Where Did the 0.3 Go? Inside an AI Routine Review

Your gymnast walks off the beam, looks at the scoreboard, and sees 9.0. Coach said it should have been a 9.3 routine. So did Mom. So did your gymnast. So where did the 0.3 actually go?

That gap — the one between what the routine looked like and what it actually scored — is where most of the frustration in gymnastics lives. The score gives you the headline. It doesn’t give you the story. And without the story, next week’s practice is mostly a guess.

AI deduction analysis is built to answer exactly that question. Here’s what it actually surfaces.

Why a Final Score Alone Isn’t Useful

A meet score is a summary. It compresses 90 seconds of routine and a dozen decisions by a judge into one number. That number is officially what counts — we’re not arguing with the judge. But for training purposes, the summary is the wrong unit.

You can’t train on “9.0.” You can train on:

• “Your switch leap was at 175 degrees, not 180. That’s a 0.10 deduction every time you compete it.”
• “You took a small step on your dismount landing. That’s 0.05.”
• “Your toes weren’t pointed on your handstand mount. That’s another 0.05.”

That’s the level of resolution coaches give to elite athletes after watching footage frame by frame. Stickd’s AI gives that resolution to every gymnast, after every routine, in under a minute.

A Real Routine, Broken Down

Here’s a Level 7 beam routine that scored a 9.0 at a recent meet. The coach’s gut said 9.3. Stickd’s AI ran the video and surfaced exactly where the difference came from:

There’s the 0.3. Not in one big mistake — in five small, totally fixable ones. None of them were obvious to the naked eye. All of them are clear in 60fps pose data when you know what to look for.

What an AI Review Sees That a Phone Replay Can’t

Filming on your phone and watching it back at half speed is genuinely useful. We’re not knocking it. But here’s the gap:

• Joint-angle math. A switch leap at 178 degrees and one at 180 look identical to a parent. They’re a 0.10 difference to a judge. Pose detection measures the angle directly.
• Frame-precise landings. Was that step on the dismount, or just a knee bend? At 60fps with skeletal data, the answer isn’t a guess.
• Form breaks you stop noticing. Toes, knee separation, body line — your eye normalizes these after watching the same gymnast for years. The AI doesn’t.
• Apples-to-apples across meets. The same model scores the same routine the same way every time, so trends are real signal, not perception drift.

From Insight to Action

Identifying the 0.10 leap-angle deduction is half the answer. The other half is what to do about it — and that’s where Stickd’s drill recommendations come in. Each deduction maps to specific corrective exercises:

• Leap angle under 180° → active flexibility drills, pike straddles, split holds
• Form break on landings → landing position holds, proprioception drills
• Toe-point form breaks → foot conditioning, theraband sequences

By next meet, those five small deductions are five drill targets in this week’s practice. That’s the loop: identify, drill, re-record, see the difference. The 9.0 becomes a 9.15, then a 9.25, then the 9.3 the coach predicted in the first place.

The Bottom Line

A judge’s score is the truth on meet day. AI deduction analysis is the explanation that gets you ready for the next one. Both matter, and they don’t compete — one closes the meet, the other opens the week of training that follows.

For a parent-friendly walkthrough of the most common deductions and what they look like, read What Your Gymnast’s Coach Doesn’t Have Time to Tell You.

Common questions

Why did a routine that looked like a 9.3 score a 9.0?

The 0.3 gap usually isn't one big mistake — it's several small deductions (a form break, a short leap angle, a step on landing) that are invisible at full speed but visible frame by frame.

What can AI deduction analysis see that a phone replay can't?

Pose detection measures joint angles directly, so split-angle and body-position faults that are hard to see at full speed still get flagged, and it catches landing steps frame by frame.

Does an AI score replace the judge's score?

No — the judge's score on meet day is the official score. AI deduction analysis is a training tool that explains what to work on before the next meet.

What happens after a deduction is identified?

Each deduction maps to specific corrective drills, so a 0.10 leap-angle deduction becomes a concrete drill target for that week's practice, not just a note.