On-water rowing video analysis

Coaches have filmed from the launch for decades. The footage is useful, but it's hard to measure from: the boat is moving, the camera is too, the gunwale hides the rower's legs, and the oar handle arcs around the pin instead of running in a straight line. Software built for the erg makes things up in a boat. It scores ankles it can't see and flags a handle path that is really just the oar's geometry.

Mocta-X has a separate on-water mode that changes the grading to fit:

• Hidden joints are unmeasured, not guessed. Shins and ankles behind the hull get no weight in the score, so they can't drag it up or down.
• Handle path isn't graded, because a curved handle path is correct in a boat.
• The catch gate is relaxed for forward body angle, which reads differently from a moving camera.
• Tracking quality is judged on the trunk, since the legs are often out of sight.
• Feet don't have to be in frame to pass the capture-quality check.

Everything visible above the gunwale is graded on every stroke: body angle at the catch, arm extension, drive sequencing (legs → back → arms), layback and trunk swing at the finish, and the recovery sequence. You get a technique grade, a performance score for how well it held up, and an AI coach that links each point to the stroke and frame it came from.

The published checks and their weights are the same as on the erg. The rowing technique guide explains each one.

In a crew boat, the question is usually timing, not one rower's angles. Upload a side-on clip of the whole boat and Mocta-X extracts each rower's stroke phase, then scores synchronisation for every pair. You can see which seat is consistently early or late, and whether they're off by the same amount every stroke or all over the place.

That's the kind of thing a coach senses from the launch but struggles to prove. With a number per pair, the conversation after the outing starts from the same evidence.

1. Get level with the boat and square to it. Film from the side, not from behind or at an angle. Past about 20–25° off side-on, joint angles start to distort.
2. Match the boat's speed in the launch so the rower stays in the same part of the frame.
3. Zoom or move in until the rower (or the crew) fills most of the frame. A tiny figure on the horizon can't be measured.
4. Keep the sun behind the camera. Glare off the water and a backlit rower both hurt tracking.
5. Record at least five or six unbroken strokes at steady pressure. For crews, keep the whole boat in shot.

More detail, and the erg version, is in how to film rowing for video analysis.

It doesn't measure boat speed, run, check or oar force. Those need a GPS or instrumented oarlocks, not a camera. It won't score a clip it can't see well: fewer than half the frames usable, fewer than three gradeable strokes, or a failed capture-quality check, and you get a reason instead of a number. And like any single-camera system, it hasn't been validated against a marker-based lab. The methodology page goes through what has been measured.