Well, to make a long story short, I finished this TAS around a year ago and simply never submitted it. The initial RNG value I used for that wasn't what I used here and unfortunately using Suuper's program doesn't work here to get any seed I want since different things are happening on startup compared to the original game. So, coming back to it, instead of trying to think harder I decided to think smarter: I picked a random date and time and made adjustments as I went along. I ended up finishing it in a mere two days.
Apologies in advance, I've forgotten most of what I've implemented when compared to Soig, so I don't have exact numbers to compare to. However, much like my previous
warpless submission, this TAS will appear slower than the current publication, but it is actually faster when loading times are factored in. It's also worth noting that there is a sound bug in BizHawk when using Luigi on the overworld map. The wrong sound effects play. It's kinda funny, but should be fixed (maybe I should make a bug report...). Regardless I hope the TAS is enjoyed and I have a few notes below:
Version Difference
Soig's TAS was on v1.04 since that's what was out at the time. This version is faster for 8-Tank because the loading zone to reach the boss is closer to the tank, thus requiring less of the screen to scroll. However, v1.04 is overall slower compared to later releases because v1.05 changed how the Bowser fight works. While I don't fully understand it, the gist of it is you can hit Bowser Jr. twice at the start, and before Bowser comes down you can hit him one last time. If done quick enough, Bowser seems to forget that he's meant to fly back up to shoot more Bullet Bills and drop more Podoboos (I'm not falling for this "Lava Bubble" propaganda), and thus all but one phase of the first half of the fight is skipped.
Newer DS Specific Mechanics
Thanks to the new-ish Hammer Suit powerup there are new things to exploit. However, a lot of stuff from NSMB carries over here as well, so I recommend reading my previous
warpless submission for a brief breakdown on those. It's worth noting that Luigi cannot do Space Jumps, which is a shame. If he could I doubt Mario would even really be used in this TAS aside from 1-1 since he accelerates a bit faster (ignoring QSG).
Mid-Air QSG
In the original NSMB, Mario would go into Shell automatically and his speed would be capped at 2.75 pixels per frame (p/f). In this game, Mario goes into Shell manually with a down press from the user and Shell speed is always capped at running speed: 3 p/f. The way they implemented this is similar to how QSG works in NSMB. Because of this, you can do QSG in the air and get the faster acceleration from it. Unfortunately, due to Newer Team's implementation, this only works when facing right. Luckily, standard QSG on ledges still work, so doing QSG facing left is still possible albeit less useful.
Hammer Space Jump
If the animation of Mario throwing a hammer ends on the exact frame his y-velocity is equal to 0, he can do a jump in mid-air. Unlike Space Jumps in NSMB, this can be chained indefinitely until either A. Mario reaches the level height limit, B. Mario hits a ceiling, or C. Mario hasn't moved far enough away to unload one of the previous two hammers. This is unbelievably useful for vertical levels and is used throughout this TAS.
Semi-Unprofessional Level Breakdown Due To Lapse in Memory And Not Wanting To Double Check Things
- 1-1: Time saved from more optimal sliding on slopes
- 1-2: Time saved from faster Shell grab and entering pipe faster
- 1-Tower: I don't remember if any time was saved but I think there was a few frames saved
- 1-Cannon: Probably nothing
- 4-1: Some time saved from a more optimal pipe entry and playing as Mario allowed for a faster acceleration compared to Luigi
- 4-Tower: Time saved from Hammer Space Jumps
- 4-4: Time saved from skipping the bone closing, unfortunately it was 25f away from skipping fireworks so I don't think skipping fireworks is possible. I actually wanted to use Hammer Space Jumps here but turns out the funny people at Newer Team decided to make the ending platform that the Castle stands on be tied to that switch, so it literally softlocks you
- 4-GH: No time saved, pacelocked due to fireworks
- 4-Cannon: Probably nothing
- 8-1: Some time saved due to more optimal Hyperspeed at the start and not using the final pipe. Using the pipe is faster for in-game time but slower RTA
- 8-2: Probably nothing
- 8-Tower: Hammer Space Jump abuse skipped a cycle and saved a lot and I'm happy about it
- 8-3: Even with the bonk before the checkpoint, using Mario is faster due to Space Jumps and saved time
- 8-4: Probably nothing
- 8-Tank: I'm so sorry, both to you, the reader, and to my family for having the audacity to lose time here :(
- 8-5: Probably nothing
- 8-6: I think time was saved from better optimizing the pipes, but I don't recall
- 8-7: Pacelocked due to fireworks
- 8-8: Time saved from being fun and fresh
- 8-A: Time saved also genuinely what is going on in this level? I swear this level is somehow RNG dependent because I would change one thing in 8-8 to manipulate RNG and all of a sudden my inputs don't sync properly and I just crash and die. It's so bizarre, but perhaps if I was smart enough I'd look at the code Newer Team posted for the minecart (I am not smart enough)
- Bowser's Castle: Time saved from aforementioned version difference, but also using Hammer Space Jumps to skip a lot of waiting sections
Suggested Screenshot (Frame 37135)
And yeah.
eien86:
Newer Super Mario Bros DS. is a really well made and historied mod/hack for New Super Mario Bros DS. The mod itself has had many revisions, currently sitting at release 1.19. This movie, though, beats the mod's 1.05 release. Typically, choosing the latest version is always recommendable, but any version of a game
is acceptable as long as there is a proper justification for it. This movie has been in the making for a long time, so it makes sense the author picked a currently old version at the time and, since it is a newer version than the currently published movie (
[3622] DS Newer Super Mario Bros. DS "warps" by Soig in 26:04.37), I see no issue with the version selection.
Now, there are a series of differences between versions 1.04 and 1.05 that have a noticeable effect on the total running time of the game and lead to this submission being overall slower than the published movie. As per the author's explanations, this is mostly because non-gameplay loading times which overcome all of the actual gameplay improvements made. Although the the explanations make total sense and the author is actually an expert player of this game, I wanted to make a quick test of this claim. I made a
video comparison of the first 5 minutes of both movies, trimming out all loading times. The result is clear, that this movie is faster gameplay-wise than the published one.
In terms of execution, this movie looks very well optimized and beats the current
RTA WR by the same author by more than a minute.
Accepting to Standard, obsoleting 3622M
Also replacing the movie file with one that trims away 1300 blank frames at the end.
