Few things are more controversial in sports than “momentum”.
Most observers insist that it is real, an intangible concept but nevertheless observed and felt as sporting events unfold. The basketball player who shoots hard continues to shoot. The player who is 4 for 4 is more likely to get that big hit on his fifth plate appearance. The team “on his heels” is in danger of collapsing. Advertisers constantly draw attention to capturing momentum.
“Psychologists will tell you there’s an evolutionary benefit to recognizing a pattern even when it doesn’t exist,” said Paul Roebber, who teaches data science courses and specializes in weather patterns at the University of Wisconsin-Milwaukee.
“Sometimes they’re false patterns, and the classic example they’re talking about is if you’re on the edge of the savannah and you see the grass waving, there’s a 99% chance it’s the wind, but If it’s a lion stalking you, it’s best to assume it’s a lion stalking you because it’ll be your end if it does. We’re programmed to recognize patterns even if they don’t. do not exist.
Can momentum be identified as tangible in football?
But here’s the thing: Roebber did some programming to prove otherwise.
Roebber, graduate student Bryan M. Burlingame, and Roebber’s football-loving friend Anthony deWinter set out to discover whether the intangible could be identified as tangible. Their findings were published in PLOS ONE Research Publication and touted by UWM with a fun statement: “The fans are right.”
“There have been studies that have tried to test whether ‘the hot hand’ exists or disprove how it exists, and I wasn’t happy with the results because I felt like they didn’t define it. really correctly,” Roebber said. “I was thinking about the performance of the team and whether the team was gaining momentum, not the individual player.”
Roebber, a Boston native who has been in Milwaukee since 1994, started working on the project when football fan Burlingame suggested they explore it as a side project to something else.
“He was interested in football, but unfortunately he’s a Lions fan,” joked Roebber (maybe it was funnier before the Packers lost to the Lions this year …sometimes it’s not just the wind that moves the savannah grass, after all).
What Project Momentum required
The project took two initial steps: developing a “win probability” calculator that recognized when a team’s chances of winning a game were improving, and defining momentum.
Once they figured out the first, they created a neural network and fed it NFL play-by-play data from 2002-12. They defined momentum as seeing the probability of winning increase over three successive series (offensive or defensive). If the probability of winning had steadily improved, a team had achieved positive momentum; if the probability steadily decreased, it was a negative dynamic.
They used seven different inputs to arrive at their probability of winning. The most important things to consider, Roebber said, were the in-game score at any time, the point spread (as in, the expectation that one team is better than the other, thus giving insight into a team’s probability of winning) and where the team was on the field.
Roebber said the advantage of “momentum” moments on a team’s chances of winning was not only statistically significant, but overwhelming.
“Teams that end up winning have about 14% more of those momentum developments or streaks in those games than teams that don’t win, which certainly seems to have some relevance to whether you win the game.” , said Roebber.
It might seem obvious – a team with three good streaks in a row has a better chance of winning. But remember, the probability of winning is relative to a team’s position in the game – teams that dominate a game will not necessarily reach Roebber’s definition of “momentum” as the bar to improve their probability of winning is already high.
Roebber said his software can predict whether a team is going to win based on elements of the game. Projections are about 80% accurate after the first quarter, then increase to 90% in the fourth.
“It’s not perfect, of course,” Roebber said. “The famous Super Bowl game between the Atlanta Falcons and the New England Patriots, at the end of the third quarter, Atlanta had a 99% probability of winning…I call these ‘black swan events’ “; you can’t predict this stuff.”
Momentum is ‘the function of all actors’, analysis says
Can it go further and predict when dynamic events are likely to occur? He tried that too.
“We built another neural network where we used the probability of win as an input…we found it does a pretty good job of knowing if a sequence is going to happen before it happens,” did he declare. “There’s more uncertainty (than with the other predictive model), and it seems like the most important thing for that to happen is if you’re lagging only slightly, have a lot of time at the counter and less than total score (in-game).”
He plans to look at other sports, but the NFL seems more conducive to an overall assessment and not a player-specific assessment.
“When you’re talking about a team sport like football, I think momentum is a function of all players,” he told the UWM report. “And so, you really have to look at the collective performance of the team.”