Well, the good news for you is that the better players can't get very far past level 24 at all. The bad news for your proposal is that 21+ aren't actually interesting from an individual-level speedrun perspective: they're all identical to level 20 in every way, as far as I know. So I don't know the official rules, but probably the most sensible thing would be for the level 20 leaderboard to accept any 21-24 level times as submissions, too.

Got a pretty nice run last night:

My best yet. But it triggers the big-combo bug in the middle:

https://www.twitch.tv/dmwit/clip/DarkManlySoymilkAMPEnergy

I don't get any actual benefit out of the bug. Which category should I be submitting this to? On the one hand, I'd be kinda sad that my best run ever doesn't qualify for the glitchless category I was intending to be running... but on the other hand, perhaps I get to be only the second person to ever submit a glitched run, which could perhaps be its own kind of honor. ^_^

So anyway I think I can find some joy either way, but what should I do here?

Ah! I see: the concern is that somebody repeatedly tops out until they see a level they like.

That makes perfect sense now.

Well, the rule is clear at least. Currently it's not obvious to me how one could use a game over to manipulate the RNG in a sensible way, if they hadn't already manipulated the RNG at the start of the run anyway. I guess it's information-theoretically possible, but seems like it would be kinda hard for a human to do the necessary computations in time.

Topouts are manipulation, too, even if there's no reset?

I've added some randomly generated problems, and let me tell you, some of them are real stinkers. Here's a video from my stream tonight of me double-checking that they're solvable and complaining a lot:

I hope you enjoy my pain. =)

P.S. I've had a chance to test on Windows now, and it works great there, too.

Hey folks! I've been wanting a tool that let me practice specific maneuvers in Dr. Mario for a while, so I made one. I finally got around to documenting it enough that I think somebody other than me could potentially use it.

Here's a little teaser video of what it looks like:

If it seems like something you'd be interested in, check out https://github.com/dmwit/bokrifulse for a detailed description, a quick guide to trying it out, and detailed documentation on creating new problems to practice your maneuvers on.

It hasn't got very many interesting problems/maneuvers available in it for the moment -- just some easy ones I put together to create a tech demo -- but even so I've found it helpful for experimenting with timing and consistency! I plan to make more in the near future; and if you end up creating some for your own practice, I'd love to hear about them and include them in the repo.

I'd be especially interested to hear from somebody who has Windows; I tried to design it so that it works smoothly on any OS, but the proof is in the pudding... (Most likely failure mode: the default configuration has paths with / in them instead of . Maybe that works, but I don't actually know!)

Hello friends --

It's a bit sad that there's no strategy guides, even though I think there is a fair bit of transferable knowledge about how to do Dr. Mario fast. So I've started working on a video series that talks about some of the tricks I frequently use while playing. It doesn't cover luck manipulation or the famous glitch, and I'm not super great yet, but some of you might find it interesting anyway. How often do you get to peek inside another player's head? =)

https://www.youtube.com/playlist?list=PLn88YPpZrrQ4rsd1NjoCwZZoyTJJdvuLI

Lots of games have glitchy speedruns. I guess that's usually handled by having categories that allow all glitches and categories that only allow specific glitches. Is it worth starting a separate category for this glitch? It might be: I can certainly imagine some people having more fun competing if they have to play the actual game, and I can certainly imagine some people having more fun competing by finding the fastest way to glitch through to the end.

What a magnificent run, Mr. inari80000! You have certainly earned some bragging rights here, and I think I speak for everybody when I say I want to sit and listen to you brag. So don't leave us hanging! It must have taken ages to plan and execute this run. I have so many questions.

I've been wanting to understand the bug you exploit for some time now (ever since I saw the famous TAS that showed it off first, of course!). It seems to my amateur eye like you've learned a lot about it. How do you prevent it from locking up the machine? Do you know the precise conditions under which it triggers? Do you have a way to predict what extra stuff it will clear?

You clearly planned your pill placements throughout the whole run very carefully. You did an impressive job of routing. Even the non-glitched levels are very efficient, manage color efficiency well, and are obviously planned many many moves ahead. How did you construct your plan? How did you practice? Did you actually memorize the exact placement of thousands of pills in a row??

How long did this whole project take, from conception through planning, practice, and execution?

Kudos again, man. I really hope you drop by and reveal your juicy secrets.

And for what it's worth: this run appears to have learned quite a bit about how the glitch works compared to the TAS. In the TAS, once the glitch is triggered it typically clears about half the remaining viruses; inari80000 appears to have learned enough to reliably clear essentially all the remaining viruses with the glitch. That's quite an accomplishment in itself, and there are several other independent things about the video that make it impressive in my mind.

There's this TAS demonstrating it on an NES emulator:

I've also triggered the glitch myself in normal play. In my attempts, and in the discussion of the TAS I linked above, it seems to be common for the glitch to lock up the machine, but it definitely doesn't happen every time, so presumably with sufficient planning you could find ways to exploit the glitch that reliably didn't lock the machine.

So that's some evidence, at least, that the glitch exists and is exploitable on the NES. What evidence do you have that the glitch is not doable?

Looks like the recent 35:30 record by thiku_nico (congrats by the way! very smooth execution!) is also a 0-10 record by quite a wide margin. Or is there a rule against using the same run in both categories?