Avoiding the Famine

Two weeks ago we took a look at some feast-or-famine offesnses from 2006. The quick-and-dirty metric for identifying such offenses was to look at the difference between the number of games in which fewer than three runs were scored or the number of games with more than six runs and the predicted number of such games based on the Weibull model.

A team that famines is particularly disadvantaged. In 2006, teams scoring a single run won only 9% of the time and teams scoring twice won 21% of the time. (Teams that didn’t score at all predictably won 0% of their games). Scoring the third and fourth runs was critical, with the winning percentages jumping to 31% and 47%. A team scoring 0-2 runs less often than expected based on its overall offensive talent is “wasting” its runs.

What, aside from random variation, would cause an offense to score a low number of runs less often than expected?

Shutouts

As a toy problem, consider the matter of shutouts. Scoring zero runs is awful, as a team can never win a game when scoring zero runs. The important thing to look at is whether a team is shut out more often than would be expected based on its overall offensive talent, which I will call dRS(0) (delta indicating the difference between the actual number of shutouts and the expected number of shutouts). In this case, “expected” means based on the Weibull distribution. The 2006 Cubs were particularly poor in this regard: the Weibull distribution predicted that a team scoring 716 runs would be shut out seven times, but the Cubs managed not to plate a single run in 15 games. That’s a lot of games in which the Cubbies had zero chance of winning. On the other hand, the 2006 Phillies were shut out three times, two instances fewer than would be expected.

Let’s say you wanted to construct a team that wouldn’t be prone to being shut down. How would you construct your roster? Do you want lots of guys on base? Lots of hits or lots of walks? Power?

Before delving into what causes a team to be prone to shutouts, I should point out that the Weibull model typically underpredicts shutouts, so that 26 of the 30 teams are shut out more often than expected for 2006. In fact, the correlation between total offensive output (as measured by total runs scored) and dRS(0) is 0.326, so one should consider that a baseline for the correlations that follow (all data is 2006 only). What this means is that indicators that correlate with dRS(0) with an absolute value less than 0.326 are not as important as overall offensive talent. The goal here is to identify attributes more important than overall offensive talent that can keep a team from being shut out. The rate indicators considered here are the three true outcomes (K%, BB%, and HR%) as well as OBP, ISO, and hit percent. The correlations between these rates and dRS(0) are:

OBP	-.200
ISO	-.489
K%	-.256
BB%	-.184
HR%	-.393
H%	-.001

(For a quick primer on correlation, click here. I’m going to play fast and loose with the statistsics here by not testing for statistical significance. It’s something that should be done and something I do plan to do when I get more data.)

The first thing you should do is stock up on luck; the correlation between dRS(0) and any of the obvious offensive rate stats is less than about 0.5, so much of the variation is probably random chance. But we can immediately see that getting on base, avoiding strikeouts, taking walks, and getting hits are not as important as overall offensive talent. A patient team doesn’t have an advantage over a hacking team when it comes to avoiding shutouts.

Where teams can really help themselves is by stocking up on sluggers; both ISO and HR% are better correlated with avoiding shutouts than overall offensive output. Of course, this makes sense: no matter how bad an offense is, if it has a few players capable of popping one into the gap or over the fence every so often, it’s going to avoid shutouts. Recall that we’re that we’re looking for correlations to dRS(0) instead of total shutouts, so what this means is that of two offenses capable of scoring five runs per game, the one that hits fewer home runs is the one more prone to being shut out.

Getting a few on the board

What about the concept of the “famine offense,” the one that scores 0-2 runs less often than expected? Once again, using 2006 data, we can look at the correlations between this phenomenon, which I’ll call dRS(<3) and the common offensive rate stats.

OBP	-.244
ISO	-.327
K%	-.000
BB%	-.333
HR%	-.388
H%	-.087

(The baseline for comparison is that the correlation between dRS(<3) and total runs scored is -.207) When you're looking to at least put a few runs on the board, getting hits and striking out really don't matter. Getting on base and working walks is a useful skill to have, which makes sense given that you need a few base runners to score a few runs. Once again, though, hitting for power, and especially hitting home runs, is the way to go when it comes to getting a few runs. Longball offenses may be boring to watch—as a fan, I've always perferred watching a gapper score a runer from first to longball trots—but home runs are the surest and quickest way to put runs on the board. How often do you hear an analyst claim that a good offensive team being dominated by opposing pitching is fatally flawed because they are overly reliant on the home runs? The analyst is probably wrong. Hitting home runs is a very good thing, and teams that rely on the home run are not more prone to being shut down. Just the opposite—teams that want to guarantee putting runs on the board would do well to keep a couple of boppers in the lineup. References & Resources
If you are interested in seeing the run distributions (both scored and allowed) for 2006, you can download them (as a zip file) by right-clicking here and selecting “save as.”


Comments are closed.