Position/velocity differences that depend on the global direction you're moving towards have likely something to do with rounding or limited number precision inside the engine. Some calculation might round away from zero when moving one way, and towards zero otherwise.
This brings to mind Quake... Quake 3 performs very aggressive truncation of player velocity to keep nineties network connections happy, which means you'll accelerate faster when moving at certain angles. There might be something similar happening here, although with modern physics engines all data types are kept very precise so any anomalies are absolutely tiny.
or could be because in some cases, like collision detection for example, you have to make arbitrary choice which axis goes first (x most of the time) so it then can lead to different behavior in some edge cases
@@gigigigiotto1673 Trackmania physics are deterministic - there's most likely at least some fixed point math. And if there are rounding anomalies, they won't care about floating point precision.
Very cool to see this now as we talked about that kinda stuff a couple of days ago in the trial & rpg discord. There are actually a couple of tricks that only work in one direction of the stadium, which I noticed while building kacky maps for kk7. Very interesting stuff!
As small as this is, it's so cool to see how far TAS is bringing us. The amount of new mechanics and tricks that no one would think to be possible even a year earlier, is incredible.
damn, sick find! question though, why are you so sure that it doesnt work in the other directions? any input set will have an altered result if you play it on the same blocks rebuilt in another direction, but it doesnt mean there isnt a different input set to do exactly the same thing in this case since the inputs to make your start trick work are so precise it's no wonder it dies since it was so hard to find it for -x direction, is it possible that inputs to do the trick in other directions actually exist but you didn't find them yet? don't know how long you tested that for^^ have you tried doing it on the right? for -x and eventually for the other directions also, does it still work if you keep the same direction but change block's height? and/or use air variation of start block instead of the ground one? aaand what happens if you remove the booster? will it still have a (smaller ofc) lead over the normal start trick or not? that would clarify if the trick works because the car for some reason takes the booster earlier than normal, or if its only about optimal grip on start ramp and the booster only magnifies the gain like on normal start tricks whoops i wrote an essay
The mirrored inputs do work yes. It's totally possible that the bug is possible in other directions with altered inputs. I couldn't know otherwise. Yes it does work at diff heights. It works on a04 for example
What I do is I go straight forward and do a double tap to both sides right as I reach the turbo so I minimize the turning diagonally distance and the wheel reaches it first, it works better and faster for me
That's how some tassers do their runs, like xeqzion and andi on fs maps, they just give the tas tool a distance, and check whatever inputs they can do to get the most amount of speed, they do it for little sections for the entire run and get an outcome, its sick actually
Hahahaha i found this in Tm Turbo just a month ago, a echelon boosting map was just a straight line with boosters and i found out that sligltly steering gave me a boost, i was gonna post it but ultimately i didnt