Join the huddle. Sign up here for Ravens updates in your inbox.

Two years ago, the Ravens entered the NFL draft needing a starting cornerback and some luck.

Clemson’s Nate Wiggins, one of the class’s fastest players, was the team’s highest-rated cover corner. He was also a consensus first-round prospect. The Ravens, after a run to the AFC championship game, had the No. 30 overall pick. Few draft analysts predicted Wiggins would last that long.

But over the draft’s first three hours, a run on offensive linemen — eight were taken in the first 26 picks — pushed corners down the board. Quinyon Mitchell was the first taken, at No. 22 overall. Terrion Arnold went two picks later. When the Dallas Cowboys took offensive tackle Tyler Guyton at No. 29 overall, an improbability had become a certainty.

“I didn’t think you were going to be there for us,” defensive coordinator Zach Orr said in a phone call with Wiggins after he’d been told he was the Ravens’ pick. Even general manager Eric DeCosta acknowledged at a news conference not long after that he thought a fall to No. 30 would “never happen.”

His numbers had said otherwise. According to the Ravens’ in-house draft model, as DeCosta later recalled, there was about a 40% chance Wiggins would be available when the Ravens were on the clock. Not a sure thing, but not a long shot, either.

“And I was just thinking about how that is such a perfect representation of so many things,” said Seth Walder, an NFL analyst for ESPN who specializes in quantitative analysis. He chuckled. Forty percent, Walder said, is “not zero. It’s not likely, but it’s really possible. … We often think probabilities are certainties and unlikelihoods are never going to happen.”

On Thursday night, DeCosta and the Ravens, along with the rest of the league, will enter the draft with perhaps unprecedented confidence in what could happen in the first round and beyond. The rise of predictive models across NFL front offices, spurred on by investments in analytics and an increasingly transparent predraft process, has made the draft easier than ever to forecast.

In Baltimore, the Ravens’ draft model seems to loom almost as large as their draft board itself. After the Ravens took Wiggins in 2024, team officials entered Day 2 hoping to draft an offensive tackle. Derrick Yam, the Ravens’ director of data and decision science, told DeCosta there was an 80% chance Roger Rosengarten would be available late in the second round.

DeCosta considered that an optimistic projection. Yam, DeCosta recalled, nervously conceded that the estimate could be off. But when the Ravens were on the clock again at No. 62 overall, Rosengarten was still available.

This year, with the Ravens set to pick in the top half of the first and second rounds for only the second time in nearly a decade, the pressure to hit on an impact starter is greater — but the obstacles in their way perhaps fewer. Even if trades throw the first round’s draft order into chaos, the Ravens’ odds of landing, say, Miami edge rusher Rueben Bain Jr. or Penn State guard Vega Ioane at No. 14 overall might not wane dramatically.

“In the draft, the first couple of rounds come off pretty easy to predict, I think,” DeCosta said Wednesday. “I don’t know exactly why that is. There seems to be more of a consensus among teams.”

It’s unclear how many NFL teams outside Baltimore use predictive draft models, though officials for several have alluded to their existence in recent years. When Ran Carthon was hired as the Tennessee Titans’ general manager in January 2023, he oversaw his first draft three months later without any algorithmic help.

By the 2024 draft, however, that had changed. Over Carthon’s six years in the San Francisco 49ers’ front office, working alongside director of football research and development Kwesi Adofo-Mensah — later named the Minnesota Vikings’ general manager — and football research and development manager Demitrius Washington — later an assistant GM in Minnesota — he’d seen how helpful predictive analytics could be.

“I was like, ‘Man, this is something I definitely want to have,” Carthon, now an analyst for SiriusXM NFL Radio and a co-host of CBS Sports’ “With the First Pick” podcast, recalled thinking. So he and Sarah Bailey, the Titans’ director of football research and development, built their own. Their model weighed inputs from three primary sources: scouts’ grades, coaches’ grades and media mock drafts. (Projections from especially accurate reporters and analysts were given more weight.)

Building a model is part art, part science. Paul Sabin helped develop models for SumerSports, an analytics company that has worked with NFL and college teams. He said he tried to capture player contributions quantitatively, which could then be tied to value propositions, which might then be reflected in how teams approached the draft. Most front offices could be expected to have models that were largely accurate in the first round and decently accurate in the second round, but offer only “a shot in the dark” later in the draft, he said.

What separates good models from great models, experts said, is not only the data they’re fed but also how the data itself is calibrated. Sabin recalled a conversation he had with an assistant GM for a team before the 2023 draft. Sabin told the assistant GM he didn’t expect running back Jahmyr Gibbs to be drafted until near the end of the first round.

“Oh, no, he’s gonna go really early,” Sabin was told. “There’s enough teams that like this guy.” The assistant GM predicted that Gibbs wouldn’t make it out of the top 15; the Detroit Lions ultimately drafted him No. 12 overall.

As much as it might help a model to know which players have been most often linked to teams in mock drafts, or what traits a front office has typically targeted over its history, draft intelligence cannot be discounted. The risk for a front office is entering the draft with overconfidence in a range of projected outcomes.

“Let’s just say you knew that the Titans were going to surprise everyone and draft, like, Makai Lemon” with the No. 4 overall pick, said Walder, referring to the USC wide receiver who’s expected to be drafted closer to the teens. “Well, that would have really big implications on all the probabilities of [picks] five and six in a way that other people might not be expecting. So that information would be huge.”

Every draft has highly volatile variables. Often it’s the quarterbacks. According to ESPN’s draft predictor model, two of the highest-probability outcomes for Ty Simpson, widely considered the draft’s second-best QB prospect, are 18 slots and one round apart. ESPN projects that there’s about a 7% chance he’s drafted No. 16 overall, a pick currently held by the New York Jets, and about a 6% chance he’s drafted No. 34 overall, a pick currently held by the Arizona Cardinals.

Trades can also be chaos agents. DeCosta acknowledged after drafting Rosengarten that he almost agreed to a deal with the Kansas City Chiefs to move back two spots in the second round, behind the San Francisco 49ers. But the deal fell through, and after the Ravens took Rosengarten, the Chiefs moved up a spot to draft another offensive tackle. DeCosta said he “had some people tell us” that both the Chiefs and 49ers had wanted Rosengarten.

“The really hard part of the draft is reacting to that in quick time,” said Sabin, now a lecturer in statistics and data science and a senior sports analytics fellow at the University of Pennsylvania. “Look, I’m biased, but I’ve always viewed it that if you go purely on emotions and not on data in the heat of a moment, you’re more likely to make a poor decision — versus, if you have good data and models to back that up, it should be able to talk you off the ledge in a high-stress situation. Any behavioral psychologist will tell you that our brains don’t work as well in quick-thinking moments in terms of thinking rationally. And that’s the job of a GM, is to be rational.”

Dynamic modeling helps. As the draft chugged along in Tennessee two years ago, Carthon watched the Titans’ odds with the prospects atop their board rise and fall. He said he approached every pick with two to four players at each key position under consideration, open to the possibilities of the board falling his way or falling apart. Stick and pick? Trade up? Trade back? With the Titans’ new model, he better understood the draft’s landscape — or he could at least claim to.

Secrecy is the standard across teams’ draft operations. Yam, whom the analytically minded Ravens did not make available for this story, called the team’s work in the field “constantly evolving” during an interview with a publication for his alma mater, Skidmore College. He also said he “can’t really divulge our secret sauce.”

As with any half-decent mock draft, there’s plenty of educated guesswork involved. In March, DeCosta said he signed outside linebacker Trey Hendrickson to a four-year, $112 million contract in part because he believed the pass rushers the Ravens would “probably covet” in the draft would be unavailable by No. 14 overall.

On Thursday, the Ravens and the rest of the NFL will test that theory. In their attempts to predict the future, they are learning how to perhaps better control it.

“I think a good organization has done 90% of the work before the draft happens,” Sabin said. “They don’t know what’s going to happen the day of the draft. But if they are prepared, they will always know what the right answer is when it comes time.”