Something weird always happens with these things!
I think you may be overestimating what my script really did for me. Let me walk you through the process.
- Manually TAS to the first gameplay input
- Use TASEditor autofire to generate a baseline of up mashing
- Sprinkle in some left and right presses to avoid obstacles without slowing down
- Use my script to verify that there were no errors in steps 2-3
So if I'm 30 frames late or whatever to the first gameplay input, that's on me. And I know you probably don't trust me after the toenail incident, but I promise you I am not that sloppy.
Let me detail some of my test results.
FCEUX 2.3.0
I have deeply personal reasons for not wanting to update my FCEUX past this version. Go ahead, ask me if I'm joking.
If you look at my inputs, they are lag-perfect. 1007 is the first frame I can press up on the first level with my menuing. Your menuing only saves 1 frame.
BizHawk 2.3.1
The version of BizHawk I happened to have. I swear this is a coincidence...
I got the same 1007 start frame as I did on FCEUX 2.3.0 (not including your menu save). I tried your movie and it doesn't sync.
BizHawk 2.9.1
I get a start frame of 1011. I loaded up your userfile, and...
it didn't sync. Huh? I saw a brief pop up that says something about "ROM hashes," sounds like nerd stuff! Probably best to ignore it.
-----------------------------------------------
Nah jk, looks like you're using Rev A. I'm using the original release. I thought the ROM name was enough to clarify that (my other one is Action 52 (USA) (Rev A) (Unl).nes), but I probably should have put something in the version box. Sorry for the confusion.
Let me rerun my tests on Rev A to confirm that these emulators aren't completely broken.
FCEUX 2.3.0
Got a start frame of 989. Your menu thing saves 1 frame.
BizHawk 2.3.1
Got a start frame of 998. Your menu thing only saves 1 frame.
BizHawk 2.9.1
Got a start frame of 994. Your movie syncs, and indeed is 2 frames faster thanks to the menuing. I still get the hash warning, though...
-------------------------------------------
I didn't have time to test past the first stage on all of these, but you get the idea. Still quite a bit of variance even up to that point. Is that normal for most games? Either way, it does seem like Rev A is legitimately faster, and your menuing tech is a legitimate save regardless of emulator or game revision.
There might be a problem in considering the yellow and gray regions adjacent, considering the purple and pink regions at the same "corner" are equally as close! (The regions to which the four-color theorem apply shouldn't be considered touching at "corners".)
I dunno, don't they just look... touchy? This is kind of how I see it. But then again, if this isn't a corner, what is? How do you touch something less than 1 pixel?
By the way, my code is online now, which includes all the graphs I created.
Meong in 1:02.06 by warmCabin
