How to analyse sports video: a step-by-step workflow

Most coaches with footage never watch it.

The bottleneck is not recording — that problem was solved by cameras that film the whole pitch unmanned. The bottleneck is everything between the upload finishing and a player doing something differently on Saturday. That gap is where footage goes to die, and it is not a discipline problem. It is a missing routine.

Here is the routine. Seven steps, repeatable in under 30 minutes per match once you have run it three or four times.

Step 1: Decide what you're looking for before you press play

This is the highest-leverage step in the whole workflow and it takes two minutes.

Write down one or two questions before you open the footage. Not "let's see how we did" — an actual question with an answer.

• Football: where do we lose the ball in build-up?
• Basketball: are we getting back in transition after a made basket?
• Volleyball: what is our reception quality against jump serves?
• Handball: how many attacks end in a shot from outside nine metres?

The reason this matters is not tidiness. Watching without a question means you notice what is salient — goals, mistakes, the loud moments — which is exactly the material you already remember. The question is what makes you see the quiet, repeated, fixable things.

A useful test: if the answer to your question could not change what you do at Tuesday's session, pick a different question.

Step 2: Record footage that's actually analysable

Resolution is the thing people buy and the thing that matters least. In order of actual importance:

1. Elevation. Get the camera up. Pitch-level footage flattens the space between players and makes shape impossible to read. A gantry, a stand, a pole, a first-floor window — anything above head height beats a 4K camera on the ground.
2. Width. Frame the shape, not the ball. Ball-chasing footage is fine for highlights and useless for analysis, because everything you want to see is happening away from the ball.
3. Stability. A tripod beats a parent's hands. Shaky footage is unwatchable at the speed you need to review it.
4. Consistency. Same angle, same height, every match. You cannot compare two matches filmed from different positions, and comparison is the entire point.

Automated systems — Veo, Trace, Spiideo, Pixellot — solve all four by design, which is most of why they took over. If you are using one, what to do with the footage afterwards is the harder question. If you are choosing between them, Trace vs Veo covers the trade-off.

Step 3: Watch it once, all the way through, without tagging

Counter-intuitive and worth doing.

If you tag on the first pass, your tags describe what you expected to find. You had a theory on the touchline, and the first pass will dutifully confirm it — you will mark the four turnovers that fit your story and skim past the eleven that do not. This is selection bias, and it is the single most common way amateur analysis produces confident nonsense.

One pass at normal or 1.5× speed, no clicking. Then tag.

Step 4: Tag the events that answer your question

Now the coding pass. Four rules:

Keep the list to five to eight categories. Twelve is sustainable for three weeks. This is the most reliable finding in amateur analysis: ambitious coding systems are abandoned, and an abandoned system produces no baseline, which means nothing to compare against.

Tag the phase, not the outcome. The turnover you want to fix started three passes before the ball was lost. If you tag the moment of loss, your clip shows the symptom. Tag the entry into the phase.

Use a timestamp buffer. A clip that starts at the event is a clip with no context. Eight to ten seconds before is usually right; for set pieces, more.

Write the definitions down. "Pressing action" means nothing until you specify it. This is what makes your tags comparable between matches — and comparable between you and your assistant. Notational analysis, including how to check your own coding is reliable.

What Sidetalk does here, and what it doesn't

Sidetalk is not a tagging tool. It does not host your video, import it, clip it or draw on it, and there is no event-coding interface. If you want a tagged clip library, you need software that does that.

What it does is the layer tagging cannot reach: what you said at the time.

You put your phone in your pocket and coach normally. Sidetalk transcribes the touchline, works out which player each remark was about — you can tap a name while you talk, or not bother — and drafts individual feedback for every player in the squad. You read it, edit anything, and send it. Nothing reaches a player before you have seen it.

Then it lines up with the video. You mark kick-off with one tap during the match, and afterwards you scrub the footage to kick-off and type the time the video shows. That is one number, once per match. From then on every remark you made carries the video's own timestamp — so you can jump to the 23rd minute in Veo and read exactly what you said there, filtered to one player.

The two things are genuinely different jobs. Tagging tells you that the shape broke down at 23:14. Your own voice at 23:14 tells you why you thought it did — and by Monday that is the half you have lost. Why the two clocks approach works with any platform.

Step 5: Look for patterns, not incidents

One clip is an anecdote. The same clip four times is a coaching point.

Once tagged, count. Not elaborately — how many times did the thing happen, and where? Four turnovers in the same channel is a pattern you can train. Four turnovers in four different situations is probably just football.

Then get a baseline. The most useful one is your own last five matches, because it controls for level, opposition quality and your own coding habits all at once. Elite benchmarks are the wrong comparison for a developing team and will tell you your U14s are bad at everything. How to read your numbers without fooling yourself.

Step 6: Build the feedback, not the report

This is where most analysis effort is wasted. A beautiful report is not an intervention.

• Three to five clips per player or unit. Not twenty. Past about five, retention collapses and the player remembers only that they were criticised.
• Pair each problem with the same thing done right — ideally by the same player, ideally the same match. This is the difference between "you keep doing this" and "you can already do this, here it is."
• Lead with the question, not the verdict. "What can you see here?" produces a player who has worked something out. "You were out of position" produces a player who has been told off. The evidence is identical; the learning is not.
• Send it individually. Team review sessions mean eighteen players watching material where perhaps three clips concern them. The quiet full-back who did their job for seventy minutes gets nothing at all, because nothing about them made the reel. Feedback that lands with young players.

Step 7: Re-measure next match

Tag the same category again. Did the number move?

This step is skipped almost universally and it is the one that makes the other six worth doing. It is also your entire argument to a sceptical club: not "we do video analysis now" but "we lost the ball in build-up nine times in April and four times in May."

If the number did not move, that is information too. Usually it means the feedback was accurate and the training did not change, which is a coaching problem rather than an analysis problem.

How long this actually takes

Honest budget for a 90-minute match, after a few repetitions:

Around 30 minutes. The first two or three matches will take twice that, because you are still deciding your categories. Anyone promising five minutes is describing a tool demo, not a workflow.

Tools at each stage

Two things worth noticing in that table. Kinovea is free and better than most paid tools at frame-by-frame technique work — if that is your need, stop shopping. And the capture row and the tagging row are separate purchases, which is the single most common budgeting mistake clubs make.

Frequently asked questions

What is video analysis in sport?

Recording a match or session and reviewing the footage to understand what happened — where the shape broke, how a technique failed, whether a press was coordinated. In practice it is three separate jobs: capturing the video, breaking it into tagged clips, and getting an athlete to change what they do as a result. Most products handle the first two.

How long does it take to analyse a match?

About 30 minutes for a 90-minute match once the routine is established, spread across a first pass, a tagging pass and building the feedback. Expect an hour for your first few while you settle on categories.

Do I need special software to analyse sports video?

No. A spreadsheet of timestamps against a YouTube upload does the job, and Kinovea is free for technique work. Software saves time rather than enabling the analysis, which is why buying it before you have a routine tends to disappoint.

What camera angle is best for match video?

Elevated, wide, and central to the halfway line. Height matters more than resolution — from pitch level you cannot see the distance between lines, which is most of what you want to look at.

How do I share clips with players?

Individually, three to five per player, with a question attached rather than a verdict. Group chats get scrolled past, and team sessions spend most of their time showing each player material about somebody else.

Where to go next

Comparing tools — how to choose video analysis software. Understanding the discipline — what performance analysis is. And if step 6 is your bottleneck, that is the part Sidetalk handles.