How to Win Your NFL Pool, Thwart a Terrorist, Maximize Delivery Routes — and Find the Perfect Kidney Match

All you need is the everything algorithm, says professor David Bergman

By Jeff Wagenheim
Illustrations by Rob Dobi

How to Win Your NFL Pool, Thwart a Terrorist, Maximize Delivery Routes — and Find the Perfect Kidney Match

All you need is the everything algorithm, says professor David Bergman

By Jeff Wagenheim
Illustrations by Rob Dobi

David Bergman grew up in White Plains, New York. His mother had immigrated from Morocco, his father from Israel. “They had no idea about American football,” says Bergman, who nonetheless became a huge football fan. It wasn’t the Giants or Jets who piqued his passion, though. Bergman’s chosen team was, and is, the Minnesota Vikings. Now what are the odds of that?

Well, if anyone can tell you the odds of that, it would be Bergman himself. An assistant professor in the Operations and Information Management Department of UConn’s School of Business, Bergman teaches and researches in the field of optimization, which falls under the umbrella of analytics. And he’s found a way to work American football into the mix.

Bergman’s specific area of academic study is computational optimization, which involves developing computer algorithms for automated decision- making — a field called prescriptive analytics. It differs from the other branches of analytics in that it doesn’t just interpret data, as in descriptive analytics, or predict future events based on that data interpretation, as in predictive analytics. Prescriptive analytics uses the ongoing collection of data to automate decision-making in an ever-changing landscape.

“The problems are super complex and they can have a really large impact,” says Bergman. But basically, “it’s a way to drive efficiency or effectiveness in your organization.”

Hence, Bergman is called upon to tackle the streamlining of quandaries in fields ranging from cybersecurity risk assessment and military equipment procurement to parcel delivery timelines and airline crew scheduling. His expertise also happens to work well in football gambling, particularly for NFL fantasy survivor pools.

“Problems that took three months to solve 10 years ago now take a second or two on my laptop.”

If you’re unfamiliar with that type of football pool, it’s really simple. And not simple at all. You enter a survival pool at the start of the NFL season, and your first task is to choose one of the winning teams from among the 16 games in Week 1 (in order to survive to do the same in Week 2, and so on). So let’s say the New England Patriots, who have played in three of the last four Super Bowls, are scheduled to open the season against the Cleveland Browns, who were 1-31 the last two seasons. You couldn’t find an easier pick than that, right?

“Yes, you would be very confident in picking the Patriots to win that game,” says Bergman. “But they might not be your best Week 1 pick.”

That’s because there’s a twist in the rules of a survivor pool: Once you make a team your weekly pick, you cannot choose that team again for the remainder of the season. If you survive to the 17th and final week of games, then, you will have used up 16 teams — one a week — and will have only 16 from which to choose a winner. It could be slim pickings if you don’t plan ahead. So you might want to save those high-flying Pats for later in the schedule, when your options are limited.

Don’t Just Take Things One at a Time

This type of quandary is called a sequential stochastic assignment problem, says Bergman. “It’s a mouthful.”

To take a bite out of this type of problem, you must make a sequence of choices in which each one places limits on the choices available to you in future decision-making. Bergman worked with University of North Alabama economics professor Jason Imbrogno to create an algorithm to automate these decisions, which can address matters far more critical than football.

The same algorithm can be applied, for instance, to such high-stakes concerns as the allocation of donor kidneys to patients. One patient in need might be a perfect match for the first kidney that becomes available, but that patient also might have a compatibility broad enough to match later kidneys. So the greater good might be served by matching the first kidney with a patient with less universal compatibility.
“How do you match a donor with a patient in a way that not only maximizes the probability of a good match in that case,” asks Bergman, “but also gives your next patient-donor decision the best chance of success?”

Bergman believes his algorithm also could enhance airport security screening. Airport security staff has a finite number of agents available at each checkpoint and a constant flow of passengers to be screened or allowed to pass. Every time an agent is with someone, that’s one fewer agent available to deal with the approaching horde. So the ongoing decisions on who to screen must be spot-on. “You want to maximize the agents’ availability without compromising security,” he says. “There are sequential decisions that can be automated.”

The same is true of military or airline crew scheduling, both of which are complex tasks that cry out for efficiency and automation. As does parcel delivery — 10,000 packages to handle in a day, picked up within specific time windows, delivered within specific time windows, going here, there, and everywhere via various modes of transportation. “The decision-making that the parcel services face every day is so large-scale and so complicated,” says Bergman, “that they employ fleets of people with my background.”

The possibilities for applying his calculations are “literally endless,” says Bergman. “There are a lot of ways we can make an impact in optimizing decisions using our algorithm. With automated decision-making using predictive models, you can maximize almost anything, as long as you know what you want to maximize.”

Illustration of a man looking at a football goal post

The Secret to Winning Your Pool

As for the NFL survivor pool, the appeal is to conquer an increasingly popular past­ime — one that is potentially enriching as well, with millions of dollars being bet each season. And the entirety of the problem is set out in front of you all at once, with the available data just a jumping-off point for finding optimal solutions. The key to success is establishing the framework for your decision-making.

If you take the most myopic approach to the pool, considering only which team has the best chance of success this week, you have a high probability of making it through several weeks. But then things will get difficult. “You run out of good options,” says Bergman, “and your probability of survival tanks.”

You also can end up being left with bad choices later in the NFL schedule if you take the opposite approach: plan out your picks for the entire 17-week season in your initial calculations. Teams rise and fall, stumble and regroup as the weeks go by. Those defending champions might look unbeatable as the schedule is getting under way, but by Week 10 — which is when your preseason analysis told you to pick them as your weekly winner — they’re in last place, their season in wreckage. “Probabilities change as the season progresses, sometimes substantially,” says Bergman. “So if you plan out your whole season, you might end up saving a team that later on makes no sense.”

“You can maximize almost anything, as long as you know what you want to maximize.”

Bergman and Imbrogno suggest that you adopt a half-season-remaining plan. Base your Week 1 pick on probabilities for the first eight weeks. For Week 2, refine the probabilities through Week 9. And so on. “A rolling horizon is the way to go,” says Bergman. “This strategy has proved better than millions of other strategies people employ. You’re planning ahead, but with the flexibility to recalculate.”

Those calculations — sorting through 15 or 16 games a week for eight weeks — might sound complicated, but they can happen in milliseconds these days. Computers gain speed every year, and algorithms have never been more efficient. In the last 20 years, Bergman estimates, optimization technology has sped up by a million times. “That is not an exaggeration,” he says. “Problems that took three months to solve 10 years ago now take a second or two on my laptop.”

But while Bergman can calculate for you the optimal picks for the next eight weeks of the NFL season, he would be selling your chances short if he did so and left it at that. The optimal strategy calls for one more step. “Even football experts are able to predict the outcomes of NFL games with only around 65 to 70 percent accuracy,” he points out, “so the probability of you lasting the whole season is minuscule.” The way to go, then, is to buy in with multiple entries — as many as the organizers will allow — and play each entry against the others. “This is the secret sauce,” he says. “Your probability of surviving in the pool increases significantly.”

Illustration of a truck with paths

Timing Is Everything

Bergman considers himself lucky to be living in this time in history when advances in technology have put analytics at center stage in the business world and elsewhere. “I got my Ph.D. in this field at an opportune time,” he says. “I didn’t pick this career path with any expectation that this was going to be a super exciting time to do it. But timing is everything.”

Of course, Bergman recognizes that timing has its own complexities. He and his wife recently became parents, and they’ve already been inundated by information and warnings about children’s screen time and access to technology. He is equipped more than most dads to recognize the benefits, and he’s in the process of gaining firsthand experience with the challenges. “As a society, we are going to have to work on being more conscious,” he says. “But that’s not my research area. I just work on making things more efficient.”

Including his own survival pools. “Football is such a complex game, with so many factors having an impact on every play. There’s so much going on that it’s a very hard game to predict. But I love trying.”

And if things get too topsy-turvy for any algorithm, Bergman can always return to the childhood technique that made him choose the Vikings and started his NFL fandom: “I think it was because I liked the team colors.”


No comments so far.

Leave a Reply