Dan Salvato
IE
Thank you very much for this analysis.Been testing replay capacity just for some reference and understanding as an end user.
The following tests were conducted with 8 minutes on the clock.
Heavy mashing with 2 Jiggly puffs holding down and mashing about 3-4 buttons each at around 2-3 times a second the replay ended with 2:43 left.
With no buttons being mashed but both the C-Stick and Control Stick being held down (used controller reset to have the same tilt the whole time) on both controllers, it ended with about 4:02 left.
Control stick and C-stick stuff registers a ton more input than buttons. High APM stuff like Fox and Falco shine shenanigans won't actually affect the replay size/length much. DI/SDI with both sticks as well as dashing and Crouch Cancel heavy playstyles will see increased replay sizes and shorter recording times.
This is accurate. It's basically because stick values are floating-point values which are 32 bits each, whereas button presses are only 1 bit each. The compression algorithm looks for 16-bit sets of zeroes, so stick values contribute much more to filling up those zeroes than buttons.
In the future, I can probably improve the compression by converting the stick integer values back to float instead of storing the floats in the replay file. However, this currently isn't a focus of mine, and I don't think file space will be too big of an issue.