How it works
No maths, promise. Here is what Bolts is looking at when you send it a clip, and what it does with what it finds.
It reads every frame
A trick lasts about half a second. Filmed at 240 frames a second, that is around a hundred and twenty pictures, and Bolts looks at all of them.
Your eye gets maybe a dozen useful ones out of that. The gap between those two numbers is the whole product: the frame your tail hits, the frame your feet find the board again — all of it is sitting right there in the footage, if something is looking fast enough.
It finds the six moments
Every flip trick has the same shape underneath, whatever the trick is. Bolts places each of these to the frame.
We did not make the six up. In 2024 a group of researchers solved the ollie as a physics problem and came out with eleven timing cues; these are the six a camera can actually see.
- crouch 0.00s
- tail strike 0.21s
- takeoff 0.26s
- peak 0.43s
- catch 0.54s
- land 0.63s
The real times from the clip on the front page. 3 of the six are scored against a hand-marked clip; the others are placed from those, and are drawn differently wherever they appear.
It describes the trick instead of naming it
Most things that name a trick pick from a list, so anything slightly unusual comes back wrong and you learn nothing. Bolts measures the parts instead.
- The flip
- How many times the board rolled over, and whether it went kick or heel.
- The board spin
- How far the board turned, in degrees. A shuvit is 180. A big spin is 360.
- Your spin
- How far you turned, which is what separates a 180 from a plain flip.
- Your stance
- Whether you rode at it normal, switch, nollie or fakie.
Put those four together and the name falls out. One flip and nothing else is a kickflip. One flip with 180 of board spin and 180 of you is a varial. Land something with no name and you still get a straight answer about what the board did, rather than the closest wrong guess from a list.
Then it reads your body against it
Knowing what the board did is only half of it. The coaching comes from what you were doing while it happened.
How high you got
Board peak and body peak, separately. The research is clear that the board peaks before you do, and how far apart those two peaks are says a lot about whether you are jumping or just lifting your feet.
How deep you loaded
Your knee and hip angles at the bottom of the crouch. Front-knee angle is the single strongest predictor of ollie height in the published kinematic work.
Where your weight was
Hips against board centre, shoulders against board line, through the whole flight. Almost every failure a coach names comes back to one of those two drifting.
How you landed
Whether your feet came down over the bolts, together, with the knees bent. Landing stiff is how people get hurt, and it is visible in a single frame.
How you film it is part of the measurement
The top of a jump happens between two frames. On an ordinary 30-a-second video that moment is not in the footage, so nothing can find it — not Bolts, not anything. Film the same trick in slow motion and it is there.
Framing matters as much. Stand too far back and the board is a few pixels across; get down low and side-on and it is big enough to measure a rotation from. That is why Bolts checks the shot before you push rather than letting you find out afterwards.
Your clips stay yours
Footage you send is yours. You choose separately whether it can be used to make Bolts better, that choice is recorded against each clip, and you can change your mind later — at which point the clip drops out of that pile for good.
Point it at your own footage
One trick is enough to see what it does.