Submission #6933: DaSmileKat's NES Extra Mario Bros. "bad ending" in 05:24.35

This run aims to access the bad ending in Extra Mario Bros as fast as possible. It improves #6774: Baddap1's NES Extra Mario Bros. "bad ending" in 05:26.78 by 146 frames (2.429 seconds). New routes and strategies were used in this run.

I think Extra Mario Bros is one of the best ROM hacks ever made, so I decided to TAS it. HappyLee is already working on an improved version of the good ending run, so I TASed a bad ending run. As I went on, I saw that Baddap’s run is littered with suboptimal movement and strategies, and most sections had been improved, both in speed and entertainment, at least in my opinion.

A main reason why Baddap’s run was rejected is because of the goal choice. I think that the bad ending should be considered beating the game, since the game states that this is the end. The game allowing you to continue playing after the bad ending is normal, since many games allows you to continue playing after reaching an ending. The invincibility after getting the good ending is just a bonus for getting the good ending. And the true final boss is a part of the different Area 5’s in the good and bad endings. Plus, the RTA community defines Any% as bad ending.

The other reason why Baddap’s run was rejected is the fact that it is very similar to the good ending run. That is not the case in this run. There is a completely new route on the way to the space jump powerup, resulting in differences including:

The deaths at the beginning is for gaming over later to warp back to the start of the game while still preserving the space jump powerup. The black screen after each death loads when Mario had passed a certain Y position, so the goal is to get Mario to that position as fast as possible. To do that, you need to die while having the Y subspeed be as low as possible and Mario’s Y position be as low as possible. By precisely manipulating the jump height to save 2 frames on each death, this run’s deaths saves 8 frames over Baddap.

As it turns out, it is faster to not grab the mushroom and not corner clip to activate the on/off block, but just barely. The framerule that this run gets is extremely tight, with 0 frames to spare, and it is probably required to save time with this route.

The second room is particularly precise, since the optimal acceleration at the beginning is only a subpixel away from losing a frame and the framerule. The fifth room is also a very difficult room, and possibly the room that I spent the most time on. Jumping while inside a wall instantly sets Mario’s X speed to 0, which is used here. L+r is used to give fast acceleration in both directions before the walljump, since otherwise, Mario won’t be far enough right to enter the pipe. Walljumping on this block is barely possible and only subpixels away from not being possible. This run exits Area 1 88 frames ahead of Baddap by not collecting the mushroom and other optimizations, but since Baddap’s run wasn’t fully optimized, it is only 59 frames faster than an optimal run using Baddap’s route.

I intentionally slowed down by 3 frames in the first room to manipulate RNG for the next room. In the second room, I found that a full jump from the lower section of the ground can clear the wide gap, and it is faster than using a cheep to bounce over it because of Small Mario’s smaller hitbox. The fastest way to get there is by jumping on the narrow strip of flat ground to the left of the pipe, since jumping from the high platform to the left of it can’t reach there. However, the spiky thing that falls in the middle of the pit can sometimes block the path, which is what the slowdown in the previous room was for.

I used a different strategy for clipping into the wall in the third room, saving some time. It involves approaching the wall from its corner, since that allows Mario to go a bit into the wall. It is faster to walljump in the fourth room as Small Mario, again because of his smaller hitbox. The reason why the space jump powerup is collected is because it is required to perform a wrong warp into Area 5 without collecting the otherwise required 5 keys. The fastest way to enter Area 5 from the space jump room is by getting to Area 2 to save progress, then dying to warp back to the start of Area 1, then getting to Area 5 from there. This run exits Area 2 still 88 frames ahead of Baddap, but it is probably possible to save 2 framerules in Area 4 over Baddap as Big Mario, so the Small Mario route only saves 17 frames.

The mushroom is collected for the Bowser fight at the end. It is faster to grab the first powerup instead of the powerup in the fourth room. Because I had unlocked Space Jump, it is now faster to go into the second exit of the first room since I no longer need to corner clip to get there. Arrow blocks gives 64 speed, which is much faster than the maximum running speed of 40, so the arrow blocks were used to speed up the fourth room.

By pushing Mario further right on the screen, the game thinks that entering the final pipe leads to Area 5 based on the screen’s position. However, the row of blocks in the middle of the room is normally not enough to push Mario far enough right. So Space Jump is used to clip slightly into the bottom of the row of blocks, which acts similarly to a bonk. In this run, I used the 16-frame window in which the screen doesn’t scroll after bonking and clipping into the left side of the row of blocks to get further right without doing the Space Jump-only clip many times, saving 42 frames over Baddap.

Because of the layout of Area 5 meaning that there is not much to optimize, that I collected the fireflower here for Bowser, and that there is a lot of enemies here, this part is very good for entertainment, and I believe I had done a good job at that. But that doesn’t mean that there aren’t any time improvements in Area 5. 2 frames were saved from Baddap's run waiting at the start of the first room. 4 frames were saved using a faster strategy for the fireflower grab. 1 frame was saved from better optimization in the third room. And finally, 7 frames were saved by doing the Bowser fight faster.

It seems like this Bowser always jump backwards initially no matter when I enter the boss room. So the strategy used here is to get behind Bowser as fast as possible, then spam fireballs and stay very close to him. Spamming fireballs from in front of Bowser is slower because of the very slow deceleration while facing to the right, and Bowser charges at you instead of running away. It is possible to use a single fireball to do double damage, but that is slower because a fireball like that spends a long time in the air, preventing rapid fireball spam.

arkiandruski: Judging.

arkiandruski: As it stands, I could probably reject this movie based on the poll results alone. I feel that would be unfair to the author, though; and there are issues I need to address. On top of that, I feel that the vote is skewed against this run based on the history that comes from the previous submission in this category. At first I was willing to wait for an eventual improved all keys run, but that is taking longer than I thought and I think it would be unfair to wait any longer.

First, whether or not this run completes the game. For this I'm ignoring infinite games and only looking at runs for games that have a defined end. More specifically, games with different endings based on completion.In the previous submission discussion, [3507] SNES Clock Tower by ruadath in 01:53.46 (as opposed to [3525] SNES Clock Tower "best ending" by ruadath in 11:53.77) and [2741] Genesis Sonic 3 & Knuckles by Aglar & marzojr in 26:53.06 (as opposed to [1387] Genesis Sonic 3 & Knuckles "Sonic, 100%" by marzojr in 55:48.80) were used as examples. In this thread [3168] NES Gimmick! by Aglar, Hotarubi & Samsara in 04:24.61 (as opposed to [1546] NES Gimmick! "100%" by Aglar & Hotarubi in 07:44.45) and [4293] PSX Oddworld: Abe's Oddysee by Samtastic in 12:59.96 (as opposed to [4310] PSX Oddworld: Abe's Oddysee "100%" by Samtastic in 1:04:16.17 and [2372] PSX Oddworld: Abe's Oddysee "maximum casualties" by Dooty in 40:00.53) were also mentioned. I personally feel one of the best examples of the movies I've found is [2186] Genesis Rolo to the Rescue by Aqfaq in 09:04.55. That has to be one of the most disappointing endings to a game I've ever seen.

After looking at the endings of all these games, I notice that the endings for all of them are still functionally similar to each other. You can see the good ending to Rolo here. Both Clock Tower endings trigger the credits, as do the Sonic and Gimmick endings. Both ending for Rolo lead to a game over. For Oddworld, you still get that same screen of text assessing your playstyle.

[3173] NES Ghosts 'n Goblins by Arc & Koh1fds in 08:07.55 is a different issue, since the bad end isn't actually functionally similar to a good end. Instead it is a setup for subsequent loops, and thus is more akin to Super Mario Brothers 2nd quest, or the Donkey Kong games, which also end after one playthrough. The ending in this submission is not setting up a new loop.

From everything I've seen, the games that apply to this situation the most are the Castlevania games like those on the GBA. A few of these games featured a normal ending that happened halfway through the game which served as a gate to the real end-game. Based on how the site treats them, even if these endings are still functionally similar, they still aren't treated as real endings by the site, and that's what purpose the end of the run here serves. It acts as a bottleneck checking whether the flags needed to continue were set.

The movie is borderline on whether it can be counted as an ending, but the precedent set by judgements so far say that it doesn't count, and I haven't seen good enough arguments that I feel comfortable breaking precedent.

However, there's another issue that was brought to my attention by another run in the workbench that I find much more compelling in relation to this runs fate, and that is how the site treats game hacks.

As it currently stands, the site requires runs of game hacks to qualify for the Moons tier in order to be published. In order for a branch to be accepted into Moons tier, it needs to differentiate itself from the main branch in a meaningful way. As applies to hacks, that means that the hack needs to be different enough from the base game to be interesting, and that the run shows off what makes the hack unique and worth publishing. If this means picking a goal that better shows off what's interesting in the hack rather than fastest completion, I believe that should be encouraged.

One of the most notable features of Extra Mario Bros is how it introduces Metroidvania and exploration elements into the Super Mario Bros engine. I would like a run of the hack to demonstrate those elements and mechanics to show the audience why they should even care about the hack in the first place.

In light of these reservations and due to low audience support, rejecting for not qualifying for Moons tier.