"""Independent recompute of the M&A ripple headline from the published CSVs ONLY.
No database, no Cymetica code. Reads flagship_events.csv + flagship_legs.csv.
"""
import csv, math
from collections import defaultdict

ev = list(csv.DictReader(open("flagship_events.csv")))
lg = list(csv.DictReader(open("flagship_legs.csv")))

# 1. Rebuild every basket return from raw leg prices, compare to the published field.
by_ev = defaultdict(list)
for r in lg:
    by_ev[(r["selector"], r["deal_id"])].append(r)
worst = 0.0
for e in ev:
    legs = by_ev[(e["selector"], e["deal_id"])]
    recomputed = sum(float(l["renormalized_weight"]) *
                     (float(l["exit_price"]) / float(l["entry_price"]) - 1.0)
                     for l in legs) * 100.0
    worst = max(worst, abs(recomputed - float(e["basket_return_pct"])))
print(f"1. basket returns rebuilt from raw prices — worst mismatch {worst:.6f} pp")

# 2. Excess = basket - SPY, recomputed from the published SPY prices.
worst = 0.0
for e in ev:
    spy = (float(e["spy_exit"]) / float(e["spy_entry"]) - 1.0) * 100.0
    worst = max(worst, abs((float(e["basket_return_pct"]) - spy) - float(e["excess_return_pct"])))
print(f"2. excess vs SPY recomputed from SPY prices — worst mismatch {worst:.6f} pp")

# 3. The headline numbers.
print("3. headline recompute:")
g = defaultdict(list)
for e in ev:
    g[(e["selector"], e["attention_bucket"])].append(float(e["excess_return_pct"]))
for k in sorted(g):
    xs = g[k]
    n = len(xs); m = sum(xs)/n
    sd = math.sqrt(sum((x-m)**2 for x in xs)/(n-1)) if n > 1 else 0.0
    t = m/(sd/math.sqrt(n)) if sd else 0.0
    hit = 100.0*sum(1 for x in xs if x > 0)/n
    print(f"   {k[0]:>8} {k[1]:>4}  n={n:3d}  avg={m:+6.2f}%  t={t:+5.2f}  hit={hit:5.1f}%")