TDA Bullpen - Our Writers' Blog

Sunday, November 16, 2003

Do Strikeouts Matter?

Jacob Luft has an interesting article over at SI.com about the evil of strikeouts. He believes that striking out is much worse than other kinds of outs, and this has an effect on winning. Some of the text:

If you take a look at it from a team-wide perspective, then it becomes clear that the strikeout is still worthy of the shame it once held. The current playoff format began in 1994, but the strike wiped out the postseason so we have to start with 1995. Since that season, no pennant-winning team has ranked higher than fifth in its league in strikeouts. But that is actually an aberration, since most pennant winners rank much lower than that.

He concludes with:

With the massive amounts of activity set to happen in the Hot Stove League, many general managers will have a chance to change the basic composition of their lineups. Recent history shows it would behoove them to try to minimize strikeouts wherever they can.

Luft continues with a table that shows the last nine years worth of pennant winners, and their position in strikeouts when it comes to their league, and concludes that strikeouts do matter.

What is the thinking behind this? People always complain that a K is a "non-productive" out, that unlike a grounder to the right-side of the infield, or a long fly ball, runners don't advance on a strikeout, and hence strikeouts are worse than other outs. Well, one third of all outs are the last out of the inning, so right away in a third of the possible cases, K's aren't any worse than other outs. When nobody is on base, why are strikeouts worse than other outs? With a slow runner on first, wouldn't a K be better than a fast-roller to the shortstop?

Suppose that strikeouts are as bad as everyone says they are. Then in theory they should have an effect on run production, no? A high strikeout team should score fewer runs than a low strikeout team if K's are so bad. Well, let's look at the relationship between runs scored and strikeouts in the American League from 1997-2001 (I picked the AL since it gets rid of pitchers batting, the timespan because I already had that data set):





The data bounce all around, showing no clear trend. The correlation coefficient, r, is equal to 0.016, about as close as you can get to being completely uncorrelated. To put this into perspective, square r. A back of the envelope calculation shows that the variation in team runs scored is effected at a 0.03% level by strikeouts. Not 3%, 0.03% In other words, no effect at all.

On the other hand, let's look at On Base Percentage and Slugging:









On Base Percentage looks like what one would expect from a variable that was closely related to run scoring. Teams with low OBPs don't score a lot of runs, teams with high OBPs score them by the basketful. They fall on a nice straight line, with a very high positive correlation coefficient of 0.925. This is a very strong correlation, at roughly (r squared) the 86% level. At r=0.864, Slugging Percentage is also strongly correlated to runs scored, around the 75% level. To put those numbers into perspective, I typed in "smoking lung cancer correlation coefficient" into Google. This page was the first item to pop up. It contains the results of a massive study by the British government in the 1970s on the effects of smoking. Their correlation coefficient turned out to be 0.716. OBP and Slugging Percentage are solid predictors of runs at the team level, unlike strikeouts.

But what about Luft's observations about the relationship between pennant winners and strikeouts? Well, there is a good reason to expect good teams to have fewer strikeouts than bad teams, but it doesn't have to do with the relative ability of batters. Take an imaginary lineup and hook it up to a rotation of Schilling, Maddux, Schmidt, Prior and Brown. If the lineup is even half-decent, this team should win two-thirds of their games. Now take the exact same lineup and hook it up to a combination of the Tigers and Reds rotation. The bad pitchers, not the good ones. This team will win about a third of their games. So what? Think about this in terms of closers. A bad team will face the other team's closer in about two-thirds of their games, a good team in only about one-third. Closers are strikeout artists. With the exact same lineup the team with the poor rotation will have more strikeouts simply because they will be facing more innings against strikeout pitchers (and more innings period - about three games worth compared to the good team due to the extra bottom of the ninths). How big an effect is this? Not large, maybe a few percent. But when the spread from the average team in the league is only about 10% in either direction, it makes a difference.

So GMs may indeed want to take another look at strikeouts.

posted by David 10:26 AM