United StatesI've decided to go ahead and add the SNES rule about RNG. It doesn't affect existing runs, but it's good to have in case RNG manipulation is ever discovered.
I'm unsure whether to add a rule specifically against starting from a savestate. I agree it's not practical to enforce, since it would remove the convenience of resetting from the pause menu. But adding the rule would help prevent new runners from unintentionally starting from the same RNG sequence.
I believe it should be clarified in the rules too. I'm just not sure what the rule should be, and I don't feel like I should be deciding the rule by myself, especially since I seem to be the only non-Japanese runner of this game in recent years, and I'm not especially active.
I don't have a reason to doubt that specific run's legitimacy, but I worry that allowing starting from a savestate would encourage every run to start from a savestate for predictable RNG.
One option is: add a rule against starting from a savestate, but allow that run to be submitted again, since it's not a top ~5 time and it predates the rule?
And maybe also add the RNG manipulation rule from SNES Star Stacker ("Any video submissions that utilize RNG Manipulation MUST include the setup done before the run begins.") While GB Star Stacker doesn't currently have RNG manipulation, it seems like a good rule to have in case it's discovered in the future, and this would also prevent cropping a savestate load out of a video of an RNG-manip run.
I'd like the active runners' opinion on this.
In my experience, savestates (including the suspend points built into Nintendo Classics/Virtual Console) conventionally aren't allowed in speedruns, even if loaded before the timer starts.
Currently, this game doesn't specifically have a rule against it. Should this be added to the rules?
If starting a run from a savestate is allowed, it would probably become optimal to do so, to start from a fixed RNG seed.
Challenge mode has been added, now that there's another runner.
If there's interest in 2000 points I can add it too. How important is it to have the same categories between versions?
I've actually considered proposing a Challenge category in the past, but I wasn't sure of a good target score. I've done a test run to 1000 (threshold for unlocking Insane). I was also considering 1200 (last music change). Why 2000? That seems arbitrary.
I thought 9999 was too long, but if there's interest in 9999, sure.
Not sure where this should go. It's not really a "guide", but "resources" doesn't have a category for game-related data either...
I've had the level data extracted from the game for years. Finally decided to give the level headers a more readable format instead of raw hex data. I know no one is speedrunning this game right now, but might as well leave the link here?
The RNG weights in particular can be useful for speedrunning-- efficient stacking can be different with more hard blocks vs more friend blocks. Can't do much about those annoying double-star/double-hard pairs, but at least you can know the levels where they're common, and try not to stack too high so you don't die to bad RNG?
Spreadsheet includes:
- Dedede fist delay per level. 0 means no Dedede. This is temporarily set to 3 in crisis mode, if it's nonzero.
- Initial star count. Only relevant for Round Clear.
- Speed, 0-7. The exact effect of this is undocumented, but it generally increases block falling rate. Difficulty seems to be part of the formula-- setting speed to 7 in a Normal level won't change anything, and Time Attack Insane always has much faster falling blocks regardless of speed.
- Fixed pairs at the start of the level (each level defines 0-4)
- RNG weights, for falling pairs and Dedede. Note that Dedede will always spawn friend blocks in the corners, and in any column that does not contain a friend, even if friends have 0 weight.
Challenge mode overrides the header data, so it was hidden from the spreadsheet.
The weights for an unused/unfinished(?) "null" block (which acts like a star, but doesn't increase star count, and uses the graphic of the tile above Kirby's head), were also hidden since they're always 0.
(Full notes spreadsheet, if anyone's interested in the technical data. I was considering creating a ROM hack of this, but stopped because this game is obscure enough I didn't think anyone would play it.)
GBAs just include a GBC processor inside the case for backwards compatibility; they're completely different hardware.
If a GBA emulator (for example, mGBA) has GB/GBC support, it's actually just independently emulating GBC, so in my opinion it'd be considered a less accurate GBC emulator. Is there a general ruling on the site about that?
Given the very short run length of Round Clear Normal, and the number of ties, I'd like to propose using more precise timing. Sub-2:20 seems like a good cutoff.
