In what will come as no surprise to those of you that follow my columns (and also to Coach K), I’m a big basketball stats nerd. Offensive efficiency, defensive efficiency, offensive rebounding percentage, Pythagorean winning percentage—I can’t get enough.

Using databases from previous years, some of the Internet stats gurus can use the score and time remaining of a given game to estimate each team’s chances of winning. The more advanced models take into account the relative strength of each team to determine that, for instance, Duke had nearly a 100 percent chance to beat Hampton after the Blue Devils took a 16-4 lead five minutes into the game.

My friend takes exception to those statistics. He asks what value the statistics have. He asks whether, if I were Hampton’s coach, I would just take my team off the floor with 35 minutes left to play in a game that was—statistically—nearly unwinnable.

I bring this up because, by one commonly-used statistical model, Arizona had just a 9 percent chance to beat Duke at halftime of the teams’ Sweet Sixteen game on Thursday. Of course, we all know how that ended.

The point, which my friend apparently had been ineloquently trying to make, is that probabilities are essentially meaningless when predicting the outcome of a single event. Perhaps if Duke played Arizona 100 times, and led by six at halftime in each one of them, then the Blue Devils would win 91 of those games. But they only played once, and Arizona won that one.

Modern medical education is focused on evidence. We are taught that, whenever possible, we should seek out a study to help us determine our patients’ diagnosis or prognosis, and that we should support our treatment decisions in the same way. These studies may follow a group of patients—half of whom receive a new drug and half of whom receive a placebo—and report on how many of each group achieved some specified outcome, like survival without having a heart attack or number of months before cancer relapse.

It is easy, as a trainee physician, to be seduced by the power of these numbers. We cite the study showing that fewer than one in 10,000 patients treated with cholesterol-lowering drugs requires hospitalization for muscle breakdown, and manage to convince ourselves that it can’t—won’t—happen to our patients. We cite the studies showing that fewer than seven out of 100,000 people aged 20 to 49 will be diagnosed with lung cancer to convince ourselves that we won’t have to give our 25-year-old patient with a cough a death sentence of a diagnosis.

Of course, it goes the other way, too. Less than 2 percent of patients with metastatic cancer of the pancreas will survive for five years, but we still cling to that hope.

This information is, of course, essential. It enables us, as doctors, to plan for our patients’ future and to choose the best course of action. But our patients only live one life, and as such are only one of our millions of data points. Someone, after all, has to be that one in a million.

The NCAA Tournament plays out this process on a smaller, but more public (and less important) stage. Duke had a 91 percent chance to win—but lost. Using similar statistical measures, Virginia Commonwealth had a 12-in-10,000 chance to reach the Final Four at the start of the Tournament—but reached the Final Four anyway.

As I sat there watching the Blue Devils lose in improbable fashion, I did not think of my patients who get crushed by an improbable diagnosis or a rare medication side effect. But perhaps I should have. It is easy, in basketball or medicine, to be seduced by statistics, to try to put order to a world that so fiercely resists it.

But it’s the space between prediction and reality where March Madness lies. And although it would be nice to have more certainty in medicine, there’s no escaping life’s unpredictability. All we can do is embrace the madness.

Alex Fanaroff is a fourth-year medical student. His column runs every Wednesday.

### Discussion

Share and discuss “On madness” on social media.