On this page
On this page
The short answer
Sports video vocabulary comes from three unrelated places: performance analysis (coding, tagging, KPIs), broadcast television (telestration, lower thirds, overlays), and computer vision and streaming (detection, homography, bitrate, HLS).
That is why it sounds inconsistent. The definitions below are grouped by origin, so the terms that go together appear together.
Reviewing and organising footage
Clip
A defined section of a longer video, marked by a start and end time. In most platforms a clip is a reference into the full game rather than a separate file, which is why creating one is instant and does not consume storage.
Coding
Systematically marking events as they occur, usually against a predefined list — every tackle, every entry into the final third, every lineout. Coding produces a structured record you can filter and count. It is the core activity of performance analysis and the thing dedicated analyst tools are built around.
Tagging
Attaching labels to a clip so it can be found later: a player, a phase, a theme. Looser than coding. Coding builds a dataset; tagging makes a library searchable.
Live coding
Coding during the match rather than afterwards, so the analysis is ready at full time or even at half time. Requires dedicated software, usually a button matrix, and somebody whose entire job it is.
Button matrix
The grid of buttons an analyst uses for live coding, each mapped to an event type. Configuring one well is most of the skill in live coding.
Playlist
An ordered set of clips, usually built by filtering coded events — every set piece, every clip involving one player. What you actually show in a film session.
KPI
Key performance indicator: a number chosen in advance as a measure of something you care about, such as ruck speed or centre pass conversion. Useful when chosen deliberately, meaningless when collected because the software offered it.
Drawing and presentation
Telestration
Drawing on video to explain it: arrows, circles, lines, zones, movement paths. From television plus illustration. See what telestration is for the longer version.
Layer
One level of annotation that can be shown, hidden or removed independently. Multi-layer telestration lets you build a picture in stages rather than drawing everything flat on one plane.
Overlay
Any graphic drawn on top of the video: a scoreboard, a clip title, a logo, a player highlight. Distinct from telestration in that overlays are usually generated rather than hand-drawn.
Lower third
A broadcast term for a graphic in the bottom third of the frame, typically a name or caption. Turns up in export settings.
Player highlight
A marker — a ring, a spotlight, a coloured cone — attached to one player to make them findable in a wide shot. Can be drawn by hand or driven automatically by tracking.
Freeze frame
A paused moment held on screen, usually annotated. Often the single most efficient coaching artefact you can produce.
Tracking and computer vision
Detection
Finding objects in a single frame and returning boxes around them, with confidence scores. Has no memory of previous frames.
Bounding box
The rectangle around a detected object. Usually expressed in normalised coordinates from 0 to 1, so it stays valid whatever resolution the video is played at.
Tracking, or association
Linking detections across frames so the same player keeps the same identity over time. This, not detection, is where tracking systems usually fail.
Track ID
The number assigned to a tracked object. Stable track IDs are the whole objective.
ID switch
When two tracked players exchange identities, typically after crossing. The characteristic failure mode of every tracking system, and hardest between teammates, who look identical by design.
Occlusion
One object hidden behind another. Sustained occlusion — a ruck, a goalmouth scramble — removes the information a tracker needs, so it must guess or drop the track.
Re-identification
Recognising that a player reappearing after an occlusion is the same one who disappeared, rather than a new person.
Homography
The mathematical relationship between the camera's image and the flat plane of the pitch. Solving it, usually from visible line markings, lets image positions be converted into real-world positions. When it cannot be solved reliably, an honest system omits world coordinates rather than estimating them.
World coordinates
Positions expressed in metres on a pitch model rather than pixels in an image. Only meaningful when homography succeeded.
Keypoints
Recognisable fixed features used to anchor the homography — corners, the centre circle, penalty-box intersections. A camera angle showing few keypoints makes pitch mapping unreliable.
Inference
Running a trained model on new data, as opposed to training it. Inference on video is GPU-intensive, which is why tracking is usually queued and why it is rarely unlimited.
Video, files and delivery
Bitrate
Data used per second of video. The main driver of both file size and picture quality. Higher bitrate means a sharper picture and a larger file.
Resolution
Pixel dimensions, such as 1080p or 4K. More pixels give more room to crop into a moment and produce substantially larger files. 1080p is the practical default for team review.
Frame rate
Frames per second. 25 or 30 is standard; 50 or 60 gives cleaner slow motion. Matters more than resolution if you intend to freeze moments.
Codec
The compression method used, such as H.264 or H.265. H.264 is the compatibility choice; H.265 produces smaller files but is less universally supported.
Container
The file format wrapping the video, such as MP4 or MOV. Distinct from the codec: an MP4 can contain several different codecs.
Transcoding
Converting video into other formats or quality levels, usually on upload. This is why a file is not immediately playable after uploading.
Adaptive streaming and HLS
Delivering video at several quality levels so the player can switch according to the viewer's connection. HLS is the common protocol. This is why a properly hosted video plays on poor mobile data while a raw file download stalls.
Progressive download
Sending one fixed-quality file and playing it as it arrives. Simpler than adaptive streaming and much worse on unreliable connections.
Unlisted
A visibility setting where a video is not in search results or on your channel but is watchable by anyone with the link. Not the same as private, which matters for youth footage — see player privacy and safeguarding.
Plans and pricing
Per seat, or per licence
Priced by the individual using the software. A second coach means a second licence. Common in analyst tools.
Per team
One price covering a whole squad, typically with a member limit. A second coach costs nothing extra. Common in participation-focused platforms.
Member limit
How many people can be in a team on a per-team plan. Count coaches, players and any parents or supporters you intend to include.
Storage allowance
How much video you can hold, in gigabytes or hours. Allowances in this market vary by an order of magnitude, and full-game video accumulates quickly. See what video analysis software really costs.
Capture-first
A platform built around its own recording hardware. The camera is required rather than optional.
Bring your own footage
A platform that works with video you already have, from any source. Most non-capture products are in this category.
Common questions
What is the difference between coding and tagging in video analysis?
Coding means systematically marking every occurrence of predefined event types during or after a match, which produces a structured dataset you can filter and count. Tagging means attaching looser labels to clips so they can be found later. Coding builds data; tagging organises a library.
What is an ID switch in player tracking?
An ID switch is when a tracking system exchanges the identities of two players, usually after they cross and separate. It is the characteristic failure of every tracking system and is hardest to avoid between teammates, because appearance is the main cue for telling players apart and teammates wear identical kit.
What does homography mean in sports video?
Homography is the mathematical relationship between the camera's image and the flat plane of the pitch. Solving it, normally by recognising line markings, lets a system convert positions in the picture into positions in metres on a pitch model. When it cannot be solved reliably, positions should be reported without world coordinates rather than estimated.
What does unlisted mean for a video?
Unlisted means the video does not appear in search results or on your channel, but anyone holding the link can watch it. It is not the same as private, which restricts viewing to specifically invited accounts. The distinction matters for youth footage, because an unlisted link can be forwarded and cannot be recalled.
What is adaptive streaming, and why does it matter for game film?
Adaptive streaming delivers video at several quality levels so the player can switch based on the viewer's connection, commonly using the HLS protocol. It matters because players watch on phones over variable mobile data, and an adaptive stream keeps playing at reduced quality where a single fixed-quality file would stall.