Skip to content

fix: a moment your athlete is not in is not your athlete's moment - #38

Merged
ralyodio merged 1 commit into
masterfrom
fix/moments-must-contain-the-athlete
Aug 10, 2026
Merged

fix: a moment your athlete is not in is not your athlete's moment#38
ralyodio merged 1 commit into
masterfrom
fix/moments-must-contain-the-athlete

Conversation

@ralyodio

Copy link
Copy Markdown
Contributor

The reel was still mostly not your kid, and this is why.

The problem

Production, with everything from #34#36 working: 7 suggested moments, 5 of which contained no trace of the athlete. He is on screen from 218s; those moments were at 49s, 68s, 158s, 175s and 182s. Every one was a real scramble under the rim, and every one was somebody else's child.

activity_near_goal and high_motion read the whole scene and never look at the athlete. Between them they carry 0.25 of weight against a 0.35 threshold and a 0.575 denominator floor — enough to clear it unaided.

So the scorer was answering "was anything happening?" when the question is "was anything happening to him?". A highlight reel of the game is not what anyone came here for.

The fix

Once somebody has said which child is theirs, a window they are not in scores zero.

Scene signals still contribute — a scramble the athlete is in is exactly what to keep — they simply cannot carry a window alone. With no athlete identified, nothing changes: there is no one to be absent, and the denominator floor remains what stops scene motion from carrying a moment.

Measured on the production project

before  7 moments, 5 of them with no athlete in frame
after   2 moments
  248.0-261.0s  0.602  player_acceleration, high_motion, ball_approaching_player, toward_goal, activity_near_goal
  263.0-272.0s  0.501  player_ball_proximity

Fewer clips, and all of them him. Both score far above the 0.35 threshold, where before the fix nothing athlete-driven had ever cleared it.

5 new tests covering the gate, the athlete-present case, and the unchanged no-athlete path. 571 tests pass, 8 skipped. Lint and build clean.

🤖 Generated with Claude Code

Production, with everything else working: seven suggested moments, of which
five contained no trace of the child they were supposedly of. He was on screen
from 218s; the moments were at 49s, 68s, 158s, 175s and 182s. Every one was a
real scramble under the rim and every one was somebody else's kid.

Two of the seven signals — `activity_near_goal` and `high_motion` — read the
whole scene and never look at the athlete. Between them they carry 0.25 of
weight against a 0.35 threshold and a 0.575 denominator floor, which is enough
to clear it unaided. So the scorer was answering "was anything happening?" when
the question is "was anything happening to *him*?", and the product promise is
the second one. A highlight reel of the game is not the thing anyone came here
for.

Once somebody has said which child is theirs, a window they are not in scores
zero. Scene signals still contribute — a scramble the athlete is *in* is exactly
what to keep — they simply cannot carry a window alone. With no athlete
identified nothing changes: there is no one to be absent, and the denominator
floor remains what stops scene motion from carrying a moment.

Re-scored against the production project, read-only:

  before  7 moments, 5 of them with no athlete in frame
  after   2 moments
    248.0-261.0s 0.602 [player_acceleration, high_motion, ball_approaching_player, toward_goal, activity_near_goal]
    263.0-272.0s 0.501 [player_ball_proximity]

Fewer clips, and all of them him.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@ralyodio
ralyodio merged commit 6a6a6c9 into master Aug 10, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant