Two streams at the same resolution can look completely different. The variable is how the encoder spends a limited bitrate, and that spending follows rules worth understanding.
Bitrate is a budget spent per second
Video compression works by describing what changed between frames rather than sending each one whole, and bitrate is how much description is affordable per second.
A static scene needs very little, so a talking head at modest bitrate looks clean. A scene where everything moves needs far more to describe the same interval.
When the budget runs short, the encoder discards detail, which appears as blocking, smearing and colour banding in exactly the moments viewers care about.
Fast games are the hard case
Rapid camera movement changes every pixel between frames, so almost nothing can be reused and the encoder must describe the whole picture repeatedly.
Particle effects, foliage, smoke and rain are similarly expensive, because each contains fine detail moving independently and compression handles that badly.
This is why competitive shooters look worse than slower games at identical settings, and why lowering in-game effects can improve a stream more than raising resolution does.
Resolution and frame rate compete for the same budget
Doubling the frame rate roughly doubles the number of pictures to describe, and raising resolution increases the detail in each one.
At a fixed bitrate, choosing both means each frame receives less, so a high-resolution stream can look softer than a lower-resolution one from the same connection.
Matching resolution and frame rate to the available bitrate usually produces a better result than sending the largest numbers the software will accept.
Hardware and software encoders make different trades
Dedicated encoding blocks on modern graphics cards are fast and cost almost no game performance, which is why most streamers use them.
Software encoding on the processor can achieve better quality at the same bitrate by searching harder for redundancy, but it competes with the game for processing time.
The right choice depends on whether the machine has spare processor capacity, which is why the recommended setup differs between a single machine and a dedicated capture computer.
The platform sets the real ceiling
Services cap ingest bitrate to control their own costs, so sending more than the cap achieves nothing regardless of connection speed.
Newer codecs compress more efficiently at the same bitrate, but adoption depends on every viewer's device being able to decode them, which lags well behind the encoding side.
Until that catches up, most quality gains come from spending a fixed budget more carefully rather than from having a larger one.