Mound Metrics AI
Guide

How to Analyze a Pitcher's Delivery from Video (Without Guessing)

You've got a video of a pitching delivery on your phone. Maybe it's your kid, maybe it's a player you're coaching. You've watched it ten times and you have a feeling something's a little off — but "a feeling" isn't much to work with.

Here's how to actually break the delivery down frame by frame, right in your browser, in a couple minutes. No account, no upload to some server, no app to install.

What you actually need

That's genuinely it.

Step 1: Get the camera angle right

This is the part people rush, and it's the part that matters most. Garbage angle in, garbage analysis out.

Step 2: Drop in the video

Upload the clip, plug in the pitcher's height and throwing hand (used to scale everything correctly), and pick an age range if you want — mechanics look pretty different on a 10-year-old versus an 18-year-old, so this helps put things in context.

Worth knowing: the video never leaves your device. It's all processed locally in the browser, so you're not uploading footage of a kid to some random server somewhere.

Step 3: Hit Auto-detect

Tap it, and the pose-tracking picks out four moments in the delivery on its own:

  1. Reference stance — a calm, upright frame before leg lift starts. This sets the scale for everything else.
  2. Peak leg lift — front knee at its highest point.
  3. Foot plant — the frame the front foot touches down.
  4. Release — the exact frame the ball leaves the hand.

Step 4: Actually look at what it picked

Don't skip this. The tracking is good, but it's reading one camera angle, and it can occasionally lose track of a joint — the throwing arm at release is the usual suspect, since it can get partially hidden behind the body. Scroll through the marked frames and eyeball whether the dots are actually sitting where the joints are. If one looks wrong, re-run Auto-detect or nudge it manually.

Step 5: Read the report — with the right expectations

Once those four moments are locked in, you get a report comparing the joint angles and ratios to general ranges pulled from pitching biomechanics research. Tap any number and it jumps you right to that spot in the video.

A few honest caveats here, because they matter:

Step 6: Save it and check back later

You can save the analysis under the athlete's name, and everything — video, marked points, all of it — stays right there in your browser for next time. No re-marking from scratch. That makes it easy to film again in a few weeks and actually see whether specific numbers are moving, instead of just guessing whether things look "better."

The honest bottom line

The report is only as good as the video you feed it. A steady, well-lit, slow-motion side shot will give you a report you can trust. A shaky clip from across the field, less so. If a report ever looks off, check the footage before you second-guess the pitcher.

Mound Metrics AI is a free tool for frame-by-frame pitching delivery analysis, and it runs entirely in your browser — nothing gets uploaded. Try it here, and take a look at the medical disclaimer before you dig in.

Once you've got a report One specific thing worth checking: late arm timing at foot plant Why "late arm" is one of the most common flaws coaches flag — and how to spot it on the exact frame the app already marks for you. Another one worth checking A collapsing front knee after foot plant A quiet power leak that's nearly invisible at full speed — why a front knee that keeps bending after landing costs velocity, and how to catch it frame by frame. ← Back to the app