Two files, one game. Do they agree?

That is Jude Menezes, talking about Paris, describing what he says to a player who has just come off (Insights from the Paris 2024 Olympics: An AMA with Jude Menezes). The laptop is on the bench. The number is physical. The instruction is tactical, and it is given quietly at the interchange rather than saved for the review. His reason, in his own words: "It's not waiting for after the game because once the game's over, it's over. Then you're thinking about the next game."

Hold onto that, because it's rarer than it sounds. Only running data is on his screen, and he reads that physical number through a tactical lens, because, in his words, "our game is based on transitions". Almost everywhere else in the sport the two kinds of data live in different hands, and the question a head coach actually wants answered sits exactly between them.

The tactical data and the physical data get collected by different people, in different tools, for different questions. The analyst tags the video. The strength and conditioning coach reads the running data for load and injury risk. Overlays that put them on one timeline do exist at the top tier, from Kitman Labs, Catapult and STATSports, built mostly for football and rugby, and none of them names a hockey customer for it. The published hockey work on joining the two is academic, not operational. And no tool turns a tagged moment and its physical cost into an answer. A coach still has to do that. What AI can realistically help with in the near term is narrower: reading one table that has both kinds of column in it.

One file sits with the analyst: the event export from Sportscode or Nacsport or whatever tool you prefer. Every circle entry, outlet and penalty corner keyed by hand. The other sits with the strength and conditioning coach: the running data from the units the players wear, Catapult at the Dutch federation, STATSports at Germany, Ireland and New Zealand.

Look at what the second one is bought for. The KNHB's July 2025 announcement with Catapult is framed around performance and "duurzame hockeycarrières", sustainable careers, with load and recovery monitoring for every national team from under 16 up. The staff around the German men's team, talking to STATSports, describe live thresholds with haptic alerts: when a player crosses an individual threshold they have set, the staff get a notification during the session. The most common threshold is a speed: high-speed running is simply distance covered above a set speed, with the threshold written into the software. Phil Moreland used above 18 kilometres an hour for the women he worked with.

None of that is a tactical question. And the tactical side is often better organised than we assume. In Belgium, as Emily Calderon put it, all six league games are coded by Monday noon. Two well-run systems. Two people. And, as far as I can tell, no shared file.

By refusing to let the physical number travel on its own. Phil Moreland, who did this for Hockey Australia and then for the Scottish women, starts with the coach: "So coach, did player x do what you asked her to do in the game? Yes. She did. The data supports that. No. She didn't. Well, here's the GPS data." His line about it is blunt:

That's a join held together by a person rather than by software. Wouter Blondeel's tournament habit of watching his midfielders' high metabolic load distance is the same thing from part two. He explains the metric himself: "It's about high speed running, and it's about explosive distance... The explosive distance has to do with accelerations and decelerations" (Mastering Physical KPIs in Field Hockey). So it counts hard accelerating and braking as well as sprinting, and it is a physical read with a tactical consequence, living in one person's head and not in the event file.

You've probably got neither file, or a tag sheet and nothing else. Hold that thought, it's the fifth section.

Cunniffe and colleagues, working with the Irish men's team, published online in 2023 what they call "the first study to investigate the influence of phase of play" in men's international hockey. They took 10 Hz GPS data, ten position readings a second, from 24 players and lined it up with the phases of play an analyst had coded from video across ten internationals. Not "how far did he run" on one screen and "how often did we press" on another. Both at once: what does pressing cost in metres and in high-speed running. Offensive pressing took about 16% of playing time but 8% of total distance and 13% of high-speed running, and it produced the highest top speeds of any phase. So on these numbers pressing costs less volume than its share of the game, and the most top speed, which is my reading rather than the authors'.

Put that next to the Sunderland finding from part two, that goals came faster when the ball was won in the attacking 25, and the coach's question writes itself. What does winning it high cost, and when does it stop being affordable?

The rest I'll leave as questions, because that's what they are. Does the outlet pattern that loses the ball in the third quarter coincide with the drop in high-speed running, or not? When the press stops working late on, is it the legs or the shape? Is the structure that concedes less also the one that runs less?

Sometimes by using the physical numbers to settle a structural decision before a tournament even starts. Mick Beunen, who by 2020 had been the physical coach for the Belgian men for over ten years, tracked high metabolic load distance per midfielder across a tournament, and the totals shift depending on whether the team rotates four midfielders or five. His question to the coaching staff is the tactical one: "If you play with four, are they able to produce what you want them to produce?" (Physically game ready - by Mick Beunen).

Notice what that isn't. It isn't a claim that fatigue caused anything. It's a physical number shaping a selection question, asked in a room where both sides are present.

You can ask the same questions with much less precision, and quarter by quarter is precise enough. More on that below.