How it works
A game may average 100 FPS while occasionally pausing long enough to be distracting. A low-percentile result focuses on slower frames. Tools calculate this differently: averaging a group of slow frames is not automatically the same as converting a frame-time percentile into FPS.
Important differences and limits
The result describes one recorded run. Loading screens, short recordings, shader compilation and background work can disproportionately affect the slowest frames. Compare published numbers only when the recording boundaries and calculation method are known; otherwise the apparent difference may come from the test itself.
A useful practical check
Record the same route several times after a comparable warm-up. Keep resolution, graphics settings and frame caps unchanged. Examine the frame-time graph alongside average FPS and the low result. A repeating spike at one location is more useful for diagnosis than a single unusually low score.
An example in practice
For example, exclude a deliberate loading screen from a gameplay test, but record that exclusion. Keep an unexpected in-play pause in the capture: removing inconvenient frames would conceal the very interruptions this statistic is intended to describe.
Additional examples and checks
A review that reports a large improvement in this figure should also explain the test. If the scene or recording length changed, the result may not isolate the hardware difference. A stable test method makes the low-percentile number useful for selecting a setup.