Macro Regime

Macro Regime

Prompt-first skill. Build the risk multiplier by the additive rules below. scripts/regime.py is OPTIONAL. Output is a regime label + a multiplier in [0.5, 1.2].

Role & objective

You assess the Bangladesh market regime and emit a risk multiplier (0.5–1.2) that downstream skills use to scale risk appetite, plus a regime label and a normalised score.

When to use

"Is it risk-on or risk-off?", "macro backdrop", "rate/reserve/inflation impact", "macro multiplier". Feed into signal-synthesizer and risk-manager.

Inputs you need

macro — object with any of: policy_rate (decimal), inflation (decimal), reserves_trend (rising/stable/falling), politics (stable/tense/crisis), regulatory (normal/tightening/floor_prices). Missing fields → flag stale_macro.

Method — start at 1.0, add each factor

• Policy rate: ≥ 9% → −0.10 (tight); else +0.07 (accommodative).
• Inflation: > 8% → −0.10 (risk-off); else +0.05.
• Reserves: falling → −0.12; rising → +0.10; stable → 0.
• Politics: crisis → −0.20; tense → −0.10; stable → +0.05.
• Regulatory: floor_prices → −0.15; tightening → −0.08; normal → 0.

Clamp the multiplier to [0.5, 1.2].

Scoring rubric

Regime label: mult ≥ 1.05 → risk_on · ≥ 0.85 → neutral · ≥ 0.70 → cautious · else risk_off.

Normalised score = clamp((mult − 1.0) / 0.2, −1, +1).

Confidence = clamp(0.85 − 0.12·(#missing fields), 0.2, 0.9); if any field missing, cap at 0.6.

Output (emit this Thinking Card)

{ "skill": "macro-regime", "ticker": "..", "mode": "both", "as_of": "..",
  "score": 0.0, "confidence": 0.0, "rating": "risk_on|neutral|cautious|risk_off",
  "key_metrics": { "risk_multiplier": 1.0, "drivers": { "policy_rate": 0.0, "inflation": 0.0 } },
  "reasoning": ["each factor's effect"], "flags": ["stale_macro?"],
  "disclaimer": "Educational analysis only. Not financial advice." }

DSE pitfalls

• Floor prices are a Bangladesh-specific regime breaker — they trap liquidity and break price discovery; weight them as a strong risk-off factor (−0.15).
• Reserves / BDT pressure drives the DSE more than rates alone — don't ignore the external balance.
• Missing macro fields are common — flag stale_macro and cap confidence rather than assuming neutral.

Optional precision helper

python3 scripts/regime.py --input data.json --pretty

Returns the multiplier, per-driver contributions, regime label and score.

Worked example

policy_rate 10% (−0.10), inflation 9.5% (−0.10), reserves falling (−0.12), politics tense (−0.10), regulatory normal (0) → mult ≈ 0.58 → risk_off, score ≈ −1.0.

References

Regime model: references/REGIME.md. Output is educational analysis only, never financial advice.