contributed to the rise in home runs over the past two decades by providing more opportunities for balls to be hit into fair territory. Coupled with the chapter I wrote in Will Carroll‘s The Juice, which showed that the impact of the newer parks on rising home run rates was minimal at best, this data tells us that it’s time to retire the complaint that newer, smaller parks have driven the boom.