MX Simulator

Generally, the physics are pretty much spot on in mx simulator. The algorithm that determines when a rider has a fall off could definitely be improved upon. I'm talking specifically about fall offs after over jumping and casing, where the best approach is to pull the clutch and sit down currently. If you were to do that in real life, the result would probably be reminiscent of Doug Henry 1995, even if you weren't jumping down a couple of stories.

Right now it's a little frustrating that you pretty much have to do the exact opposite as in real life, where you'd want to stand up and gas it during the landing. See a couple of videos for reference:
Jason Anderson jumping the wall jump at washougal: https://www.youtube.com/watch?v=AKz_J5mOjv0
Glendale Practice clip, see wall jump before the finish line: https://www.youtube.com/shorts/IyNBzUSOLJk

This also affects me sometimes on on-off jumps, when you come up slightly short but still gas it to make the off, you'll sometimes get a legs off.

Generally, you should be more likely to have a crash when landing while sitting and off throttle, then landing while standing and with throttle applied.

It's more difficult than it looks to make a good test for it. Basically I want to be able to detect it early so I can detach the rider early for a natural looking fall, so one test condition is ride velocity relative to the bike. That cancels when you're sitting down since the rider and bike are essentially a single body in that case. So it also tests G force. Only problem with the G force test is some things that look innocent actually put a lot of Gs on the rider. So if you get too picky about the G test the rider will fall off on a little curb hit or something like that.

I think it is possible to do better but it's definitely not simple!

I already had a feeling it wasn't an easy fix

The rider velocity relative to the bike test definitely sounds like the main issue here, maybe adding a static penalty value/multiplier if the rider is sitting down would be a quick workaround for this?

Regarding the g-force test, does it only take into consideration the amount of g-force during a physics step or an average over a certain amount of time? Is the direction of that force taken into account as well?
Without having any data to prove my point, I feel like certain situations such as hitting a curb will result in a big, short spike in g-forces, while the forces are a bit more spread out during over jumping/casing. Example: Hitting a small curb would result in a peak of 100g for 0.01s and an average 5g for 0.1s, over jumping Laroccos Leap would result in a peak of 30g for 0.01s and an average 20g for 0.1s. Fall off would only occur for a sustained 15g for 0.1s. Again, without any data this is pure speculation.

It's the average over 4 steps (1/32 second). You're right that a longer average might help. I think I must have experimented with that when I wrote it though.

I guess I could always just hack it and raise the G limit when you haven't been in the air recently.