Adaptive bitrate explains why a picture softens exactly when the action speeds up

One event, several parallel encodings

A live stream is not encoded once but several times, producing a ladder of versions that range from very low quality to the highest the service offers. Each version is cut into segments aligned with the others, so a player can move between them at a segment boundary without any interruption. The player, not the service, chooses which version to request, based on how quickly the previous segments arrived and how much video it has in reserve.

This design is what allows one stream to serve a fast fixed connection and a weak mobile one without either being told it cannot watch. It also means the quality a viewer receives is the outcome of a continuous negotiation rather than a setting chosen once.

Why motion is the expensive case

Video compression works largely by describing how each frame differs from the previous one, which is very efficient when most of the frame is unchanged. A static wide shot of a crowd compresses to almost nothing, because the encoder can reuse most of the previous frame and send only the small differences. A rapid camera pan changes every part of the frame at once, so the encoder has almost nothing to reuse and the data required rises sharply.

Sport is full of exactly this kind of motion, and the moments of highest motion coincide with the moments the viewer most wants to see clearly. The result is that a stream is under maximum pressure precisely when quality matters most, which is an unfortunate but unavoidable alignment.

What a downswitch looks like

When the player detects that segments are arriving more slowly than they are being consumed, it requests a lower quality version to avoid running out. The viewer sees this as the picture softening, with fine detail such as grass texture, crowd faces and distant markings turning mushy. Because the switch happens at a segment boundary, it is not accompanied by a stall, which is the entire point of the mechanism.

The player will switch back up once conditions improve, though it does so cautiously to avoid oscillating between levels. A viewer who sees repeated softening and recovery is watching a connection that is marginal for the quality being attempted.

Why the same connection behaves differently at different times

The capacity available to a household is shared, both within the home and across the local segment of the provider's network. During a widely watched fixture, many households in the same area are streaming simultaneously, which loads the shared portion of that path. Inside the home the same effect happens on a smaller scale as other devices compete for the connection and for airtime on the wireless network.

The stream measures the result rather than the cause, so it simply sees slower delivery and responds by lowering quality. This is why an identical setup can deliver excellent quality on a quiet evening and mediocre quality during a heavily watched fixture.

What a viewer can actually influence

The elements within a viewer's control are the local ones, namely how the device connects, what else is using the connection and how far it sits from the access point. A wired connection removes wireless variability entirely, which is the single most effective change available in most homes. Reducing competing activity during a fixture helps for the same reason, since the player is responding to the capacity actually available to it.

Manually fixing the quality level is possible on some services but usually counterproductive, because it removes the mechanism that prevents stalls. Beyond the home, the path belongs to the provider and the distribution network, and nothing at the viewer's end changes what happens there.