Identify the kind of delay
| Symptom | Measure or observe | Useful next step |
|---|---|---|
| Progress stays in download | Bytes/progress and exact network error | Use loading troubleshooting. |
| Files finish but engine is preparing | Preparation stage and renderer | Keep tab foreground; record time separately. |
| Scene moves in visible jumps | Repeat the same scene at the same resolution | Compare render workload. |
| Mouse moves normally outside game | Pointer capture and sensitivity | Check controls before blaming FPS. |
| Entire tab crashes or reloads | Map, duration, other tabs and browser crash notice | Close duplicate heavy tabs; save details. |
Run a useful comparison
- Write down OS, browser/version, computer model or CPU/GPU, map, renderer, window size and fullscreen state.
- Keep the same scene, round, camera direction and test duration. A menu and a crowded scene are not comparable.
- Change one setting, such as resolution or renderer. Restart only as required, and label whether the map files were already cached.
- Repeat both conditions. If no frame counter or repeatable measurement is available, describe the observed stutter rather than inventing an FPS number.
- Include failures as well as improvements. A run that enters gameplay but freezes shortly afterwards is not a successful long-run test.
Settings to inspect
Close unnecessary heavy tabs and applications, connect power where appropriate, and check the browser’s own graphics-acceleration setting. If you change it, restart as the browser requests and record the before/after state. Do not disable security protections.
A lower render resolution can reduce graphics work, but the improvement depends on the bottleneck. WebGL2 and automatic rendering should be compared on the same machine once authorized hosting is available. No fixed improvement is promised.
What our old timings mean
Historical ready times include downloads, preparation, caching state and the particular network environment. Some probes intercepted requests or ran alongside other work. Those values cannot establish a fair renderer comparison, a minimum specification, cold-download throughput or an expected time for another visitor.
A warm OPFS run with zero downloader network bytes does not mean that the browser made no HTTP requests. Browser-check results report feature availability, not smooth play.