One of the basic viewpoints with roots in the statistical analysis movement that is attributed to Billy Beane in Moneyball is the idea that the playoffs represent too small of a sample size to mean anything real. That is, playoff performance is decided mostly by luck. All one can do is build a team capable of making the playoffs with two or three quality front end starters to run out in the 5 or 7 game series.
It is a common notion, on the other hand, that teams willing to play 'small-ball,' that is teams that bunt, steal, play hit and run and so forth are better playoff contenders. Now, we all know that statistical analysts have shown how these kinds of plays are not really worth it. They are low percentage plays that, in general, cost more than they produce. As a result of this, views like that above of 'small ball' teams are often met with disdain from the statistical community.
However, I am not so certain that these two views are not reconcilable, if not even complementary.
What it comes down to is this: If playoff results are determined by luck, why is it advantageous to play towards the high percentages? These plays certainly work out better in the long run, but it is not clear that they lend any real advantage over the course of the 10-17 games one will play on the way to a title. All of the teams in the playoffs are good baseball teams. So, by the small sample size argument, either your team is hot, and they win, or it isn't and they don't. Longer term trends or probabilities are irrelevant.
It follows, then, that if you are depending on luck to get you to the WS, why not push your luck? If you steal and hit and run, the percentages are against you, but as I said, that is irrelevant. The important fact is that if you steal and hit and run your potential payoff is bigger. So, if you are lucky and everything works out, you come out in the best position possible.
Look at Dave Roberts's stolen base in game 4 (you know which game 4). If he had been caught, they most certainly would have lost that game. But if he had not stolen that base, they also most likely would have lost anyway. Just one example, to be sure, but it serves to illustrate my point. The extra risk taken brought extra reward, and the extra reward in this case was the difference maker between a win and a loss; between a playoff exit and an historic World Series win.
So it seems that there is good reason to think that much of the sabremetric wisdom can be thrown to the wind in a short series. After all, statistical analysts do not even purport to be saying anything about very small samples. The traditional risk/reward payoff is thrown off balance, and it may very well make sense to err farther to the side of greater risk than would be wise over the course of a full season.
If this is the case, the traditional view of 'smallball' teams as good playoff teams does not contradict results given by sabremetric analysis. In fact, I even think that the view should be seen as complementary.
Subscribe to:
Post Comments (Atom)
1 comment:
I agree with the general idea that certain plays that aren't useful in the aggregate (i.e. stealing bases if you aren't stealing them at a high percentage or sacrifice bunting) are definitely worthwhile in a short series where luck plays a huge component. However, I want to present an important nuance that adds to Theo's point.
"So it seems that there is good reason to think that much of the sabremetric wisdom can be thrown to the wind in a short series… it may very well make sense to err farther to the side of greater risk than would be wise over the course of a full season."
This is a really important point since it illustrates, at least to me, the importance of using sabremetrics but not relying on it as the only or even the main factor in decisionmaking. Whereas Theo says that we might as well throw out a lot of the sabremetric data, I think that this is precisely when it comes in handy. Let's take sacrifice bunting as an example.
Let's say your team is in the playoffs and one of your hitters leads off an inning with a double. Many traditionalists would say that you should bunt him over. On the other hand, most people who know about sabremetrics think that, in the aggregate, sacrifice bunting is a great way to kill a rally. what do you do next? This is where using sabremetrics situationally becomes important. If your team is down a run or tied in the 7th and you know the other team has a really good, well rested bullpen you might want to bunt. Here is why, even though the expected runs for a man on second with no outs is significantly higher than that for a man on third with one out, the chance that you will score at least one run is slightly higher if you bunt the man over. Thus, when you apply the knowledge of gleaned from using sabremetrics to the situation and combine that with Theo's astute point about risk/reward, you have an even stronger reason to decide to bunt the man over.
I guess this is less of disagreement with Theo's post and more of an addition. To sum up: sabremetrics can be useful is almost any situation but should always take a backseat to the specifics of that situation when determining a course of action; especially in a short series where each out counts even more since you can't just get them back next week.
Post a Comment