Bringing home the BACON
February 19, 2009 1 Comment
Let’s talk about contact.
I’m sure you’re familiar with the concept of BABIP. Now, instead of looking at balls in play, let’s look at contact. (The difference is that we include home runs – excluding HR may make sense for pitchers, but in some contexts, not for hitters.) To that end, let’s introduce two “new” stats, or at least two stats that have probably been invented before that I’m giving names to:
BACON – Batting Average on CONtact. Yes, it’s pronounced “bacon.” This was unintentional, but freaking awesome.
SLGCON – SLuGging on CONtact. Sadly not a breakfast food.
I’ve adopted the terms batting average and slugging, but am not actually using them the way they’re traditionally defined. In this case, I’ve defined a hit as “reaching base safely on anything other than a fielder’s choice,” thus including reaching on an error. For slugging percentage, I’m using the furthest base attained to by the batter on a hit (as defined above). So if a runner hits a single and reaches second on an error, he gets credited two bases. (If he gets thrown out at second, he gets an out, not a hit.)
So what are we going to use these things for? Are we going to try to create the latest, greatest DIPS? Are we trying to project future hitter performance?
Nah. Let’s talk ‘roids.
I’m sure you know the conventional narrative on steroids:
- Baseball was a wholesome, pure sport, handed down to us from Valhalla (or Cooperstown, I guess) – perfect in every way, shape, and form.
- OMGZ ROIDS ROIDS ROIDS.
- Big Damn Sluggers use the home run to treat your pure, untouched childhood memories like the Vikings treated seaside English villages.
Act IV of this little shadow box play is apparently “turn baseball into competative cycling,” which isn’t my concept of a good idea. But I digress. Here’s an intersting question: is it true? Let’s take a look at the change in SLGCON from 1988 to 1999:
SLGCON is very stable before and after 1993, but it undergoes a massive jump at that point in time. This is essentially the great story behind the boost in HR rates during the so-called “steroid era.” Let’s zoom the graph out a bit (instead using HR/PA as our unit of measure):
Check out the two regression lines there. We have essentially very stable HR/PA numbers both before and after 1993, but a massive jump at 1993. This is not an original observation, and seems to hold true no matter what measure of power hitting you care to use.
And now we come round to Dan Rosenheck’s recent NYT article.
Where numbers can be somewhat more useful is in alerting us to the fact that something unusual — steroids or otherwise — was going on around the turn of the last decade.
Assuming that pitchers and hitters used performance-enhancing drugs equally, there is no reason to expect that overall run-scoring levels would change.
But what would go up because of steroid use is the degree to which players’ on-field results are separated from each other. The juicers’ production should, on the whole, exceed the league average, while the clean players should lag.
This is exactly what happened between 1993 and 2004. Using the standard deviation, a common measure of how tightly a set of numbers is bunched together, performances by both hitters and pitchers were more spread out during that time than in any 12-year period since World War II. Although some of the difference was caused by adding new teams, expansion was much more rapid in the 1960s than the 1990s, and standard deviations were still lower back then.
Dan is correct that something unusual was – and is – going on. I don’t think it’s steroids, though. Look at the changes in the graph – the change in run-scoring environments was an event, not a process. It strains credibility to try to explain the data with steroids – you’d have to have all of your juicers start getting help from their cousins at roughly the same time, sometime during the 1993 season, to explain the events on the ground.
What about the standard deviations? Can we explain that as something else as the growing gap between the clean and dirty players? I think so – and I think we have to. First, I don’t think that the evidence sustains the idea that (as a whole) clean players were significantly worse than dirty players. The positive tests for steroids since 2003 have largely been scrubs, not stars.
So what could explain the increase in variance along with the shift in slugging levels? I suspect that the increase in slugging levels is to blame. This has the benefit of being the obvious answer, but I suppose I should try and explain.
For whatever reason, it became a lot easier to hit baseballs hard and far right around the 1993 season. (The most likely culprit is actually the baseballs themselves, so from here on out we’ll simply refer to the “livelier ball.”) This livlier baseball was of greatest benefit to people who could hit the ball in the air to begin with. Let’s look at our on-contact numbers now, along with breakdowns by ground balls, air balls (fly balls, line drives and popups) and fly balls:
The predominant benefit of the livlier ball was to fly balls – an increase in on-contact slugging of .180 points! Ground balls saw a much more modest boost in on-contact slugging, only .018. That’s, well, 10 times the difference.
And that’s what I suspect is driving the increasing standard deviations; there was a league-wide change in offense that seems to favor power hitters, who were already the top talent in the league.
The other interesting thing to note is that the ground ball rate increased, and the fly ball rate decreased, when we moved into the modern era. Why is that? I suspect (but haven’t tested yet) that it’s because as the fly ball became a dramatically more dangerous thing, teams and pitchers came to place greater emphasis on getting ground balls.
But the evidence for a massive, steroid-fueled change in baseball offense simply isn’t there. Rosenheck notes:
None of this means that steroids are necessarily the cause of the separation. But the game’s fans are probably in no mood to write off the association as mere coincidence.
But that’s almost certainly what it is – a coincidence.