When it comes to analytics, whos the real Mike McCarthy?

June 28, 2024

The Athletic has live coverage of Cowboys vs. Bills

PHILADELPHIA — Dak Prescott launched himself head-first and took flight just as two Eagles defenders closed in on third down to send the Cowboy careening to the 1-yard line.

Dallas wasn’t dead yet. Prescott knew without having to look to the sideline that he would stay on the field for fourth down. The Cowboys were down 11 points with 10 minutes left, and going for the touchdown here and then two points was the analytically sound strategy.

But Luke Schoonmaker, the big rookie tight end, didn’t take his route quite deep enough, and he went down with the ball just a few inches short of the goal line. After Dallas’ eventual 28-23 loss on Nov. 5, head coach Mike McCarthy was asked about his decision to try for the failed touchdown.

“There’s a distance once you cross a certain yard line,” he said. “Whether it’s fourth-and-2, fourth-and-3 … that’s echoed before the series, during the series, we have individuals that are responsible for that.”

McCarthy dropped an important crumb in that vague answer: We have individuals.

The Cowboys hired five people this offseason to make up a brand-new analytics staff, the first analytics expansion in Dallas during the McCarthy era. There are more Brains in Dallas than ever before, and this year, they’re actually welcome to sit with the Football Guys. Literally: The Brains are now given a seat in the coaches booth on game day (three seats, to be exact) and are included in all meetings.

Just before he was hired in Dallas in January 2020, McCarthy touted his interest in analytics — a buzzy topic and something he wasn’t particularly known for in his 13 years in Green Bay, where he was often criticized for clock management. He told reporters he’d spent his year away from coaching studying league-wide trends and watching every play of the top 10 NFL offenses. He visited Pro Football Focus’ office in Cincinnati to learn more about the information the sports analytics company could provide, and he made an analytics vision board imagining his dream eight-person staff, complete with an imaginative role called the Coordinator of Mathematical Innovation.

“Everybody has analytics,” McCarthy told NFL Network at the time. “But it has to be part of your everyday operation to show up on Sundays.”

In 2023, the NFL is firmly established in its analytics era. Broadcasts have evolved to regularly reference win probability and network analysts often criticize conservative coaching. This analytics shift isn’t new, and these staffing additions wouldn’t be much of a story if they happened somewhere else. But McCarthy is a special case, because he’s become the mascot of the analytics battle taking place inside every head coach.

Analytics can mean a lot of different things. It can be used in decision-making in games and in player evaluation, and in managing the salary cap. Job titles are often intentionally broad to protect proprietary strategies.

However you define it, though, the things McCarthy has said and done about analytics have never quite lined up. He played it down during his Green Bay tenure, when he was actually diving into it behind the scenes. Then he played it up ahead of being hired in Dallas — but nothing substantially changed with his analytics staffing for three years.

“We’ve been slow to it,” Cowboys executive vice president Stephen Jones told The Athletic in October. “I will say that. I don’t know that we’ve been behind, but we damn sure hadn’t been setting the pace.”

The Cowboys look like a contender every regular season under McCarthy, but have not advanced past the divisional round. The Jones family is impatient.

“It’s been a long time since we’ve won a Super Bowl,” Stephen Jones said. “So, if you’re not getting better, you’re going backwards.”

McCarthy’s reputation as someone who does lip service to analytics isn’t totally fair, though it’s largely self-inflicted. Is this the season McCarthy finally figures out his relationship to analytics and everything lines up for these 9-3 Cowboys?

Mike McCarthy (photographed in Philadelphia during a Nov. 5 loss to the Eagles) has often had a disconnect between his words and his actions on analytics. (Tim Nwachukwu / Getty Images)

In McCarthy’s introductory news conference at The Star, the term “analytics” came up six times, “technology” twice, and “innovation” twice more. It was a central theme to his first appearance as Dallas’ head coach. The narrative was that this interest in numbers and trends was new for McCarthy, part of his reincarnation.

But in Green Bay, McCarthy’s staff had actually been near the forefront of the analytics takeover that swept the NFL during the 2010s.

At some point around the 2010 season, the year Green Bay won the Super Bowl, McCarthy went to Packers director of research and development Mike Eayrs with an observation.

“The scouting report that we have today is almost identical to the scouting report we had when I was working for the Chiefs,” Eayrs remembers McCarthy telling him. “I’ve got to believe that there’s more to it than that.”

McCarthy’s curiosity pushed Eayrs to look past basic stats like yards per play and points per game and get creative. “One of the things that I noticed about scouting reports was they covered from the snap to the whistle, “ Eayrs said. “And that really isn’t giving you the entire picture.”

So Eayrs focused on what happened in between the whistle and subsequent snap. He started timing that in-between by hand, and he enlisted pro scouts attending the Packers’ upcoming opponents’ games to do the same. They took down time stamps: When the huddle was formed, when the offense got over the ball, when the ball was snapped. He called this new stat “tempo” and he identified seven speeds, from “presto” to four-minute.

Eayrs, now retired, is considered the grandfather of football analytics. While working for the Minnesota Vikings in the 1990s, he was the first to mine the NFL’s play-by-play data and organize it into a searchable database with 76 data points (now hundreds more) on every game played. He first coined and defined the “explosive play” stat. When McCarthy took the head coaching job in Green Bay in 2006, he fired many of the existing staff to install his own guys, but he retained Eayrs after meeting with him and hearing his list of analytical ideas the Packers should incorporate. Eayrs worked for McCarthy until he retired in 2015.

“Coach McCarthy never got enough credit for his role in implementing (analytics) into our culture,” Eayrs said.

A little later in his Green Bay tenure, McCarthy discovered PFF because, as PFF founder Neil Hornsby tells it, McCarthy wasn’t happy with the grades his defensive coaches were giving their players. “You’re all giving your guys winning grades, but we’re getting the s— kicked out of us,” Hornsby recalled McCarthy thinking. “How do you square that circle?”

So McCarthy had Eayrs compare his coaching staff’s player grades against PFF’s grades to chart the differences. Eayrs said McCarthy once invited a NASA rocket scientist to observe training camp practices with an eye for anything that Green Bay could improve on.

“He wasn’t a guy that was going to sit and read a ton of reports,” Eayrs said. “But he was always on the lookout for ideas and concepts that he thought would make us a better team, and a better organization.”

When McCarthy visited PFF for a half-day in 2019 during his year of self-improvement, they discussed fourth down decisions. One person who was there, who requested anonymity because they weren’t authorized to discuss the visit, said that PFF staffers pulled up some old Green Bay tape to prompt the coach to reflect. Among the clips were the 2014 NFC Championship Game, when the Packers kicked two field goals from the 1-yard line to go up 6-0 in the first quarter. They lost 28-22 in overtime. Did McCarthy wish he had those decisions back now?

Per ESPN analytics, the Packers win probability if they went for it on that first fourth-and-1 was 49.7 percent. Kicking brought it down to 45.6 percent. On the second fourth-and-1 the win probability to go for it was 64.6 percent and to kick was 61 percent. That’s a combined difference in win probability of 7.7 percentage points, with a Super Bowl on the line. (It’s worth noting that because football is a sport with small sample sizes, these models have an unknown margin of error.)

Instead of acknowledging that going for a touchdown would have been a reasonable or better choice, McCarthy challenged PFF’s fourth-down modeling and defended his decisions from five years earlier, citing the weather, the wind, and the crowd noise in Seattle. Yet, in the same meeting he demonstrated his interest in having more information to help him coach better, like measuring individual player value with Wins Above Replacement. He’d even passed along some of his own ideas of what PFF should include in their models.

Every week in Green Bay, McCarthy held a staff meeting the day after the game to review specific game management situations. “We had a fourth-and-4 here,” Eayrs said. “And should we have punted? Should we have kicked the field goal, should we have gone for it?” Eayrs would then consult his database and set parameters for teams in the same situation as Green Bay. How many chose to punt and what was the outcome of the game? How many chose to go for it?

“I actually probably prevented him from becoming a better analytics guy,” Eayrs said. “I’m just being totally candid. Because when we’d look at the different alternatives, I’d say, historically, you’re looking at a difference of 5 or 6 percent. And I said, honestly, there’s no clear scientific trend here.”

So Eayrs encouraged McCarthy to go with his intuition — enough to send any current NFL analytics staffer reading this screaming into the nearest pillow, since they’d all consider a 5 or 6 percent swing in win probability significant.

But within each NFL head coach are two wolves. One wolf coaches by gut instinct shaped by years of anecdotal job experience and uses words like “trust” and “feel” to defend choices. The other wolf makes evidence-based decisions grounded in numbers, like win probability, and doesn’t bend when the choice doesn’t work out as planned. Which wolf does McCarthy feed more?

“McCarthy has not been shut off to this stuff,” says Eric Eager, vice president and partner at SumerSports, a sports analytics startup, who met with McCarthy when he visited PFF for a half-day in 2019. “He’s a very smart coach. He’s not a meathead. … But he’s also an incredibly successful coach, who probably trusts his gut. So those two things are always at conflict.”

In McCarthy’s debut game in 2020 at the L.A. Rams, he left his offense on the field on fourth-and-3 at the Rams 11-yard line with 11:46 left in the fourth quarter. Dallas was down 20-17 and could have tied the game with a chip-shot field goal. Instead, wide receiver CeeDee Lamb caught the ball on a shallow crosser that was a yard too shallow, and the Cowboys lost. Afterward, reporters asked McCarthy about the decision.

“Mike, what role did analytics play in that?” a reporter asked. “I know you are a big believer in that.”

McCarthy chuckled at the words “big believer.”

“Analytics will tell you to go for it there,” he said. “But there are other factors. … You have to look at your variables and trust your players. The first time going out into a game, I want our offensive guys to play wide open. We have that capability. I’ll pull back when I feel like it’s in the best interest of the team.”

You can hear the two wolves inside McCarthy tearing at each other in that answer. It’s not easy to withstand the media hounding when the analytically sound decision doesn’t pan out. And this inner turmoil is not unique to McCarthy (hello, Brandon Staley). But he’s become emblematic of it because he occupies the biggest and most important seat in football; because he’s regularly criticized for clock management decisions; and because his public comments are often at war with his in-game choices.

This offseason, before the new analytics staff was hired, McCarthy gave an answer that sounded like an anti-analytics bingo card when discussing moving on from his former offensive coordinator and play-caller Kellen Moore.

“I’ve been where Kellen has been,” McCarthy said. “He wants to light up the scoreboard. But I want to run the damn ball so I can rest my defense. … I don’t desire to be the No. 1 offense in the league. I want to be the No. 1 team in the league with a number of wins and a championship. And if we gotta give up some production and take care of the ball better to get that, then that’s what we’ll do, because we have a really good defense.”

That stone-age football statement illustrates the dichotomy between how McCarthy presents himself and what he’s actually doing — because Dallas is passing significantly more than it did last season, when Moore was calling plays. This year, the designed pass rate with garbage time eliminated is 61.5 percent, 12th in the league. Last year, Dallas had the 26th-lowest pass rate at 54 percent (league average was 59.3 percent). And during McCarthy’s tenure in Green Bay (2006-2018), the last time he called plays, his Packers were constantly passing. They were second in the NFL in designed pass rate over that time (63.5 percent).

In his second season in Dallas, McCarthy was asked how he’s affected by criticism of his clock management and decision-making. “I get it, analytics are great, but analytics has become a weapon for the media,” McCarthy said.

“I don’t think it’s his fault that people made it out to be like he was going to take differential equations and numerical analysis for a semester,” Eager says. “He came to PFF for one day, and he got some information on how to use us.”

Because Dallas beat Seattle last Thursday, McCarthy didn’t face a blistering analytics interrogation from reporters postgame. But there was plenty to wonder about. McCarthy didn’t go for two after a touchdown brought Dallas within two points of Seattle’s 28-26 lead at the end of the third quarter. And he decided to kick a field goal on Dallas’ final possession to go up six points inside the two-minute warning, leaving the Seahawks 1:43 on the clock with a chance to score the game-breaking touchdown (on third down before the kick, Prescott threw an incomplete pass instead of running out more clock).

“There are some aspects of the broad swath that is analytics that coaches see value in,” says Adam Vonder Haar, who was one of two full-time analytics staffers in Dallas until he left in 2022. “Like fourth-down models and going for two down eight, where they’re like, yes, this makes sense. That’s something McCarthy did while I was there. And there are other (aspects), where the burden of the evidence hasn’t quite swayed them from their years of football experience.”

The Cowboys declined The Athletic’s request to speak to McCarthy or John Park, the new leader of the analytics team, to discuss what the new staff is doing differently this season and the reason they were hired now.

NFL teams typically treat information about their analytics staff like state secrets, and that’s especially true in Dallas. The Cowboys probably know the conclusion we’re ready to jump to: That back-to-back head-scratching, one-score losses to the 49ers early in the playoffs led to McCarthy finally doing the thing he said he wanted to.

Stephen Jones cautions against jumping of any sort. “Probably too much of a leap of faith to say, ‘Well, it’s two losses to the 49ers,’” he said. “I think it’s overall (performance). How do we improve as an organization?”

Dak Prescott and the Cowboys suffered their second straight one-score playoff loss to the 49ers in January; the expanded analytics team will try to prevent a third early exit. (Thearon W. Henderson / Getty Images)

Analytics staffers around the league took note of Dallas’ summer hiring spree. This looked like a team doing whatever it takes to get an edge by hiring respected talent. The Cowboys hired Park away from the Colts to head up the department and gave him freedom to build out his own staff. Park played high school football and has degrees from Duke and Columbia. He first worked as an insurance actuary, until he pivoted back to football at age 30 and picked up an entry-level coaching role at Rutgers. From there he went to the league office and then, with the Colts, found his footing applying his math and data analysis strengths to in-game decisions and player evaluation.

Park hired one of his analysts, Sarah Mallepalle, from the Ravens, a team known for its robust analytics department, and another, Bryant Davis (who has a doctorate in statistics) from the Tampa Bay Rays, who have one of MLB’s biggest analytics staffs. (In baseball, that means something else entirely — the Rays reportedly have more than 40 analytics staffers.) William Britt, who played football in college, came to Dallas from the Steelers, and engineer Max Lyons from a video game developer.

When McCarthy arrived in Dallas in 2020, he inherited a two-person full-time analytics squad, and then designated four members of his coaching staff into hybrid coaching and analytics roles.

Kyle Valero was one of those coaches until this offseason, when his contract wasn’t renewed. Valero was an offensive assistant for tight ends on Jason Garrett’s Dallas staff, and McCarthy re-hired him as quality control/analytics. “But nothing really changed other than it was giving a title to things that had already been done,” Valero said.
He still spent about a third of his time working on the analytics part of his title, creating reports on opponent and self-scout tendencies, and assisting other coaches in pulling data from platforms like PFF.

“Coach McCarthy pushed the term ‘analytics’ more than the other coaches I worked with,” said Valero. “I would attribute it to, right now, if you’re starting a company and you say we use cutting-edge AI, the times today, (people) are going to hear ‘AI’ and get excited about it. I think Mike, with the times now, is using the term more.”

The previous analytics staff met with the coordinators individually but weren’t in team meetings with full staff and players, because the hybrid coach/analytics staffers bridged that gap. In Indianapolis, Park eliminated the chasm between Football Guys and Brains, and he’s integrated his staff in the same way in Dallas.

“They are in every meeting,” said 10-year Cowboys veteran Zack Martin. “Every week, Coach is putting numbers up in front of us, tendencies, production numbers.”

What Martin mentioned is pretty basic analytics work, but, as Stephen Jones said, “there’s a lot more in the soup.” Jones said Park and his staff are involved in “everything” — from coaching decisions and game-planning to scouting and draft evaluations and free agency.

“You’re still probably years away from really seeing it end-to-end,” Eager said. “Here’s a data pipeline, here are decisions.”

Even if it’s too early to definitively evaluate the new analytics staff, there are a few analytically accepted on-field concepts that hint at its impact.

As mentioned, McCarthy is calling more passes this season, and the Cowboys are passing way more on first down, traditionally a low-risk run down, which is a sign of analytics influence. Per PFF, this season Dallas is passing at the fifth-highest rate above expectation on first down, 5.4 percent, the highest in McCarthy’s tenure.

They Cowboys have used play action more this year than any other under McCarthy (17 percent of plays, fourth in the NFL), up from 14.7 percent in 2022, and they’ve diversified their personnel groupings. Dallas has already run 30-plus snaps with an NFL-high six personnel groupings; it had never before run more than three personnel groupings at that rate.

In 2020, per PFF, Dallas’ offense motioned slightly above league average, but then dipped below league average in 2021 and 2022. That trend continued until their Week 7 bye this season. After the bye, a time of introspection when an analytics staff and coaching staff will combine to incorporate any edge they can find, Dallas has jumped up to using motion on 56.1 percent of plays, ninth in the NFL. This looks like the analytics staff’s impact, but as one of the wolves in a coach’s head might argue: You can twist the numbers to tell any story.

Dallas actually is lighting up the scoreboard this year, averaging 28.2 offensive points per game. Prescott is playing at an MVP level, and the Cowboys are looking like they could be Super Bowl contenders. But the Seahawks are the only team with a winning record they have beaten, and they have lost to two other Super Bowl contenders, the 49ers and Eagles. They face Philadelphia again this week on “Sunday Night Football.”

“It’s a game of inches,” McCarthy said in boilerplate coach speak after the 28-23 loss to the Eagles in Week 9. “We came up an inch or two short on a couple of plays.”

This new analytics staff has a big job to make sure these Cowboys advance further this postseason: Find those missing inches.

(Top illustration: Eamonn Dalton / The Athletic; photo: Andy Lyons / Getty Images)

ncG1vNJzZmismJqutbTLnquim16YvK57lGpnbXFhbXxzfJFsZmpqX2WFcLnIpJxmpZOYrrPAx7JkmqaRoca1tcKsZJ2ZnKGutHnCqK6bp6mofA%3D%3D