[#674] Fix user scoring logic bugs

[realproject7]

Summary

• Fixed spam_label handling: spam_label=2 (verified) no longer zeroed out; only spam_label=1 disqualifies; 0/blank gets -3 penalty instead of instant zero
• Fixed quotient_score: used as-is on 0-1 scale, no longer divided by 100
• Re-ran scoring and updated output CSV

Changed files

• scripts/score-users.mjs — 3 logic fixes
• archive/dropcast-users-scored-20260330.csv — updated output

Verification

• project7 (FID 8106): Score 64, Tag Writer ✅
• 483 Writers, 5757 Readers, 1095 None
• Top scores in 65-73 range (reasonable)
• Summary posted on external #674

Test plan

• project7 scores 60+ with Writer tag
• spam_label=2 users not zeroed
• spam_label=0 users penalized but not zeroed
• quotient_score on correct 0-1 scale
• npm run build passes

Fixes #674

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
plotlink	Ignored		Mar 31, 2026 1:25pm

[project7-interns]

T2b APPROVE

Script fix (+8/-5) is correct:

1. spam_label=2 no longer zeroed — old code treated any non-zero/non-blank as spam. Now only spamLabel === 1 disqualifies. ✅
2. Graduated spam penalty — unknown (0/blank) gets -3 penalty, verified (2) gets +2 bonus, spam (1) already exited. Clean ternary. ✅
3. quotient_score fix — was divided by 100 when data is already 0-1 range. Removing /100 is correct. ✅

Archive CSV is the scored output — reasonable to include for reproducibility.

[project7-interns]

Verdict: APPROVE

Summary

The PR fixes the two scoring bugs described in issue #674 with a minimal, focused change to scripts/score-users.mjs, then re-runs the generated archive output. The updated logic now treats spam_label as intended and uses quotient_score on its native 0-1 scale.

Findings

• No blocking findings.

Decision

Approve. The code matches the requested bug fixes, keeps the change set focused, and preserves the existing scoring structure outside the corrected logic.