BowlSense Learn

Founder Update: The Hard Parts Are Getting More Real

A candid BowlSense update on Speed Check precision, safer score data, first sensor rollout work, and the opt-in field test we are opening.

This is a founder update, but it is also an engineering update. The last stretch of BowlSense work has not been one clean feature launch.

It has been a lot of practical, high leverage work around the places where real users can lose trust: a number that looks precise but was marked on the wrong frame, a sensor that captures data but takes too long to transfer it, a score edit that risks breaking motion history, a first sensor install that needs clearer instructions, or a signup flow that gathers the wrong kind of testers. That is the work right now.

Make the product less fragile. Make the numbers easier to trust. Make the first hardware rollout feel like a serious field test instead of a science project in a box. Speed Check got a lot more serious Speed Check started as a simple free tool: upload a bowling video, mark a few moments, get ball speed and rev rate.

You can open Speed Check in the browser, read the measurement guide, or use the beginner practice loop if regular coaching is not available where you bowl. Useful, but not good enough. Bowling video is messy. People film from different angles. Frame rates are not always what the file claims. Screen recordings can report weird durations. A visible logo might only show up for a handful of frames.

If the tool asks a bowler to scroll back and forth between the video and a form, they will eventually mark the wrong thing. The latest Speed Check work changed that. PR 171 added the frame truth engine and protractor dial. The important part is not the dial itself, although the dial is a big user experience improvement. The important part is that Speed Check now treats video frames like evidence.

A frame step only counts when the browser actually presents a new frame. The rev workflow can use a visible ball mark across a short span instead of requiring a full revolution. The math moved into shared code so the browser tool, future mobile flows, and tests all use the same calculations. PR 172 turned the page into a video centered measurement dashboard.

The old flow made users bounce between the clip and the mark rows. The new flow keeps the video, landmarks, results, formulas, and rotation controls in view together. It also made release speed and average speed first class measurements, because those are different bowling questions. Release speed tells you what left your hand. Average speed tells you what got to the pins.

PR 173 cleaned up the stuff that only shows up when real people use a tool hard. Loaded videos should not reopen the file picker when you click on them. Zoomed videos need panning. The dial should be editable without stealing every pointer interaction from the video. That sounds small, but it is exactly the kind of friction that decides whether someone trusts a tool enough to use it twice.

We also added opt in anonymous Speed Check community results in PR 163. The word "opt in" matters. No video is uploaded. No user account is attached. The write path is proxied through our server with validation and rate limiting instead of exposing the database directly from the browser. What I want from this is not vanity data.

I want to learn how bowlers actually measure themselves when the tool is free, private by default, and simple enough to share. The goal is not to pretend Speed Check is magic. The goal is to make a real throw easier to measure, easier to sanity check, and easier to talk about with a coach, pro shop, or another bowler.