TASVideos discussion board

[NesVideoAgent]

This is an automatically posted message for discussing submission: #1897: Swordless Link's N64 Super Mario 64 in 05:39.37

See details at: http://tasvideos.org/1897S.html

Submission status: published
Last edited by: Bisqwit

---

The submission message follows:

• Aims for fastest time
• Abuses programming errors
• Manipulates luck
• Takes damage to save time

The goal of this run was to beat the previous run by at least 2 seconds (60 frames). Thanks to a series of improvements I found, this was more than possible, and I actually managed to save more than 8 seconds in the process, which I never expected would happen initially. The exact number of frames saved over the previous run is 263, which is 8.8 seconds.

Comments…

This is the first run I've done without AKA. I found a rather large timesaver in the lobby, and told AKA about it, but he said he'd "let me claim this one", so I did it myself. It was quite a good experience, since it's the first full run I've done on my own, but overall, I preferred coauthoring this game over doing it myself.

Overall, I'm very happy with the outcome of this run. I think it looks a lot cleaner than the last one.

Improvements over the previous run…

Before the castle

A new strat was used here, which utilised jumping and diving. This saved one frame over just doing long jumps to the castle. I spent days trying to come up with a faster strategy, but after everything else I tried ended up being at least a frame slower, I came to the conclusion that what I had there was optimal.

Inside the castle

The first timesaver I found, as I mentioned above, was in the lobby. This was done by BLJing on the opposite set of stairs from what one would normally BLJ on (the ones outside Bob-omb Battlefield). After doing a series of quick BLJs, I did one large BLJ, then managed to fit in another quick BLJ before shooting off. When Mario left the stairs, I guided him into the porch, and managed to get him merged in the vertex, so he could turn around enough to make it to the stairs in the center of the room. If he hadn't gotten merged with the vertex, he wouldn't have been able to turn around enough to make it onto the stairs. I tested this extensively. I also tested doing something similar on the JRB stairs, which proved to be several frames slower. This new strategy saved 50 frames over the previous strategy, which is just over 1.5 seconds.

However, I then started thinking about beginning the BLJ from further away from the wall, just in case I'd be able to fit more BLJs in on the step, potentially preventing the need for that "slow" BLJ. I also thoroughly experimented with which step to BLJ on, and the second from the bottom turned out to be the fastest. After many hours of tedious experimentation, I finally got it to work, and this saved an additional 28 frames (just under 1 second), bringing the total number of frames saved at this stage up to 77, which is just over 2.5 seconds.

Bowser in the Dark World

The next timesaver I found was something which, after I did it, I was surprised that me and AKA hadn't done it in the last run. In Bowser in the Dark World, rather than killing some of the speed after BLJing by doing a normal jump, I kept Mario at full speed, and just took a slightly different path, utilising diving rather than jump kicking. This saved 28 frames, which is just under 1 second.

However, after redoing the section before BitDW, I decided to redo BitDW too, because I felt I could've made it more entertaining. So I did, and also managed to save another 4 frames, which brought the total number of frames saved in BitDW to 32, which is just over 1 second.

As for the first Bowser battle, it has been optimal for many runs already, so it's nothing new.

From BitDW to BitFS

After BitDW, I tested a lot of different move combinations, and concluded that a jump kick followed by a long jump was the fastest. It got Mario into the correct position for doing the BLJ quicker than any other combination of moves. This meant I was able to start BLJing earlier than in the previous run. 6 frames were saved in this section.

As for the basement, no frames were gained here. I tried for days to eliminate the need for the "slow" BLJ outside the 30 star door, but it just wasn't happening. The reason for this is that, if I did do another BLJ on the stairs, although I was actually able to do it, it was only possible through obscure analogue movements, which effectively zeroed Mario's speed anyway. And as was the case in the last run, a very precise speed was needed to bypass DDD, and I wouldn't have maintained that speed all the way from the stairs to the entrance to DDD, even if I had somehow accumulated it on the stairs. The buttslam is also still required, because it's the only move which cancels his speed suddenly enough for him to fall into the level. If you dive, he just shoots to the back of the area. So all I did differently here was change the camera angles while Mario was falling into BitFS.

Bowser in the Fire Sea

AKA suggested trying a series of dives at the start of the level, similar to what was done outside the castle. I tried this, and sadly, it ended up being slower than the traditional method, so I went with that. No frames were gained on the first tier of the level, but 4 were gained on the second tier and moving platform area, thanks to a new combination of moves, and curving Mario around the moving platforms, rather than doing a jump dive. (also suggested by AKA) At the end of BitFS, I did a slide kick into the lava, instead of the traditional jump dive. mr_roberts_z came up with this strat, and it saved 2 frames.

In the actual battle, I managed to save 1 frame, thanks to my positioning while swinging Bowser. I have no idea how we missed this in the previous run. So overall, 7 frames were saved in BitFS over the previous run.

From BitFS to BitS

While entering DDD to get back to the lobby of the castle, I managed to save 4 frames. The only explanation for this I can come up with is that in the previous run, we paused 4 frames too late without realising it. I have no idea how we managed that, but there's no other explanation for the frames I gained.

No frames were saved in either of the next two rooms, but in the room with the first stair BLJ, 2 frames were saved while BLJing. In the blue room, an additional frame was saved on the way to the star door at the top of the room.

In the room with the Endless Stairs, 5 frames were saved. This was thanks to not LJing onto the stairs, but rather doing a jump kick, because it put me within reach of performing the BLJ much quicker than the LJ did. This saved 10 frames on starting the BLJ, but in the end, only 5 were actually saved, since Mario was closer to the bottom of the stairs when he began BLJing, so he had more distance to travel before reaching BitS.

So overall, 12 frames were saved in this section of the run.

Bowser in the Sky

Initially, I thought that this part of the run was already optimal in the previous run. I did manage to get to the lifts at the end 3 frames quicker, but this gain meant that the lifts were in a bad position, and ultimately cost way more time than it saved. So I just did what was done in the previous run. As for the stuff after the lifts, it too was optimal. I tried some new strats, like stutter jumping after the second LJ, but that proved to be really slow. I used a slightly different strat right before entering the pipe, but it didn't save, or cost any time. I just did it because it looked cooler.

However, after I had submitted the run (¬_¬), mapler90210 found a new strat towards the end of the level which saved an additional 4 seconds, so I cancelled my run and implimented this into it before resubmitting. Sadly, it meant that the lifts were in much worse positions, but I still managed to get through them quickly enough. BLJing is possible on the lifts, but thanks to their positioning, it wasn't even possible without waiting for the next lift, so it was much slower than taking the traditional route.

As for the battle, again, no frames were saved, because it was already optimal. We spent hours doing that in the last run, making sure we hit the correct bombs on the correct frames, etc. All I changed was the stuff I did while waiting for Bowser's speech after killing him.

With thanks to…

AKA, for coauthoring several runs with me in the past, which gave me the experience necessary to create a run as optimal as this. He also introduced me to Memory Hacking Software, which allowed me to monitor Mario's speed, which helped a LOT with the BLJing. He also provided constructive criticism at every part of the run, and suggested several new methods of improvement, particularly in the last 2 Bowser levels. He also showed me how to hex edit towards the end of the run, which saved me from bugging mr_roberts_z or him when I needed some hexing done from that point on.

mapler90210, for discovering that route change towards the end of BitS. This saved a lot of time. I just wish he'd found it before I submitted the first version of this run. ;) But thanks a lot anyway.

mr_roberts_z, for helping me test a lot of stuff relating to the post-BitDW BLJ. He also provided constructive criticism for every part of the run, and consequently helped me improve it by telling me his ideas for improvements. He also did some hex editing for me, and stayed awake into the early hours of the morning on several occasions so that I didn't have to stop working on the run for that night. (in case he was able to find improvements in what I had done)

Rikku, for coming up with that strategy for getting to the castle. He found this while working on the new 120 star TAS with mr_roberts_z. It may only save 1 frame, but it's still something.

---

Bisqwit: Refreshing for forum topic update

---

adelikat: Accepting for publication as an improvement to the published movie.

---

Bisqwit: Processing…

[Soulrivers]

Perfect? I think so. :)

Yes vote.

[AKA]

Not that I'm saying the game is maxed out but I can't think of any additional improvements.

That's faggotry.

suggested movies

[Tompa]

Run is too good, voting no.

[Aglar]

I somehow felt there was a reason the previous run was canceled. Finding 4 more seconds of improvement is truly amazing.

[ZeXr0]

Avi or Youtube anyone :( pllzzz

[Aqfaq]

WFT? Almost 9 seconds?

[mr_roberts_z]

Well see the thing about Super Mario 64 is that it's just been done TOO many times and the cycle clearly should've ended at 6:42 but I guess this is a respectable improvement and although I'd hate to see Spezzafer's run get obsoleted I suppose it may be worth it so I will have to give this a Yes vote.

Yes vote.

[Mukki]

[Comicalflop]

8.8 yes votes.

[Raiscan]

I feel its necessary for me to abstain from voting this time. Not quite sure while I feel this to be required, but along my lines of reasoning are that its pretty much the same as the last submission which I voted yes for.

Too many mini-improvements :(

[DK64_MASTER]

Getting repetitive, but enough for me to vote yes.

I've said it before, and I'll say it again, I would really like to see a 16 star run with this level of optimization. Drawing boundaries on what glitches not to use might be a bit tough.

[superjupi]

I don't even know where you keep finding these frames. x_x
Voted yes on the last submission, and I don't see a reason to change my vote for this one.

[Dark Cloud]

Great improvement! Yes vote!

[hopper]

Same Yes vote as before. Even Yesser now.

[petrie911]

petrie911 wrote:

[Twelvepack]

Really nice. This run gets faster all the time!

[GeminiSaint]

Obvious yes.

[mmbossman]

Captain Obvious to the rescue with another yes vote.

[mapler90210]

if i had the ability to vote, i would vote yes. a great run, with extreme optimization as well as very entertaining. great job, SL!