Probably more akin to an IL than a full game category, but Practice Mode allows some things you can't do anywhere else in the game. So I figured let's max it out:

All 6 shapes 5 layers 99 cores

I've been calling it "The Centurion" because I want to :)

Here's an initial submission of 5:50.

Farming atoms appears to be the way to go.

I'm currently running on a gamepad (8bitdo SN-30+) but I have an arcade stick with a quad gate joystick and I've been debating switching to it because 1) it's easier on my thumb and 2) it reduces the number of inadvertent diagonal inputs... but I'm wondering if that would be considered an unfair (NES) advantage?

I should also note that at a certain point a save file tips over and starts showing the 2-minute-long "on-stage" cutscene where it used to show the bot-specific cutscene. This seemed to happen to my file after I used the all-levels unlock code and did the above test, so it might be related to seeing every bot-specific cutscene or otherwise clearing things with multiple bots. If it happens, you'll have to start a new save file for speedrunning non-ILs (which isn't really a problem because you can just unlock everything pretty easily).

Update update: Here's the times for the bot-specific cutscenes, from first black frame after 5-10 to last black frame before credits. + means I didn't account for the Elgato delay and stopped recording early, so they could be as much as .750s short, but they aren't the fastest so it's not worth re-recording. Rocket 54.684 Gear 43.550+ Gyro 51.150+ Stomp 35.300 Wheels 29.500 Turbine 54.083 Jak 28.217

So it looks like Jak is the winner. They might be off by a few frames but it's not so close that it matters. Meanwhile, I've been running with the absolute worst one...

Now, this adds an actual consideration for Hide and Seek... do you save more time using Rocket to find drills/towers quickly than the 26s you lose with the cutscene? (It turns out all of the speed bots have slow cutscenes...)

Update - FFR told me that the bot you choose determines the length of the cutscene after 5-10 (at least on certain revisions of the game), and he says Wheels is the fastest. I will check it out myself later.

I played around with the bots and for Rescue I found it 100% didn't matter to me because the speed and power only affect when you hold the D-Pad and I'm always tapping.

For Hide+Seek though, there are many games where you need to spin the ball to find things - Drill, Brick, the two Towers - so I'm using the bot with max speed for those. (With this bot, you can actually tap-drag faster than you can hold-drag!)

I've got a feeling it will less be "people" and more be "me"... but if I don't ask it definitely won't happen.

I think individual episodes make sense but individual stages would just be an RNG fest. Some of the early ones, in particular, are barely seconds long. I definitely know people who would like that though.

I've been submitting this to marathons and not having a ton of luck because it's long and a bit repetitive. MWSF accepted Rescue, but I'd really like to get Hide+Seek somewhere. So I'm wondering if we could get a shorter category, like Episodes 1-2, or some individual episode categories - ep. 3 would be great for a marathon.

Reposting from the discord. This is somewhat hand-in-hand with my comment about the WR being too dismissive of its own quality.

If you get 11 coins out of both coin blocks, you ultimately have to choose whether to get the 50th coin on the first life (after exiting the pipe), the second life (the third block where the two goombas are), or the third life (same).

It is possible to save a frame rule on WR getting the coin on the second life. Times are off a bit because I'm recording at 60 and not 59.94.

• Exit the pipe such that the blue sky appears at 14.366 or earlier (alternatively, the time should appear at 15.983) or earlier.
• Run straight into the goomba and die at 17.066 or earlier
• Start second life at 23.066
• Execute WR pacing through the 10-coin block, getting the 11th coin at 29.716
• Run right and get to the second of three coin blocks at 31.450
• Jump and get third coin block at 31.816
• Die to the goomba without stomping it at 32.000 (frame rule is at 32.066 so there's 4 frames of leeway)
• Start 3rd life at 38.083
• Repeat that execution on the 3rd life, as far as the 2nd coin block, and get 46.450

It is also possible to exit the pipe fast enough on the first life to get the coin, die to the Goomba and still catch the early frame rule (14.183 blue sky, 15.800 timer appears). This would add more leeway on the second life to catch the 32.066 framerule, but wouldn't be enough to save us another frame rule overall.

So you either need a ridiculously fast first life, or a ridiculously fast second life - your choice. And in either case you need to snag a coin and fall into a Goomba without stomping it, which I still find tricky.

Getting the coin on the third life is easier because you don't have to mess with the goomba stomping, and it's easier to catch the above framerules because you're doing less on the first two lives. But then you have to spend extra time on the third life, and you can't catch an EARLIER framerule. So this route is just going to be slower. Based on the WR and my own PBs, 46.816 is going to be close to what you can get this way.

Also note that the time from second to third block on life 3 is longer than the difference between dying to a goomba and getting a coin on the other lives, since it takes a few frames to get from the coin to the goomba in either case anyway, and on the 3rd life you end immediately at 50 coins.

I totally understand that reluctance.

I'm still doing more runs and more analysis. I did catch an early framerule on life 1 (the "god pipe" as I'm calling it) again yesterday (once in 150 tries) so I can say it's beatable, but not easily.

I'm also exploring another route that I thought would save one framerule over WR comparatively easily, but I haven't been able to make it pay out yet, so my math might be wrong there. I'll update later.

The SMB-only WR says "Bad underground and easily beatable" in the comment. Can we remove that? It's false because of frame rules on deaths. It seems POSSIBLE to catch an earlier frame rule there, but it requires a nigh perfect pipe - the kind I've only gotten once in a thousand tries.

Catching an earlier frame rule on the second life also seems like it might technically be possible (it's "only" 4 frames) but would require basically perfect play.

I don't want somebody reading that and seeing "free WR" when it's actually a very good time given the limitations of framerules.

I've got more analysis of this on the Discord, could paste it here if anyone wants.

After posting this, I wanted to check that the issue was "Emulators" and not just "Project64", and I can confirm that other emulators have similar lag. I tested PJ64, along with Mupen64Plus 2.0r in RetroArch (couldn't get the latest version figured out today because no frontend) and ParaLLEl 2.0 in RetroArch. The results are pretty similar.

loading 1.6.1 => 1.6.2 Real - 1.933s PJ64 - 4.233 Mupen - 4.200 ParaLLEl - 4.200

Cutscene Real - 17.334 PJ64 - 26.300 Mupen - 24.333 ParaLLEl - 25.533

By about 9-10 frames

Probably worth missing your last note if you've got an FC.

I don't think this qualifies as a "guide" but it's something I wanted to write up so people are aware.

I downloaded some VODs and did some frame counting and I noticed two major areas of slowdown for emulators:

1. The cutscene that plays between Hide and Seek episodes lags really noticeably in emulators, and it seems to cost about 6 seconds per episode. The other cutscenes don't appear to be appreciably slower.

2. The loading time between levels isn't as noticeable because nothing is happening, but it seems to be about double (!) in the emulator. This varies depending on stage complexity, but it goes from 1.500s to 3.000s for Rescue 1:1 => 1:2 and 2.400s => 5.000s for Rescue 10:9 => Rescue 10:10. If I take a guess that the average level lag is 2 seconds, and there are 100 level loads, that's 200 seconds, or about 3 minutes 20 seconds (!) lost on emulator over the course of the Rescue run.

The other thing I noticed is that FFRPro's Rescue PB plays the "Stage" cutscene (about 1m53s long) after Rescue 5:10, and both mine and RebeccaRE's plays the "Love" cutscene (only about 25s long). I'm curious if there's multiple versions of the game with different cutscene ordering, or something else.

I optimized the 63-piece route a bit more, to delay line clears and play the stack as high as possible as long as possible. Still started from MRM's but deviates a bit earlier, and man does it include some counterintuitive drops.

It's still possible that 65 pieces could be faster, because it might keep the stack higher longer, but I haven't routed that one.

I was learning MRM's Tetris route and came up with what I think are improvements (but my Tetris play isn't fast/consistent enough to know for sure). They're the same through the first 40 or so tetrominoes but there are some changes near the end.

MRM's route uses 66 tetrominoes and 8 line clear delays. This route gets a Tetris with the last I, getting it down to 65 tetrominoes and 7 line clear delays, and it doesn't substantially change the stack height. The only way this isn't a timesave is if this takes longer to position the pieces.

This route uses the optimal 63 tetrominoes (63 * 4 is 252, the smallest multiple of 4 >= 25 * 10) and 7 line clear delays. However, it's a lower stack for much of the later part of the run, so it may not actually be faster. My personal best is faster with this than the previous route, but agan I'm not consistent enough to tell for sure.

Sent you an invite on Discord.

Does All Time Trials require that the levels be done in order, starting with Championship 2 and ending with Half / Highway 3? Or can we choose the order?

This is coming out of a discussion on the Discord; moving Half earlier would make trying riskier double-bomb strats viable, but it also feels thematically appropriate to be the "boss" of the run.

Goal: Clear a time trial without eating any ghosts. Bombs are OK. This is surprisingly harder than it sounds.

Proof of concept on Championship 2 (embed is broken):

www.twitch.tv/videos/647287942

In my trials, this is possible on every level without dying to get more bombs, even Big Eater. (Whether it's faster on Big Eater to die or not is a separate question that I don't know the answer to.) Championship 1 is a royal pain though.

A single-segment run would also be possible but I'm not sure how many people would want to run it because it's really easy to invalidate the run.