#!/usr/bin/env python3 """Bayesian Intrinsic Growth — probabilistic 3-5yr growth hypothesis with priors. Source framework: references/serenity/bayesian-intrinsic-growth.md. Maintains 5 growth hypotheses with prior probabilities, updates them from observable evidence (revenue history, margin trend, TAM CAGR, market-share direction, valuation percentile, FOMO signals), then compares the posterior-weighted intrinsic growth rate to the market-IMPLIED growth rate derived from current valuation. Hypotheses (3-5yr revenue CAGR): H1 STAGNANT 0-5% H2 MODERATE 5-15% H3 STRONG 15-25% H4 ACCELERATING 25-40% H5 EXPLOSIVE 40%+ Output: - posterior probabilities for each hypothesis - posterior-weighted intrinsic growth rate (expected CAGR) - market-implied growth (rough reverse-DCF from EV/Sales × terminal multiple assumption) - gap = intrinsic - implied (positive = market is underpricing growth) - FOMO score 0-100 (high = price moved more than fundamentals justify) Usage: compute_bayesian_growth.py raw-data.json --output bayesian_growth.json compute_bayesian_growth.py raw-data.json --tam-cagr 0.18 --recent-price-return-1y 0.85 """ from __future__ import annotations import argparse import json import math import os import sys from datetime import datetime, timezone HYPOTHESES = [ {"id": "H1", "label": "STAGNANT", "cagr_min": 0.00, "cagr_max": 0.05, "midpoint": 0.025}, {"id": "H2", "label": "MODERATE", "cagr_min": 0.05, "cagr_max": 0.15, "midpoint": 0.10}, {"id": "H3", "label": "STRONG", "cagr_min": 0.15, "cagr_max": 0.25, "midpoint": 0.20}, {"id": "H4", "label": "ACCELERATING", "cagr_min": 0.25, "cagr_max": 0.40, "midpoint": 0.325}, {"id": "H5", "label": "EXPLOSIVE", "cagr_min": 0.40, "cagr_max": 0.80, "midpoint": 0.55}, ] def neutral_prior(): """Flat prior across the 5 hypotheses, biased slightly toward MODERATE-STRONG.""" return {"H1": 0.10, "H2": 0.30, "H3": 0.30, "H4": 0.20, "H5": 0.10} def normalize(d: dict) -> dict: total = sum(d.values()) if total <= 0: return d return {k: v / total for k, v in d.items()} def likelihood_for_revenue_history(observed_cagr): """P(observed CAGR | hypothesis) — Gaussian-ish bump around hypothesis midpoint.""" if observed_cagr is None: return {h["id"]: 1.0 for h in HYPOTHESES} out = {} for h in HYPOTHESES: mid = h["midpoint"] sigma = 0.08 # tolerance band likelihood = math.exp(-((observed_cagr - mid) ** 2) / (2 * sigma * sigma)) out[h["id"]] = max(likelihood, 0.05) # floor return out def likelihood_for_tam_cagr(tam_cagr): """A high industry TAM CAGR boosts higher hypotheses.""" if tam_cagr is None: return {h["id"]: 1.0 for h in HYPOTHESES} out = {} for h in HYPOTHESES: if tam_cagr >= h["cagr_min"] - 0.05: out[h["id"]] = 1.0 + min((tam_cagr - h["midpoint"]) * 2.0, 0.5) else: out[h["id"]] = max(0.1, 1.0 - (h["midpoint"] - tam_cagr) * 3.0) return out def likelihood_for_margin_trend(margin_now, margin_3y_ago): """Margin expansion supports higher hypotheses (operating leverage).""" if margin_now is None or margin_3y_ago is None: return {h["id"]: 1.0 for h in HYPOTHESES} delta = margin_now - margin_3y_ago out = {} for h in HYPOTHESES: if h["midpoint"] > 0.15: # strong+ hypotheses out[h["id"]] = 1.0 + max(min(delta * 4.0, 0.5), -0.4) else: out[h["id"]] = 1.0 - delta * 2.0 out[h["id"]] = max(0.3, min(1.5, out[h["id"]])) return out def compute_market_implied_growth(inputs: dict, years: int = 5): """Rough reverse-DCF: what revenue CAGR justifies current EV/Sales × terminal multiple? Simple model: assume terminal EV/Sales = 4x (a generic "mature growth" exit multiple). Solve for CAGR such that current EV / future revenue ≈ terminal multiple. EV = revenue_now × (1+g)^years × terminal_multiple (1+g) = (EV / (revenue_now × terminal_multiple)) ^ (1/years) """ ev_sales = inputs.get("ev_sales") if ev_sales is None or ev_sales <= 0: return None terminal_multiple = 4.0 ratio = ev_sales / terminal_multiple if ratio <= 0: return None return ratio ** (1 / years) - 1 def compute_fomo_score(price_return_1y, eps_cagr_pct, revenue_cagr_pct): """FOMO = % by which price return exceeds fundamental growth, capped 0-100.""" if price_return_1y is None: return None fundamental = eps_cagr_pct or revenue_cagr_pct or 0.10 gap = price_return_1y - fundamental fomo_raw = max(0, gap) * 100 return min(100, int(round(fomo_raw))) def extract_inputs(raw: dict, args) -> dict: ticker = list(raw.keys())[0] if raw else "UNKNOWN" c = raw.get(ticker, {}) info = c.get("info", {}) or {} annual = c.get("annual", {}) or {} revenue_cagr_pct = info.get("revenueGrowth") eps_cagr_pct = info.get("earningsGrowth") # 3-5yr historical revenue CAGR from annual ann_rev = annual.get("revenue") or [] hist_cagr = None try: vals = [float(x.get("value", x) if isinstance(x, dict) else x) for x in ann_rev[-5:] if x is not None] if len(vals) >= 3 and vals[0] > 0: years = len(vals) - 1 hist_cagr = (vals[-1] / vals[0]) ** (1 / years) - 1 except (ValueError, TypeError, ZeroDivisionError): pass margin_now = info.get("operatingMargins") revenue = info.get("totalRevenue") ev = info.get("enterpriseValue") ev_sales = (ev / revenue) if (ev and revenue) else None return { "ticker": ticker, "revenue_cagr_recent": revenue_cagr_pct, "revenue_cagr_historical": hist_cagr, "eps_cagr_pct": eps_cagr_pct, "margin_now": margin_now, "tam_cagr": args.tam_cagr, "price_return_1y": args.recent_price_return_1y, "revenue": revenue, "ev": ev, "ev_sales": ev_sales, } def main(): parser = argparse.ArgumentParser(description="Bayesian intrinsic growth valuation") parser.add_argument("input", help="Path to raw-data.json from fetch_financials.py") parser.add_argument("--output", help="Output file path (default: stdout)") parser.add_argument("--tam-cagr", type=float, help="Industry TAM CAGR (decimal)") parser.add_argument("--recent-price-return-1y", type=float, help="1y price return for FOMO calc (decimal, e.g. 0.85 = +85 percent)") args = parser.parse_args() with open(args.input) as f: raw = json.load(f) inputs = extract_inputs(raw, args) prior = neutral_prior() posteriors = dict(prior) # Sequential Bayesian update update_steps = [] for name, likelihoods in [ ("revenue_history", likelihood_for_revenue_history(inputs["revenue_cagr_historical"])), ("recent_revenue", likelihood_for_revenue_history(inputs["revenue_cagr_recent"])), ("tam_cagr", likelihood_for_tam_cagr(inputs["tam_cagr"])), ]: posteriors = {h_id: posteriors[h_id] * likelihoods[h_id] for h_id in posteriors} posteriors = normalize(posteriors) update_steps.append({"step": name, "posteriors": dict(posteriors)}) intrinsic_cagr = sum(posteriors[h["id"]] * h["midpoint"] for h in HYPOTHESES) implied_growth = compute_market_implied_growth(inputs) gap = (intrinsic_cagr - implied_growth) if implied_growth is not None else None fomo = compute_fomo_score(inputs["price_return_1y"], inputs["eps_cagr_pct"], inputs["revenue_cagr_recent"]) verdict = "INSUFFICIENT_DATA" if implied_growth is not None: if gap > 0.05: verdict = "UNDERPRICED_GROWTH" # intrinsic > implied by 5pp+ elif gap < -0.05: verdict = "OVERPRICED_GROWTH" else: verdict = "FAIRLY_PRICED" result = { "ticker": inputs["ticker"], "retrieved_at": datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ"), "methodology": "Bayesian intrinsic growth: prior over 5 CAGR hypotheses → sequential update from revenue history, recent growth, and TAM. Posterior-weighted CAGR vs market-implied (reverse-DCF on EV/Sales). Source: references/serenity/bayesian-intrinsic-growth.md.", "inputs": inputs, "prior": prior, "update_steps": update_steps, "posterior": posteriors, "hypotheses": HYPOTHESES, "intrinsic_cagr": intrinsic_cagr, "intrinsic_cagr_pct_display": f"{intrinsic_cagr * 100:.1f}%", "market_implied_growth": implied_growth, "market_implied_growth_pct_display": f"{implied_growth * 100:.1f}%" if implied_growth is not None else None, "intrinsic_minus_implied": gap, "fomo_score_0_100": fomo, "verdict": verdict, "interpretation": { "UNDERPRICED_GROWTH": "Posterior intrinsic CAGR exceeds market-implied by 5pp+. Watch for re-rating catalyst.", "OVERPRICED_GROWTH": "Market pricing in more growth than evidence supports. Vulnerable to disappointment.", "FAIRLY_PRICED": "Intrinsic and implied growth align within 5pp. Edge requires variant perception.", "INSUFFICIENT_DATA": "Missing EV or revenue — cannot compute market-implied growth.", }[verdict], "notes": [], } if inputs["revenue_cagr_historical"] is None: result["notes"].append("No 3y+ revenue history — posterior dominated by recent growth + TAM.") if inputs["tam_cagr"] is None: result["notes"].append("No TAM CAGR — TAM update step uses flat likelihoods.") if fomo is not None and fomo > 50: result["notes"].append(f"FOMO score {fomo}/100 — price has run ahead of fundamentals.") out_json = json.dumps(result, indent=2) if args.output: os.makedirs(os.path.dirname(args.output) or ".", exist_ok=True) with open(args.output, "w") as f: f.write(out_json) else: print(out_json) sys.exit(0) if __name__ == "__main__": main()