← Back to Options DeskWeek plan, Oct 19-23 →
L3 MIXED
Find + Grade
Morning scan, combined list, A/B/C score card · as of Mon Oct 19 2026, 3:50 PM EDT
Implementing Files
FileLOCDescriptionStatus
scripts/beam_v5/morning_scan.py4025Discovery engine: wide net → options filter → NLP discovery → deep pull. Ran with the money lane 2:26-2:28 PM.built earlier
scripts/options/combine_lists.py279One graded table from the three sites, with a contract pick and PotterBox for A/B names.built today
data/options/graded/graded-2026-10-19-1432.md—Today's graded table: 5 A, 12 B, 249 C out of 266 liquid names.built today
data/beam_v5/conviction_scores.json—The old combined conviction score. Last written Jul 28; the scan does not write it now.not built yet
Data Flow
Inputs
→
Fidelity scansdirection from the money
→
tastytrade metricsIV rank, IV percentile, earnings date
→
Webull rankingsattention
→
Yahoo + PotterBox5-day move, box state
Outputs
→
graded-*.md / .jsonone card per name, every field written, one-line reason
→
A/B picksgo to the journal with a contract
Decisions + Tradeoffs
Grade rules
A = direction from 2+ money scans, IV rank ≤ 60, a tradable contract, no earnings before expiry. B = 1 scan, or expensive options, or earnings before expiry. C = everything else.
Flow can't tell a hedge from a bet
Put buying on a stock people own is often insurance. Every page says so and shows PotterBox vs pick as a second opinion.
Known Gaps
Combined conviction scoreNot rebuilt: one score per ticker from flow, IV, attention and catalyst.
C names have no contractSo there is no C control group yet; the proof test compares A with B and with zero.
Commands to Invoke
Grade today's listsuv run python scripts/options/combine_lists.py --data data/research/options-desk-2026-10-19 --out data/options/graded --budget 150
Ideas / TODO
Edit this textarea to capture ideas. Copy-paste the contents to an agent to generate the implementation.