@VB Thanks, I see. By the way, would you like to upload the correct bios file to resources and write a guide about how to set Bizhawk? I learned it when I played MZM, but it's useful for other players who want to run this game in the future.

@sorp About part 1, I'm going to count the frames besides the lag just to make sure. About part 2, I forgot about the 60 fps video. Maybe getting data from one 60 fps video would work instead of taking the average of six 30 fps videos. We can look at more than one 60 fps or check it against some 30 fps videos to make sure there's no weird differences between runs, but I think just one would probably be good enough. About part 3, I think if we're taking the average we won't need exact video capture, so it can work with even bad camera recordings. The start and end of the loading times shouldn't be systematically early or late, so I think all those inaccuracies will cancel each other out. About part 4, it's only VBA-Next and original hardware that might have the damage-lag. So you should try that test for VBA-Next.

And something else I forgot... There are possible slower screen-transitions from Chapel1 to Chapel2 for the easier route, and the flickering transitions between Aqueduct and Clock Tower. So these loading times will have to be timed too. I suspect that the transition frames for entering are different depending on if it's the first time you enter the area or not, so we can't assume it's the same. Since Wii U VC is the standard we should do it for that, but for others the slower screen transitions only need to be timed if there are runs doing them.

Sorry for taking awhile. Here's the frame data for some of the TASes and the 60fps WiiU VC video. https://docs.google.com/spreadsheets/d/1LWuBBZI2n42-9yXrHN0UYlBLPQERUZq4_k1BEacYPeM

I took a base TAS and edited it for different emulators. So far VBA-RR v24 without bios, and the newest version of BizHawk (2.2.2) for mGBA and VBA Next cores, both with bios. I added and removed frames of input in the TAS during screen transitions so that the TAS would sync and be as fast as possible. This change in frames of input turned out to be exactly the same as the change in black frames, so I think we can safely consider black frames to be making up all the difference in loading times. I can upload the TASes if anyone wants them. I also downloaded the 60fps video sorp mentioned and counted the black frames of the video.

The totals in my results: VBA RR v24 (no bios) - 151 frames of black loading screen BizHawk 2.2.2 mGBA (yes bios) - 180 BizHawk 2.2.2 VBA Next (yes bios) - 184 WiiU VC - 166

By these results sorp's run should get +15 frames, which is 0.25 seconds. And interestingly, mGBA and VBA Next turned out to be very similar.

There's some differences between this data and what charleon had. It could be due to BizHawk updating, counting loading times differently, or something else, besides the more precise 60fps counting I used.

So if there aren't any objections we can use this for conversion. The other runs need a little more work though.