Game objectives
- Emulator used: lsnes-rr2-beta23
- Aims for fastest time
- Forgoes arbitrary code execution
- Abuses programming errors
- Takes damage to save time
Time saved is displayed as '(a-b)', where 'a' is the amount of frames saved in the level, and 'b' is the total amount of frames saved so far.
Introduction to Ledges and Edges (110-110)
I improved acceleration in the slope in the beginning. I didn't go as fast as possible, though, because Mario needs to grab a throw block in the middle of the level. In order to grab a throw block without stopping, Mario needs to pass by a specific pixel. So I had to slow down in the slope to pass 1 pixel away from the mushroom, and I lost another pixel in the flying ? block, which set up a good throw block grab. Then, as suggested by J1h7e7, instead of using the springboard, I used a throw block jump. It was possible to avoid the chuck without a screen scroll, which saved a good amount of time. After avoiding the mushroom, I could keep full P-Meter, which allowed me to have higher speed in the last half of the level, and skip the grids entirely. I had to wait a bit in the end for the moving turn blocks because I didn't screen scroll, but getting past them was slightly improved. In this level, 110 frames were saved.
Mushroom Ledge (109-219)
I improved the turning around in the beginning, and I brought a shell into the pipe. This shell was used to shelljump, which skipped the majority of the intended ?-blocks maze. In my first run, I made Mario hit the P-Switch and threw it while it was flat to open a way out. However, as J1h7e7 demonstred, it was possible to skip the whole maze by jumping in the P-Switch in a lower and further position. In the last room, 2 frames were saved with the keyhole trick, 109 in total.
Overworld - A note about saving the game or not.
Sometimes the game asks if you want to save or to continue without saving. It is faster to select 'saving the game', because you don't need to press down for a frame before. However, it lags the game for a frame. If you spend a frame pressing down, and select 'continue without saving', then the game doesn't lag. So you might think it doesn't mater if you save or not, because you'll lose one frame anyway, but it does mater a lot, because lag frames are different real frames, and the game has many frame rules. Before the third level, I chose to save the game, so that I saved one real frame. There is a frame rule of interaction between fireballs and sprites, and by saving this real frame the fireballs could kill the chuck faster in the third level, so that Mario didn't need to decelerate that much, and 3 frames were saved there overall. In all the other save windows I chose to not save, because if it's impossible to beat the next big frame rule (which was always the star warp frame rule), it's better to reduce lag as much as possible before you get there, so that the real frame counter goes faster.
Chuckleton Plains (44-263)
Acceleration was slightly improved, and some corner boosts were made. The big improvement was to not damage boost near the chuck, and instead decelerate and kill him. It is faster because the 'getting hurt' animation is too long.
To avoid being pushed back by the chuck I fired him with 2 fireballs, so that he dies when Mario touches him. As stated previously, saving the game was important to kill the chuck faster in this bit. 44 frames saved.
The Untitled Sky Level Place (26-289)
I killed the goomba at the start with a fireball, so I could get P-Meter faster. A jump and an walljump were improved in the middle, and some corner boosts were included. 26 frames saved in total.
Springity Sproing (5-294)
Small improvement in acceleration, some corner boosts and a platform boost summed up to a 5 frames save.
Bastion in Blue (1549-1843)
Some frames were saved in the bits Mario had to clear way up by removing throw blocks. There is a pair of springboards in the beginning of the level, on a pipe. You are not supposed to be able to grab one of those, but it's possible to interact with one of them for one frame if you step the other; so, I grabbed one there. In this level there is also a secret exit: a pipe in the boss fight leads to the top of the first room, where you find a goal tape. I used that springboard in the first room to clip through the ceiling and touch the goal tape's hitbox, without needing to enter the boss. This saved about 26 seconds.
Baseball Mountain (164-2007)
I didn't stop in the water to scroll the screen, it was possible to avoid the pokeys without it. Avoiding a lag frame there without slowing down was really, really tough. A few subpixels were saved when Mario had to decelerate to wait a platform to sink or to avoid the pokeys, and some corner boosts were added. 164 frames saved.
Disco Dance Party (1400-3407)
I used a shelljump I got from a disco shell to skip the whole first room. In the second room, there were many small improvements when getting P-Meter, and I didn't get an unnecessary shell before a long jump. The big improvement, pointed out by J1h7e7, was to use a throw block to jump through the spikes, instead of disco shells. Reducing lag around the throw block, and saving subpixels to get it without needing to decelerate was pretty difficult. 23 seconds were saved here.
Chains of Atrophy (69-3476)
P-Meter obtaining was improved by decelerating for one frame in the ice, allowing Mario to stay in the ground for one extra frame. I clipped though the floor when getting the key. In my first submission I had picked the key ASAP, but it was actually better to pick it as later as possible. Despite you only get to accelerate later, it makes Mario to fall faster. Keeping flying speed after squeezing under the pipe saved a bit more than half a second. The keyhole trick also saved a couple frames, and having Y position as low as possible in the keyhole reduces lag. 69 were saved in total.
Disarray (1778-5244)
In the first room, I used walljumps instead of waiting for the moving platform, and I used the disco shell to go up faster. Killing the chuck and fishes, bouncing in the note blocks and walljumping were improved. To get to the clouds, I used a shelljump instead of the vine, and I screen scrolled to get past the platform in the first cycle. Killing the green koopas with a spin jump reduces lag. A couple more lag frames were saved after the update. Probably random factors such as real frame allowed this.
In the second room, the jump in the naked green koopa was slightly improved. Instead of getting a item box mushroom as the level asks in the info box, I entered in a reset pipe with big Mario and some invincibility time, because Mario had just taken damage. This time is just enough to go through the munchers - it finishes when Mario has clipped inside the last muncher, and what seems to be a new trick (at least I had never heard of it), Mario doesn't get hurt if he is inside a muncher in some certain positions. So the item box mushroom wasn't necessary, which allowed me to finish the level much earlier. In the updated movie, I saved a few frames in this level by slowing down before the muncher that damaged Mario, so that I took damage from it one frame later, but still entered the pipe in the same frame. This made Mario to have one extra frame of invincibility, which was very useful to go through the munchers faster. Lag and real frames were carefully traded of here to optimize as much as possible the star warp frame rule. 29 seconds were saved in the level.
Bowser's Castle (3758-9002)
In the first room, it is actually better to not get P-Meter and go with speed 37. I went for room 1 instead of 3 (maybe for room 3 it would be better to get P-Meter), and I took a platform boost. In room 1, there was some luck manipulation for the mechakoopas. The walljump below the turn blocks is really precise, it's hard to keep P-Meter there because I had to decelerate while waiting for the block's timer. In the update I entered room 1 with a different frame parity, which allowed me to save 1 extra frame. In the next room, a few frames were saved with more precise jumps. In the 4th room, I could keep take-of meter before hitting the cape block, so I didn't need to get P-Meter again there. A lot of time was saved with more precise flying, and by keeping flight in the whole room.
Room 6 is faster than room 5, and the way to it was improved. In the next room it was faster to go with 37 speed, getting boosts in the stair, instead of flying (one lag frame saved here in the update). In the 7th room, I flied below the level. There is a water tile in the first lighting (it was put there to hold the platform), and I used it after taking damage to recover flying status.
Room 8 was the hardest to optimize in this TAS. I improved the way to get P-Meter, and the way to the grids. The trick there is to make the camera to stop scrolling for a while, and then make it to scroll again, so that it will keep following Mario even without take-of meter. In the grids, it's possible to reach the camera ceiling, so I need to wait for it. After them, I get P-Meter again, and some jumps in this bit are very precise, even subpixel perfect. Choosing the platforms to land took a lot of testing, too. In the next room, I entered it with take-of meter, so I could start flying in the beginning, and it was possible to go up fast without decelerating. Some corner boosts were added here. In Bowser's fight, finally, I saved a couple frames before each Mechakoopa throw, saved some lags in the updated movie, and I ended input earlier (J1h7e7 was not aiming for short input, though). More than one minute was saved in this level: 62.6 seconds.
Major thanks to J1h7e7, for his run, and for pointing out improvements in 3 levels! His ideas with throw blocks saved a good amount of time!
Masterjun: Replaced file with a 122 frame improvement.
Masterjun: Let me get the easy and obvious things out of the way. The run itself is very optimized in SMW terms, and standing by itself it is also very entertaining. The entertainment usually comes from the fact that fast runs in the SMW engine are entertaining by default. But the levels in this run have especially difficult layouts and seeing them just rush by is really worth a watch. So in terms of technical quality and entertainment value, the run is on par with the other published SMW runs.
Now, since this is a hack, it has to be judged on its own. With this hack of SMW, which already has
three other hacks with publications, a new run has to add something to that current selection of runs.
Most of the levels being played (as well as the first overworld) feature the original SMW graphics and enemies. There is some new music, but the majority of the tracks in the levels is original SMW music. It doesn't help that this game doesn't seem to be too popular. But then in contrast, the final level really shows entertaining and new ideas. It suggests the variety this hack might have hidden in the unplayed levels. This seems to be confirmed by the posts in the discussion thread. However, a TAS that completes the game using the fastest route to bowser just doesn't show off anything we don't already have with the other hacks.
This is why my final decision is to reject this run. However, as the viewer feedback was quite good and the hack in general has some really good and new ideas for levels, I encourage to create a run with a different goal which results in showing off more of what the game has to offer.
Memory: Honestly a neat romhack, would love to see a 100% TAS, but this warps TAS can still be Accepted due to the new rules.
