A live stream sends a fixed amount of data per second regardless of what is happening on screen. That budget is what determines quality, and different games spend it very differently.
Compression works by describing change
Video compression stores occasional complete frames and describes the rest as differences from what came before. A static scene requires very little data.
When the whole image changes at once, there is little to reuse, and the encoder must describe far more to keep the same fidelity.
With a fixed budget it cannot, so it reduces detail instead. This is why streams become blocky during rapid movement and sharpen when the camera settles.
Certain visual effects are expensive to encode
Particle effects, foliage, rain, smoke and confetti fill the frame with small independent movement, which is close to the worst case for compression.
Dark scenes with subtle gradients suffer differently, showing banding where the encoder lacks the data to represent gradual changes smoothly.
Broadcasters playing such games often reduce resolution rather than frame rate, since fewer pixels leaves more budget for each one.
Resolution and frame rate compete for the same budget
Doubling frame rate halves the data available per frame at a fixed bitrate, so a smoother stream is necessarily a less detailed one.
Fast action generally benefits more from frame rate, while slower games with fine detail benefit more from resolution. There is no setting that suits both.
This is why broadcasters change settings between games rather than finding one configuration, and why the same channel varies in apparent quality.
Platform limits are set by delivery costs
Services cap bitrate because they pay to distribute every stream to every viewer, and the cost scales with data rather than with audience enjoyment.
Caps are often tiered, with established or partnered broadcasters permitted higher rates, which is a bandwidth allocation decision rather than a quality judgement.
Transcoding to lower qualities for viewers on weak connections also consumes resources, which is why it is sometimes reserved rather than universal.
Newer codecs shift the trade rather than removing it
More efficient compression standards achieve similar quality at lower bitrates by analysing the image more thoroughly.
The cost is computation on both ends, and adoption depends on viewers having hardware that can decode it, which lags behind the broadcasters who can produce it.
Platforms therefore adopt new codecs slowly and in parallel with older ones, so the practical improvement arrives years after the technology does.