It’s that time of year again.
March Madness is upon us, with all its basketball, school spirit, gambling, money on the line, eyes glued to tiny phone screen sports apps, hidden browser tabs at work, legendary triumph, crushing loss, breathtaking excitement, tears pouring down the face of a six-foot-tall college senior, mad grads, heart of a champion, cinderella tale, buzzer-thug fame. (Furious.)
And now, a week into the tournament and heading into the Sweet Sixteen, your group is busted.
I’m not psychic, just data driven. We’ve passed 48 of the tournament’s 63 total matches and the odds of your group still being in the game are incredibly slim. It could even be anywhere from 1 in 280,000,000,000,000. (That’s trillions, to save you from decimal counting).
At least that’s what the odds would be if your bracket were completely random – if you designed it by tossing a coin to decide which team wins each duel. With this strategy, you have a 50 percent chance of picking correctly; multiply that chance by 48, and your odds drop exponentially—to 1 in 1/(0.50)^48, to be precise. The whole tournament? You land this shot once every 1/(0.50)^63 = 9.2 trillion tries.
The number trillion is so irrelevant in life that it can hardly be fathomed. according to a Report from the NCAA website, when you consider that there are an estimated 7.5 trillion grains of sand on Earth, that means that if you had to guess what particular grain of sand from any of the world’s beaches I was thinking of, you’d have a 23% better chance than a perfect mount. Or if you had to guess where a single acorn is hidden in one of the planet’s three trillion trees, your odds would still be three million times better than a perfect bracket.
As Tim Chartier, a math and computer science professor at Davidson College (where NBA star Steph Curry once played), recounts Fast companyyou could do a billion brackets per second and it would still take 300 years to cover the trillion possible versions of events.
But again, the quintillion is largely irrelevant. Nobody actually picks a bracket by tossing a coin, and if you’re even trying to identify the better team, you’re probably over 50% and the odds of the perfect bracket increasing exponentially.
Could it reach the realm of possibility? According to Chartier, who has researched and even developed the “bracketology” for years fan-friendly March Mathness website that produces weighted brackets, the answer is not really. There are modeling engines that process data from millions of factors to determine the most likely outcome of each game, but Chartier says their accuracy is around 70% despite years of tinkering.
If you picked the correct winner of each game 70% of the time, your bracket perfect odds would be 1 in 1/(0.70)^63 = 5.7 billion. Still a hard hit.
This makes it clear why Warren Buffett — who made his fortune investing wisely — has made a sport of offering wealth to everyone which achieves a perfect hold: Once it was $1 billion, another time $1 million a year for life. As a numbers person, he knows it will never happen.
Algorithms bring us closer – but no cigar
From a trillion to a billion is still orders of magnitude larger. Prediction algorithms can achieve this by analyzing a set of statistics. For a Duke vs. Kansas game, consider how many times Duke wins against Kansas, the difference in points, how many games Duke has won all season, how many it has won in a row before this game (a hot period), whether players are injured, the coach’s track record, how often 2 seed (Duke) beat 1 seed (Kansas) and millions of other factors of increasing complexity, down to three-point baskets in the early game and free throw percentages in the late game. And each of these factors is adjusted in order of importance.
The power these models can accomplish has grown massively in recent years, according to Mark Ward, a statistics professor at Purdue and director of the Data science focused data minetold Fast company. They can now sift through information from newspapers, social media, or Wikipedia to uncover insights, better than humans: “It’s able to spot things that you and I might think are just qualitative — no hard numbers, no quantitative information.” – but it can determine if Written Sentences are favorable to a team because of the way they have been trained.”
To get the point across, Ward brings in AlphaGo, an AI that has learned to beat humans at complex play Walk. It does this by analyzing millions of possible outcomes resulting from a single move and using data from past games to decide which are the most likely. AIs can do this for sports – like EA Games’ video game Driving me crazy cast for the NFL’s Super Bowl by simulating countless possible games, each player like a pawn with custom stats. However, as Chartier notes, the Driving me crazy This model isn’t that useful for March Madness since most of its players only have stats from one and at most three tournament appearances.
At the end of the day, says Ward, “they’re just models,” and luck can always throw a spanner in the works.
This is confirmed by Chartier, who cites the inevitability of the unknowable – the X-Factors which he cannot explain. For one thing, emotional stress can wreak havoc: “Higher seeds, when they get into trouble, feel the weight of history’s shadow falling on them because you don’t want to be the #2 seed losing to the #15 seed . Not to mention, the #15 seed sees the rays of hope and success that shine upon them. This can sometimes shake the level of the game. It’s hard to quantify.” And sometimes mid-tournament teams falter and start behaving differently. Then there’s always the Cinderella stories rising out of left field – this year the St Peter’s Peacocks; and last year the Oral Roberts Golden Eagles. No Chartier model anticipated this.
Even if your election wins, if it was a last-second buzzer-beater, “Did you really predict that?” Chartier laughs, “because it could have gone differently. . . . Sometimes people watch money ball and think it’s possible to get 90% accuracy, but there’s always an element of luck that you can’t predict.”
What matters most to him: The rigors of a team’s regular-season schedule (“it’s not just your record, it’s the strength of the teams you beat”); so-called “math mojo” (when you win against good teams during the season); and don’t miss home (when you win against good teams along the way). “If you play all your tough teams at home early in the season, it’s less predictive of March Madness’ performance,” he says.
The NBA, he concedes, is more predictable: players are technically closer – making it harder to keep up – but they are also more consistent. (For example, if a player gets “hot hands” at the college level, the variability is far greater than in the league.) And NBA championships are a seven-game series compared to the NCAA’s single elimination. The longer the playoffs, the more likely the better team will win.
But in the end it’s a percentage game. “Even if you’re just 1-2% more proactive,” he says, your odds of getting the perfect braces increase by billions.
According to the NCAA, which collects millions of entries each year for its official bracket challenge, no bracket last year stayed perfect past Game 28 (a surprise COVID-cancelled game that forced a team to abandon.) The Record was set in 2019, when an Ohio neuropsychologist correctly predicted 49 games in a row. The probability of this was 1 in 38 million to 560 trillion. That run was slammed two games into the Sweet Sixteen when 3-seeded Purdue defeated 2-seeded Tennessee in overtime. (Among tournament stans, the bracket titled “Center Road” was briefly famous.)
We may only have made it to Round 2 this year, but we still have a week and a half of wild twists and turns ahead of us. And broken brackets, dammit, we’re here for the drama. Bring on the madness!
https://www.fastcompany.com/90734778/mathematicians-explain-why-predictive-algorithms-still-wont-get-you-a-perfect-march-madness-bracket?partner=feedburner&utm_source=feedburner&utm_medium=feed&utm_campaign=feedburner+fastcompany&utm_content=feedburner Is a perfect mount possible?