Absolute Accuracy testing

First things first. Over the years, we have learnt one important lesson: Validating jump height scores is not a trivial task. At least not, if want to take it seriously.

Sure, with smaller jumps, particularly when done on flat water, you can get a decent estimate off a video. However, it gets incrementally more complex as the waters get rougher and choppier, for there is not an even water level to measure off of, and using reference objects for visual measurement becomes difficult. Factor in lense distortions, video angle, moving camera perspective, dynamic rider and kite positions, and any meaningful "measuring" becomes impossible.

Having gone through various different processes over the years, we know what it takes. To make sure your product is providing accurate jump height scores, you gotta go deep. Simply setting up a camera next to the water and guesstimate off line length to validate your measurements won't cut it.

In our mission to tune WOO scores to best be in line with actual jump height, there's many things we've tried. This list includes simple optical processes, laser based techology, 3 camera setups for triangulation, and even high-precision GPS systems costing thousands of dollars. The final procedure that we use nowadays is a complex combination of them. Every step needs to be executed to perfection.

Location-wise, you want a spot that allows for a camera to be placed far away from the rider, yet has an undisturbed view of the action. You also want enough space to place reference objects. Flat water and consistent wind of course help too.

For your setup, you station a camera at the beach, at a given location and height. You got to make sure it’s set up rock solid. A tripod won’t do. Angle, focal length, focus all have to be fixed too to collect good data. In our case, the camera was placed more than 800 meters away from the action, resulting in a flat image with little distortion.

Next, you place reference objects for the camera to pick up, as well as in the jumping spot, and collect their respective heights and positions. The ground posts ("Post 1" & "Post 2") are set up at the exact same height as the camera, to level the image out perfectly.

Additional references are set up at a higher point behind the rider ("Top"), as well as next to the takeoff spot ("Launch Marker") for extra checks and balances. You also want to keep track of the dynamic rider position. For this and exact reference point placement, we are using a so-called RTK (Real-time kinematic) system, which is rather costly but yields milimeter-accurate GPS data.

All you gotta do now is send out riders and actually record the data. You collect high resolution video data on your camera (4k60), record WOO data through multiple sensor placed in the center of the boards (thousands of data points per second), all while keeping track of the rider position. As mentioned, the latter part is achieved through the use of the RTK system.

When collecting data, you want a good sample size that covers different maneauveours on various kites in all sorts of conditions. So, among other things, we perform straight jumps, sent jumps, woosts, various rotations, kite loops, one-foot and board-offs, and test with straight takeoffs and pre-pops. We test for different kite sizes and models, including LEI and Foil kites.

Having collected all the relevant data points, we can now apply math to validate the scores coming off the WOOs.

Based on the previous illustration of our setup (see "Setup & Location", includings formula for steps 5 and 6) and this program screenshot, the process can be explained as follows:

1. Create reference grid for video frames in illustrator program, based on the set up reference points. Horizontal lines corresponding to water level, camera height as determined by reference posts,

2. Find apex through WOO and video timestamps and add this frame as a layer to the program.

3. Draw a line from water level to WOO position at the exact center of the board and note the amount of pixels.

4. Take note of distance measurement to camera based on high-precision GPS data from RTK System.

5. Calculate the jump height by converting pixel count back to millimeters.

6. Apply height correction based on distance to camera to get the final height score.

We repeat this process for every jump and have our data set(s).

Even with such a level of sophistication, certain limitations apply. Accounting for things like optical inaccuracies, heat blur, ocean movement, reference object alignment, camera movement, and human imperfections such as when selecting board center, you end up with a range of uncertainty of around 30 centimeters.

This margin of error is good enough to validate that WOO scores correlate with actual jump height - and do so across the board, i.e. across different gear setups, maneuevers or tricks . This process for example confirmed that kicking your legs up in various ways mid-jump, commonly referred to as "WOOsting" in fact does not provide a significant score boost apart from the natural height gained from the move.

Testing for Absolute Accuracy is not trivial - it requires a sophisticated process involving a unique setup of location and visual references, extensive high-tech gear, a solid testing procedure, down to the level of pixel-counting. When serious about confirming scores to actual height, this is the lengths to which you have to go.

We do think such testing is important, which is we make the effort. We hope this behind the scenes could provide some insights into what we are doing to ensure the game is not only fair - WOOs measuring the same - but also grounded in reality.