AI video analysis in football: what a grassroots club can do

What AI video analysis in football actually does today

The term covers several different technologies, and they are not equally mature. Mixing them up is the most common reason clubs buy something that disappoints.

• Automatic camera framing. A fixed wide-angle camera, or two lenses stitched together, records the whole pitch and software crops a virtual "broadcast" view that follows the ball. For a full-size pitch filmed from a good height this works well enough for coaching.
• Event detection and auto-clipping. Models trained on thousands of matches flag shots, goals, set pieces, possession changes and sometimes passes, then cut clips around them. Accuracy is good for clear events (a shot, a corner) and weaker for subtle ones (a missed pressing trigger, a poor body shape).
• Player tracking from video. The software tries to follow each player and build x/y positions over time. This is where amateur footage struggles: plain kits, low resolution, players blocking each other and no readable shirt numbers lead to identity switches.
• Derived metrics. Distances, sprints, heat maps, xG and passing networks are calculated from the tracking or event data. They inherit every error in the layer underneath.

For context, the professional end uses a very different set-up. FIFA announced that each stadium at the 2026 World Cup would have 16 optical tracking cameras producing more than 150 million tracking data points per match. A club with one camera on a pole at a council pitch is solving a different problem, and should not expect the same precision.

A one-camera set-up that works for a grassroots club

Most of the quality of grassroots football video analysis is decided before any software runs. A poor angle cannot be fixed by an algorithm.

1. Height. Get the camera as high as you safely can, on the halfway line. A telescopic pole or a stand at the top of a small terrace is better than a tripod at eye level, because the far side and both penalty areas stay visible.
2. Position. Centre the camera on the halfway line, a few metres back from the touchline, away from the technical areas and from where spectators stand.
3. Framing. Check that both goal lines and the far touchline are in frame before kick-off. Record a 10-second test and play it back.
4. Resolution and power. Record at least 1080p. Bring a power bank and enough storage for the match plus warm-up; a dead battery at 60 minutes is the most common failure.
5. Consistency. Use the same position every home match. Comparable footage matters more than perfect footage.

Plan for the time cost honestly. Setting up and taking down takes 10–15 minutes. Uploading and processing usually runs overnight. A coach who reviews only auto-generated clips for two or three themes can do a useful review in 30–45 minutes; a full manual tag of a match takes several hours and is rarely sustainable for a volunteer.

From clips to decisions: ask fewer, better questions

Software can produce dozens of charts per match. Few of them will change what happens at Tuesday training. Start from the game model and choose two or three questions the coaching staff actually want answered, then pick the simplest output that answers each one.

The fourth row matters. Physical metrics from a single wide camera are estimates built on top of imperfect tracking. If load is the question, see the article on wearables, GPS and injury prevention instead.

Checking the output before you trust it

Every automatic system makes mistakes. The coach's job is to know how many, and where. A simple validation routine takes about 20 minutes per match for the first few weeks and much less once you know the tool.

1. Spot-check a sample. Pick 20 auto-tagged events at random and watch each one. Count how many are correct, mislabelled or missing. If more than a few in 20 are wrong, treat that event type as unreliable.
2. Look for missed events. Watch one 10-minute block in full and tag it yourself. Compare with what the software found. Missed events bias the data more than wrong ones.
3. Check identity. If the tool assigns events to players, follow two or three players for five minutes. Identity switches are common when kits are similar or players cross.
4. Do not read one match. A single game is mostly noise: the opponent, the pitch and the weather drive the numbers. Look for patterns across four to six matches.
5. Write down what you changed. If a training block follows from the data, note it and look at the same question four weeks later.

Running tournaments is one of the most time-consuming jobs at a grassroots club: registrations, the draw, the schedule and results. DorsalCUP does it for free, and the tournament fixture generator builds a one-day schedule in minutes, no sign-up.

Turning a finding into training

Video is only useful if it changes the session. Using the example above, the problem is the right-back receiving with his back to play under pressure from the winger and the inside midfielder.

Build-up under pressure: 6+GK v 5

• Space: 50 m wide x 40 m deep, from the goal line; two mini-goals on the halfway line for the defending side to counter into.
• Players: goalkeeper plus back four and two midfielders against five pressers.
• Rules: the build-up team scores by dribbling or passing through a 10 m wide gate on the halfway line; pressers score by winning the ball and finishing in the full-size goal within 8 seconds.
• Duration: 4 x 5 minutes, 2 minutes' rest; rotate pressers.
• Coaching points: full-back receives on the half-turn with the back foot, open to the pitch; centre-back passes to the outside foot; the nearest midfielder offers an inside angle before the ball arrives.
• Progression: the presser on the full-back's side may press from behind; then the full-back has two touches maximum.

Show the players six to eight short clips, not a full half. With under-14s and above, a 10-minute video meeting before the session is enough. With younger players, keep video short, positive and focused on the team, as the FA's filming guidance also recommends for grassroots youth football. Related tactical detail is in the article on build-up against a high press.

Costs, risks and filming minors

Hardware for automatic recording ranges from a smartphone on a pole to dedicated AI cameras sold on subscription. Prices and licence models change often, so compare the total cost over a season: camera, mount, storage, subscription per team, and the volunteer hours to run it.

• Footage of players is personal data. Under UK data protection law the club needs a lawful basis, a privacy notice and secure storage.
• Tell people first. The FA's guidance on photographing and filming children says to tell players and parents that filming is part of the coaching programme, to offer an opt-out, and to inform the opposing club before filming.
• Check the contract. Ask where footage is stored, who can access it, how long it is kept and whether the supplier uses it to train its models.
• Limit access. Share clips through the platform with named accounts, not through public links or group chats.

The full picture on consent, retention and the EU AI Act is covered in filming youth football and data protection.

Frequently asked questions