Momentum Screen

Momentum Screen

When to use

Activate when a user wants to know whether a DSE stock qualifies as a momentum leader: "run the momentum checklist on GP", "does BEXIMCO pass the Minervini trend template?", "is this a Driehaus growth setup?", "screen this for relative strength". This is the momentum-leg companion to technical-analysis; feed its grade into signal-synthesizer for a final call.

What it does

Scores the 25-point momentum checklist (PRD-001 REQ-038) and groups it into the five weighted categories of REQ-032:

1. Trend (25%) — the 8-rule Minervini SEPA trend template: price above the 50/150/200-day MAs, the MAs stacked 50>150>200, the 200-day rising, price within 25% of its 52-week high and at least 30% above its 52-week low.
2. Momentum Power (25%) — RSI(14) in 40-70, MACD above signal, ROC positive and rising, ADX>25, MFI(14) in 20-80, Bollinger %B in 0.5-1.0.
3. Volume (20%) — OBV trending up, volume above its 20-day average, the Accumulation/Distribution line rising, Volume ROC positive.
4. Relative Performance (15%) — Driehaus growth: earnings acceleration, positive earnings surprise, institutional accumulation (relative volume), relative strength vs the market index.
5. Risk (15%) — ATR/price under 6%, within 8% of support, no major resistance within 10%.

Each criterion passes/fails with a beginner-friendly explanation. The category fractions are combined (TREND 25% / MOMENTUM POWER 25% / VOLUME 20% / RELATIVE PERFORMANCE 15% / RISK 15%) into a 0..1 momentum score and mapped to a Momentum Grade: A+ ≥0.9, A ≥0.8, B+ ≥0.7, B ≥0.6, C ≥0.5, D ≥0.4, F <0.4. Output is a Thinking Card (see suite README).

How to run

# input must follow the suite data contract and include at least 30 OHLCV bars
python3 scripts/screen.py --input data.json --pretty
cat data.json | python3 scripts/screen.py

Reads ohlcv (required, ≥30 bars; ≥200 for the full MA stack), optional fundamentals (eps_history for earnings acceleration, earnings_surprise) and optional market_index (relative strength). Returns score (0..1), confidence, rating (the grade), key_metrics (overall count plus criteria_met/total per category), reasoning (✓/✗/? per criterion) and flags.

Try it now with the shared fixture:

python3 scripts/screen.py --input ../_fixtures/sample_input.json --pretty

Interpreting output

• rating A+/A — a textbook momentum leader; B+/B — qualifying with caveats; C and below — momentum is incomplete or fading.
• key_metrics.overall_count (e.g. "18/25") is the raw count of passed criteria; the weighted score can differ because categories are weighted.
• Criteria that need missing data (e.g. no earnings_surprise, no market_index) are marked ?, not counted, and surfaced as a missing:<field> flag — never silently dropped.

Notes

Indicator math lives in scripts/indicators.py (pure Python, copied from the suite so this skill is self-contained). All 25 criteria, with beginner explanations and Minervini/Driehaus attribution, are in references/CHECKLIST.md. Output is educational analysis only, never financial advice.