Introducing phases of play
Context is everything
As futi gets closer to user testing, we’ll be sending out newsletters (and posting videos to our brand new YouTube, Instagram and TikTok channels) to introduce the data models that power our app.
Along with each newsletter, we’ll release accompanying data from the 2025 MLS season in online tables where you can explore or download stats.
Today’s release is from an important new model that informs a lot of futi’s analytics: phases of play.
How do you put football data into tactical context?
Most of the time, people don’t. A pass is a pass and a tackle is a tackle no matter how they happen. Football stats typically treat each action as an isolated event, ignoring what’s going on in the game around it.
That’s because most raw football data — what’s known as “event data” — doesn’t contain information about what’s happening away from the ball. Somebody watches the game and logs every on-ball action, adding details about when and where it happened, who did it, whether it was successful, stuff like that. There’s a lot you can do with that data, but it doesn’t tell you how the teams are set up or how players are moving off the ball. As somebody once put it, event data is like listening to the game on the radio.
To help fill in the missing context, futi developed a model based on a common framework that coaches and analysts use to describe the game: phases of play.
You’ve heard of the buildup, right? That’s a phase of play. So are counterattacks, corner kicks and so on. These are labels for familiar situations in a football match that help us describe what’s going on beyond “Well, a bunch of people sure are chasing a ball around.” Phases break the game into bite-size chunks to make it easier to digest.
Futi’s phase of play framework starts by divvying up the game into a few high-level categories: organized possession, transitions, set pieces and contested phases (when neither side has control of the ball). Each category contains multiple phases. Organized possession, for example, can be broken down into buildup, progression, fast break and finishing phases.
There’s no definition of these phases in the laws of the game, no official rules about when one ends and another begins. The names and exact number of phases can vary slightly from one coach to another, but you pretty much know them when you see them. They capture natural divisions in the game. When you get down to it, a phase of play is just a description of where and how the ball is moving and — this is the important part — roughly how the sides are organized.
Their relationship with team structure makes phases super useful for addressing the tactical context problem. Imagine two crosses with the same start and end location, same passer and receiver, same result. In isolation they look identical in the data. If you know one cross came during a fast break and another during the second phase of a corner kick, though, it’s obvious that they’re not similar at all. The sides are set up very differently, with players and the ball moving at different speeds in different amounts of space. The level of control is different. Pressure is different. The two plays probably have very different chances of scoring.
Knowing the phase of play isn’t quite turning on the TV, but you’re at least listening to the radio closely enough to have a pretty good mental image of what the game looks like.
Determining what phase the game is in from on-ball data is as valuable as it is difficult. Futi’s phase of play model was painstakingly assembled from hundreds of rules that determine when the game moves from one state to another based on information such as where the ball is, how fast play is moving, what’s happened in the last few actions, whether there was a set piece recently and so on. We spent weeks refining the rules against match footage, touch by touch, frame by frame, to make sure our phase labels accurately captured what teams were doing off the ball.
The resulting phase of play context is important to a lot of futi’s analytics, including the possession value model that powers our player ratings. It’s also pretty interesting data in its own right, since understanding what phases teams are most active in and how successful they are at them can be a powerful tool for understanding their style of play.
If you want to explore some phase of play data for yourself, you can do it in futi’s new online tables, where we’ll be releasing data from the 2025 MLS season to accompany this series of model explainers.
The phases tab contains three season-level stats for every team in each phase:
Count: The average number of phases of that type a team had per game.
Success: The percentage of those phases that the team “won,” meaning they ended the phase with a higher possession value than when they started.
Share: The percentage of the team’s total actions that happen in that phase.
Here’s a visual guide to phase definitions:
Organized possession1
Transitions
Set pieces
Contested
Phases are described here from the perspective of the team in possession, but each also has a corresponding defensive phase for the team out of possession. During contested phases neither team is considered to be in possession.
My god they’ve done it
👏👏👏