Automated generation of Manic Miner speedrun/walkthrough

[crem]

Does anyone know any tool to dump the contents of input frames of .rzx file?

I wanted to compare the keypress information (and actually take a look at the Danny's recording as both SPIN and Spectaculator are Windows only and I don't have Windows), but so far couldn't find a convenient tool for it.

Fuse cannot open that RZX file, the rzxdump from Fuse utils only dumps number and length of RZX frames but not the input data inside frames, and libspectrum doesn't have very convenient API.

[JianYang]

Are you maybe trying to play back recordings made with a Sinclair/Amstrad ROM against an OpenSE ROM? That leads to problems. Since I figured that out I have no problems replaying Danny's RZXs in FUSE.

 

[crem]

6 minutes ago, JianYang said:

Are you maybe trying to play back recordings made with a Sinclair/Amstrad ROM against an OpenSE ROM? That leads to problems. Since I figured that out I have no problems replaying Danny's RZXs in FUSE.

 

Thanks, it was indeed the case! For MM itself that has not been a problem, but the launcher seems to use more of ROM stuff.

[jetsetdanny]

I'm glad the FUSE problem has been solved 🙂. I've had no experience with FUSE (never used it). Unfortunately, I don't know how to dump the contents of input frames of an .rzx file, either 😮.

[crem]

I think I found bug in my code.

I use address $87F1 as start of a frame, while in reality the movement happens earlier in the game loop (at address $875C it calls Move WIlly (1) routine which calls Move Willy (2) which reads keyboard).

That means that my script skipped the very first frame of the level and started to move from the frame 2.

That also means, that even if you start moving one frame later in levels 1, 2, 3 and 6, you still will be able to get the maximum score.

[crem]

Sorry for posting this level out of order, but I think it's an improvement of the best known score by almost three seconds (interestingly, with the "only start moving from the second frame" bug, it was actually almost 3 seconds slower than the best known score, due to having to wait for the blue amoebatron in the very beginning).

Level 9 ("Wacky Amoebatrons")

Score: 1376

lvl9_20210312-0255.rzx

Edited March 12, 2021 by crem

[crem]

This time no surprises for Level 4.

Score: 1854

Hmmm now I wonder if I should regenerate Levels 1, 2, 3 and 6 after the bugfix.. Maybe there's something.

lvl4_20210312-0744.rzx

[IRF]

It won't make a difference in some caverns, where you have to wait around later on for a moment anyway - e.g. for the bottom penguin in the Cold Room to move out of the way of the portal.  But it may well tweak the score up in other caverns?

[MtM]

That route in Wacky Amoebatrons is a shocker! 😉 It looks on like on several occasions WIlly has made proper contact with baddie sprites to me, he changes colour due to proximity. Further, I suppose one huge advantage that the AI has is that it can make perfect jumps every time off the very end of ledges without having to worry about keypresses being correctly read.

[IRF]

If Willy comes under the 'shadow' of a guardian then he (or part of him) can change colour. He only gets killed if one of his pixels collides with a guardian pixel though.