A post on chance, on the off chance that it will make Myles mad

Now how likely is that? I gave Myles a hard time about Greg Maddux's apparent tough luck after he suggested that a recent stretch of well-pitched but seldom-won games is a rather common event for Maddux and quite uncommon for every other pitcher. I thought that having such a stretch (in which Maddux threw 20 strikeouts, 2 walks, 1 win, 2 losses, and 3 no-decisions with a 2.57 ERA) should not at all be uncommon in any pitcher as great (or lucky) as Maddux who has been able to endure such a tenure in the major leagues.

Steven Dubner recently wrote about the Texas Rangers scoring 30 runs in one game and how unusual he thought it was that in setting the Major League record for runs scored the Rangers actually scored in only 4 of 9 innings. That means that in the course of one game this team failed to score in more than half of all attempts (5 of 9) yet still scored the most runs ever in a 9 inning game. How odd is that?

Well, maybe not so odd after all. What is more unusual, he suggests, is our ability to predict randomness. On a similar topic Steven Levitt muses that for the Kansas City Royals to tie a record losing streak is actually not that noteworthy. Though a worthwhile read I'll summarize it for you: we expect too much uniformity out of randomness. That's why we expect the Royals, who we don't expect to win very many games, to win at least one out of four. It doesn't matter that they'll end up averaging one out of four for the year--we expect one out of every four. Furthermore we like to think that when they've lost 15 or so consecutive games that this is not a normal behavior but an unusual circumstance.

Likewise we expect that when the Rangers post thirty runs in nine innings they would do such in a "more orderly" manner. We just don't tend to like random events and feel much more comfortable with uniformity. If the Royals should win one out of every four games but haven't won in twelve contests, then they're due to win the next four. Although we expect Greg Maddux to lose well-pitched contests from time to time, we don't expect a string of such contests to happen so close together.

Levitt suggests this: predict the results of 25 coin flips, then flip a coin 25 times. Compare your prediction versus the actual results and see how comfortable you feel about randomness. Maybe then you'll have a little less surprise, though no less empathy, for Greg's tough luck. Or maybe not.