Scorecasting (35 page)

Here’s the interesting part. The players themselves—both in the NBA and at Cornell—firmly
believed
in the hot hand effect. They
felt
hot or cold, as though the result of the previous shot would go a long way toward determining the result of the next one. Considering how often coaches instruct their players to “feed the hot hand,” it’s clear that many of them believe in the phenomenon of momentum. And as fans, most of us do, too. Before the professors began looking at basketball players’ stats, they surveyed fans and found that 91 percent agreed that a player has “a better chance of making a shot after having just
made
his last two or three shots than he does after having just
missed
his last two or three shots.” In fact, the fans estimated that his chances were 20 percent greater if he had just made his last two or three shots. Even for free throws, 68 percent of the surveyed fans agreed that a player has “a better chance of making his second shot after
making
his first shot than
after
missing
his first shot.” A full 84 percent of fans believed that “it is important to pass the ball to someone who has just made several shots in a row.”

Because the researchers’ data was so thoroughly at odds with perceptions in the sports world, they and other researchers refined the study further. Maybe shots taken within one minute of each other would exhibit more persistence. But that didn’t turn out to be the case. They tried replicating their findings using more data on more players over more seasons. Still, the results remained unchanged. Others looked at the results of the Three-Point Shootout held during the NBA’s All-Star Weekend, in which the most accurate shooters on the planet—absent defense, in a controlled environment—compete in a straightforward contest. Again, as often as announcers declared, “He’s on fire!” there was no evidence of momentum.

There was one academic study that, initially anyway,
did
find evidence of a hot hand. Irony of ironies, the results were driven largely by … the shooting of Vinnie Johnson. The key piece of supporting evidence, though, was the Microwave’s run of seven consecutive baskets in the fifth game of the 1988 NBA finals. Unfortunately, that seven-out-of-seven streak never happened. The data had been miscoded. (He missed his fourth shot, though a teammate tipped in the miss.) Once the data were corrected, Johnson—again, the player most notorious for shooting in spurts—was shown to be no more or less streaky than any other player, no more or less likely to make a shot after a hit as he was after a miss. But the researchers did show that Johnson and his teammates
thought
he was a streak shooter. He tended to shoot more after making a basket and was fed the ball more frequently after each make. The problem was, he wasn’t more likely to score.

More recently, John Huizinga and Sandy Weil, who also investigated the value of blocked shots (see “The Value of a Blocked Shot”), updated the hot hand study by looking at all NBA games between 2002 and 2008. They, too, found no evidence of any hot hand effect. However, they did find something else. Despite there being no greater likelihood of accuracy, shooters making their last
several attempts act as if a hot hand exists. After making a shot, they take harder shots—and shoot about 3.5 percentage points below their normal field goal percentage. They also shoot 16 percent sooner than they do after a missed jump shot and are almost 10 percent more likely to take their team’s next shot if they made their last shot than if they missed it. (Both of these effects are much stronger for point guards and swingmen, which stands to reason: No one talks about a “hot hand” in conjunction with dunks, layups, and other short-range shots.) The authors concluded that if everyone on the team behaved this way—shooting more frequently and taking more difficult shots after a previous make than after a previous miss—it could ultimately cost a team 4.5 wins per season on average.

Okay, so momentum doesn’t exist on the level of the individual, but what about at the team level? Does momentum exist for NBA teams? We considered about 3,500 NBA games between 2005 and 2009, examining the play-by-play data and paying special attention to scoring runs. One can define scoring runs in any number of ways; we chose to look at teams that scored at least six unanswered points in the previous minute and called them hot (or their opponents cold). We then asked what happens over the next minute in those games. Did the hot team continue to remain hot by increasing its lead (or decreasing its deficit) against the cold team?

In a word, no. In fact, we found the opposite. If a team scores six or more unanswered points in the previous minute, it will on average be outscored by its opponent (by 0.31 points) over the next minute. This implies that there isn’t merely an absence of momentum; there is a
reversal
. A team that gets hot is more likely to do worse, not better.

But perhaps a minute was too short a time frame to consider momentum, so we looked over the next two, five, and ten minutes. But we found the same effect: reversals of fortune, not evidence of momentum. The following chart sums up our results.

In every instance—no matter how far back we defined hot or cold teams and how far forward we looked—we found strong evidence of reversals rather than momentum. Hot teams tend to
get outscored after going on a run; cold teams tend to catch up. When he hosted ESPN’s
SportsCenter
,
Steve Levy had a catch-phrase: “It’s the NBA; everyone makes a run.” Turns out, he was absolutely right.

This, of course, could be for a variety of reasons. Maybe streaking teams expend more energy when they make a run and then get fatigued. Perhaps after a big run coaches are more likely to send in inferior bench players. Or perhaps players exert less effort after they’ve built a comfortable lead, or opponents exert more effort when behind. Whatever the reason for the reversals, the evidence supporting momentum is simply not there.

Next, we looked at comebacks. A team is down by, say, ten points in the waning minutes of a game and stages a furious comeback to tie the game and send it into overtime. Does that team have a better chance of winning in OT? The answer, we’ve found, is no. Its chance of winning in overtime is no different from that
of the team that gave up the lead (or, for that matter, than it is for two teams that were neck and neck the entire game before heading to OT). We found no evidence that teams that are on winning streaks of two, three, four, five, six, seven, eight, nine, or even ten games have any better chance of winning the next game. (The same is true for teams on long losing streaks.)

Even the postseason seems immune to momentum. Often we hear how important it is that teams get hot or sustain momentum going into the playoffs. We find no evidence for that. Controlling for a team’s overall regular season record, we find that a team entering the playoffs on a winning streak—even as much as ten games—does no better than a team entering the postseason on a losing streak. (Sure enough, as we were studying the data, the Boston Celtics were marching to the 2010 NBA finals, having lost seven of their last ten regular season games. Their opponents, the Lakers, had lost six of their nine final regular season games.)

Nor is the absence of momentum unique to the NBA. In
baseball, hitting streaks seem to be no more predictive of future success than slumps are. Batting averages of players are just as likely to be higher after cold streaks as they are after hot streaks. The same thing goes for pitchers. We found no evidence that starters get into “a groove” that enables us to predict future success. Researchers
have
found the existence of momentum in two niche sports,
bowling and
billiards, but those sports (games, really) rely on a repetitive motion and take place in the same physical space. That’s a lot different from sinking jump shots in the face of a defense or hitting 95-mph fastballs.

Why do we attribute so much importance to “sports momentum” when it’s mostly fiction?
Psychology offers an explanation. People tend to ascribe patterns to events. We don’t like mystery. We want to be able to explain what we’re seeing. Randomness and luck resist explanation. We’re uneasy concluding that “stuff happens” even when it might be the best explanation.

What’s more, many of us don’t have a firm grasp of the laws of
chance. A classic example: On the first day of class, a math professor asks his students to go home, flip a coin 200 times, and record the sequence of heads and tails. He then warns, “Don’t fake the data, because I’ll know.” Invariably some students choose to fake flipping the coin and make up the results. The professor then amazes the class by identifying the fakers. How? Because those faking the data will record lots of alternations between heads and tails and include no long streaks of one or the other in the erroneous belief that this looks “more random.” Their sequence will resemble this: HTHTHHTHTTHTHT.

But in a truly random sequence of 200 coin tosses, a run of six or seven straight heads or tails is extremely likely: HTTTTTHHTTTHHHHHH.

Counterintuitive? Most of us think the probability of getting six heads or tails in a row is really remote. That’s true if we flip the coin only 6 times, but it’s not true if we flip it 200 times. The chances of flipping 10 heads in a row when you flip the coin only 10 times are very low, about 1 in 1,024. Flip the coin 710 times and the chances of seeing at least one run of 10 straight heads is 50 percent, or one in two. Flipping the coin 5,000 times? We’d see at least one string of 10 in a row 99.3 percent of the time. At 10,000 times, it’s virtually a lock (99.99 percent) that we’ll see at least one run of 10 heads in a row.

In 1953, the psychologist
Alphonse Chapanis of Johns Hopkins documented how badly human subjects understand randomness by asking them to write down long sequences of random numbers (0 through 9). He found that almost no one chose to use repetitive pairs such as 222 and 333. Subjects instead tended to alternate digits and avoid repetition. In short, they couldn’t create random sequences. This bias can be gamed or taken advantage of. State
lotteries, for instance, have an overwhelming number of tickets with alternating numbers and very few with repetitive digits. Since most lottery pots are split among the winners, your chances of having the pot all to yourself are greater if you pick 22222 versus 65821. Nobody picks 22222. But why not? The lottery is random. Drawing 22222 is just as likely as drawing 65821. You have to
get all five digits correct and in the same order in either case. But people don’t see it that way. (Imagine if the Powerball numbers actually came out 22222. Most people’s first thought would be that something was suspicious.)

The same thing is true with flipping a coin. If you get ten heads in a row, what’s the likelihood that the next flip will be heads? Don’t be fooled—it’s 50 percent, the same as it is on any single coin flip. Most people think the chances of getting heads will actually be
lower
than 50 percent—the opposite of momentum. They know they should see roughly the same number of heads as tails (50–50), so they feel that if they’ve seen ten heads in a row, they’re
due
for a tails. A tails has to emerge to balance things out. But it doesn’t. There is no law of averages. If the process is random, there’s no predictability. This is also what drives the “gambler’s fallacy.” Gamblers on losing streaks erroneously believe they’re due and keep
gambling, thinking that their luck has to balance out. But if the whole thing is random, you aren’t due for anything. Your chances haven’t changed at all.

The casinos, of course, are happy to exploit this failure to understand randomness. Some of them even post the recent results of the roulette wheel spins, hoping to dupe gamblers: “Hey, it’s landed on odd five straight times. We’re due for even!”