This TAS is part of the WIP 70star A Button Challenge TAS.
The first part of the TAS (until 00:17) was executed by fifdspence (https://www.youtube.com/channel/UCXub.... He managed to store 56 sliding speed and 13 speed for bswimming, which I think was quite difficult due to the spacing on the rock, though I don't know the details.
After that, Mario swims to the pillar, where he repeatedly reactivates his sliding speed and slides for 1 or 2 frames to increase it (   • JRB Hyper Speed Using Pillar Attempt  ). Once I've built up enough speed, I swim to the other side of the pillar, reactivate the stored speed with Z+CUp and punch, wait a bit, then dive to the platform and rollout to get the star.

This star was very challenging for me: the trick used to build up speed is very awkward, it can give you either 1f or 2f (better) of sliding, but whether you are able to set up for a 2f slide depends on spacing, and after each iteration of the trick the spacing is different, and especially when you have high speed it's quite difficult to get 2f slides, so it's probably not feasible to avoid all 1f slides, but there's probably at least a few frames of timesave here.
I end up just building up to 97.18 speed for the pillar, which is as low as I managed to take it. Saving 1 more speed cycle would mean you would have to do it with 96.5, which I couldn't manage even with ideal starting position and angle (I was about 10 units short vertically, so it MIGHT be possible).
When landing on the platform, I have to slide for about 10 frames so I can get far enough to roll out to the star. It's actually possible to dive from the pole one frame sooner, but that forces you to use a worse angle, so even then I couldn't save that one frame, but it might be doable with extremely optimized straining + sliding and a perfect initial setup (the initial position for the pillar speed reactivation was already tight, like a 23 unit margin of error).

This strategy has a fun story. Up until this week, we assumed that 100.x speed was strictly necessary to be able to get to the platform (   • JRB Blast to the Stone Pillar 0xA  ), which meant that we would need to use Hyperspeed Punching to achieve the minimum speed, since the sliding trick can't generate more than 100.0. But, once that assumption was brought into question, I decided to look into whether 100.0 was possible, and I was able to do it once I figured out that after the punch on the pillar, having neutral joystick input conserves speed better without reducing the amount you move (in the end, I was able to take the necessary speed down to 97.18, which cuts out a significant amount of speed building cycles). Then, fifdspence timed the strats and confirmed that using just the sliding trick would be faster than doing an HSP and then the sliding trick, because although the HSP builds up speed quicker, it makes you have to swim around to do anything after it, which is slower than just staying in the same place and repeatedly doing the sliding trick.

Note that the SPD value on the top left corner of the screen shows the sliding speed taken as the hypotenuse of the sliding velocity's components, rounded down to the nearest integer. I decided to put that there because it might make the speed building process more insightful.

Huge thanks for fifdspence for entrusting me with optimizing this star, it's the first time I've TASed anything for a project with such a high level of optimization, and I feel happy for being able to contribute to the 70 ABC TAS.