Cult of Collaboration Bingo!

With new general manager Kwesi Adofo-Mensah and head coach Kevin O’Connell, the Minnesota Vikings have made a conscious decision to approach the draft like a gamble. And they want to be the house.

For organizations devoted to process, success or failure isn’t defined by one outcome, but by a series of practices that produce good outcomes more often than not. This is how the casino wins in Las Vegas. A bet on a roulette wheel might produce an outcome that looks dramatic at the time, whether it’s a big win or a big loss. But with enough players and enough spins, Vegas comes out ahead, earning a bit over 5 percent on an average spin.

Throughout the casino in any number of games, these small edges exist for the house’s benefit — anywhere between 10 percent to less than 1 percent. Despite how small these edges can be, the long time frame and high volume allow casinos to rake in massive profits year after year. Those edges on the margins, which exist solely because the green zeros on the wheel hit just over 2 percent of the time, are what make the business viable.

And those small edges are what the Vikings hope to replicate as they approach player acquisition, team-building strategy and on-field tactics.

For Adofo-Mensah, it’s a natural extension of his background in economics and finance. As he explained, “Nobody liked economics when they learned it, but you talk about economics, it’s optimizing utility, right? There’s a very simple equation and constraints that you have, and you solve for those constraints. The NFL puzzle is different because you don’t know what that equation looks like, you don’t know exactly what those constraints look like. So there’s a little bit of art and there’s a little bit of science. That’s what makes the puzzle fun.”

While it might be difficult to construct a Pareto-optimal NFL roster, this desire to produce an optimized roster has driven Adofo-Mensah through his fast-rising NFL career. In this case, it means recognizing the interconnectedness of an NFL puzzle and making bets stacked in the Vikings’ favor. This approach to betting on the margins has been there since he started working in football.

The Vikings GM knows that the draft can be random and that no general manager substantially outperforms any other general manager over the long run. When asked about it, he enthusiastically embraced this oft-studied finding. “The first step is understanding that and not treating it like you have some sort of golden magic wand. In San Francisco, I would always tell them: I can get you the house advantage. I can get you marginally better. But if we’re marginally better at every pick, every decision we make, that adds up to a big deal.

“Yes, the draft is random, but it’s more random in certain spots than others, right? Trying to address certain needs that you need, you probably should do it in the more high-probability places in the draft. And, obviously, there are probably certain positions that are harder. That’s where some of the studying, the data that’s come in, there are certain things that you learn about best methods going forward.”

For Adofo-Mensah, that data doesn’t have to come in the form of quantifiable, spreadsheet-friendly sources. Everything about a draft prospect is data, and so his definition of “analytics” or a “data-based” approach to the draft might look a lot different than the arguments we see pop up on Twitter.

“Data is really important. But I think people sometimes lose the sense of being open-minded about what data is. Data is observations. Data is noticing how I walk into a room. Data is also a number. And we use all those things at our disposal,” he said. “It’s just one information source, and you have your other information sources, and you combine them. At the end of the day, when these things don’t agree, that’s the call to process. Let’s get in a room and figure this out — because great decisions are made when you learn. Decisions are about learning environments. So I think data helps in that process.”

That doesn’t mean all that data holds the same relevancy. He was quick to point out that the complexities of any particular play in football make passing yards a fairly crude way of evaluating quarterback performance, and he was wary of the value of traditional interview settings, pointing to a study performed by Google about the effectiveness of standard interviews versus structured interviews that replicate on-the-job experiences.

But even if data isn’t particularly clean or relevant, he’d still like access to it so he can make judgments about how to weigh it. When asked about the NFL removing the Wonderlic test from the NFL combine, Adofo-Mensah lamented a lost data point. “I’m a data guy, so I don’t want you to remove data from our datasets. That’s just how I go. I want to add; never take away. I don’t know that it’s necessarily correlated precisely to what happens on the field. It’s more pattern recognition and things like that. I do think that some of the new tests that have come out are pretty good, but the Wonderlic is not a bad test. Also, it probably just shows some level of preparation and care, no different from anything else.”

The research largely backs up what Adofo-Mensah says. Aside from some outlier studies, the vast majority of the literature on the subject tells us that Wonderlic scores do a poor job of predicting quarterback and non-quarterback performance. But having as many sources of data as possible is important to Minnesota’s new general manager, and he sees it as his job to parse out the importance of different data points later on in the process.

That’s one reason Adofo-Mensah brought former Colts general manager Ryan Grigson into the building. “He’s a talented evaluator,” Adofo-Mensah said. “I want the most people around me that are great in different skill sets. My job is to combine them all into one decision to help the team. I want to combine as many talented people as I can.”

“Sometimes it won’t just be thinking in terms of a straight average or something like that, it’ll be ‘Hey, this person has been doing this for 40 years. They might be more believable than my own thought process.’ That’ll just come from learning, that’ll be on the job.”

The key insight from his work as a quant is not the importance of numbers but rather a method of thinking about the world that understands the gray instead of seeing things in black and white. “It’s a way of thinking. That’s why I don’t like the term ‘analytics’ — because that implies that other people don’t do it. I promise you Bill Walsh has a way of thinking, I promise you Bill Belichick has a way of thinking. It’s just about being very detailed, using evidence to confirm your hypothesis and what you think and your ideas. That’s all it is.

“To be the best GM,” he continued, “I have to be able to combine all those data sources together. And at its very simplest, a model is just a decision rule. It’s a way to formally and consistently take the same inputs of information and end up with the same answer. I think I have great people in the room, so I want to hear them out and we’re going to make decisions collaboratively and come to great conclusions together.”

That emphasis on evidence and collaboration might mean asking uncomfortable or unusual questions, like trying to go back over a decade of scouting reports to see if nuanced route running matters as much as off-the-line explosiveness when projecting college receivers into the NFL. Or trying to see if hustle plays are overrated in a scouting report and length is underrated for a defensive lineman. Adofo-Mensah’s background also allows him to perform the same analysis on the statistical data provided by the analytics department, too.

“Minnesota has an incredible team,” he said of the analytics group in Minnesota. “Scott Kuhn, Rex (Johnson), Frenchy (Chris French) — they have a really good team of people. They’ve done some great things. So, it’s been cool for me to now be in this seat. Like ‘No, come in my office, tell me everything, don’t hold back. Tell me what’s in the black box, tell me everything,’ and it’s a new dynamic for them.”

He joked that “Everybody thinks they want an analytics-friendly GM until he gets there and he starts asking you for a lot of things,” but it’s clear that the kind of external rigor applied to these models provides a new dynamic for how the analytics team and the front office interact with each other. And though he’s bound by contract not to bring statistical models developed by the 49ers and Browns to the Vikings, he does have the insights gained from the development process that he can apply to the insights generated by Minnesota’s analytics department.

That insight could lead to challenges at every level of operation. It’s one thing to question the traditional pillars of film scouting and backtesting those claims, but it’s another thing to turn it inward and perform the same analysis on the analytics department.

In the end, it’s a way to hone the process and make sure every step in the journey toward finding another player is a well-studied one. And that kind of internal rigor might lead to conclusions that don’t match those of the third-party analytics community. Positional value, a constant debate for those who study football, isn’t imagined in the same way for Adofo-Mensah. He instead looks at it in the framework of how many teammates a player can effectively replace.

“A good player, if they make a good play, how many of their teammates can they save at once? How many of their organization can they save? Can they save me? Can they save the coordinator? Can they save everybody? I try and look at it from a very simplistic sense in that way,” he said. Though he added that it naturally lends itself to quarterbacks and pass rushers, he wasn’t concerned about the particulars of a specific position’s value because “at the end of the day, great players are scarce.”

Statistically-oriented thinking extends beyond double-checking the evidentiary basis for evaluation — it also understands the world, and football, as an interconnected series of systems more than individual parts that can be isolated in a vacuum. It also acknowledges the deep uncertainty that one operates in when working in football. That’s why Adofo-Mensah likes to leverage his stock- and commodities-trading background when considering players and football strategy.

Like with the stock market, “you’re just guessing,” when it comes to player outcomes, he said. “You’re predicting. We have information today and you watch a player and then you see in five years what they are like. Those things don’t always line up. People come through, they improve their mechanics, they get different coaching, they get in different schemes that fit their skill sets.

“You’ve only got so many bullets, right? You’ve only got so many first-round picks, second-round picks and money. So you want to use all your best bullets on the things that are the most certain and come out that way.”

That approach has filtered into the hiring of O’Connell as head coach, whom Adofo-Mensah has called his “thought partner.” O’Connell has embraced a version of statistical thinking and the interconnectedness of the game. His job is to create consistent performance on the field, so his approach is to make sure players are enabled to make the right decisions.

O’Connell has emphasized that players must understand the “why” of how they’re doing things. Even unintentionally, O’Connell seems to be employing another process-oriented insight: This could be described as a form of “anti-fragility,” where a system can withstand unexpected shocks or changes.