""" Do Kalshi and Poly find the SAME bets or DIFFERENT ones? Quantify overlap. poly_backfill rows carry BOTH kalshi_{side}_mid/ask AND poly_{side}_price, so each book's [0.05,0.12) multiplicative-de-vig strategy runs on identical fixtures. For each fixture (game_date,league,home,away) decide whether Kalshi fired a bet, whether Poly fired, and if both, whether they picked the SAME side. Buckets: Kalshi-only / Poly-only / Both-same-side / Both-different-side / neither Reports counts AND each bucket's WR/ROIturn, so we see if the UNIQUE bets (the ones only one book finds) are real edge -> i.e. does running both books add independent n. Two universes: (a) ALL backfill fixtures (Poly covers fewer leagues, so coverage gap shows up) (b) SHARED-PRICE fixtures only: both books quoted a price (fair overlap test) Read-only. Multiplicative de-vig (live default), both-sides dedupe to higher CLV. """ import csv, glob, os, math from datetime import datetime from collections import defaultdict DIR = r"E:\soccer\poly_backfill" BAND = (0.05, 0.12) _FMTS = ['%Y-%m-%d %H:%M:%S', '%m/%d/%Y %H:%M', '%m/%d/%Y %H:%M:%S', '%Y-%m-%d %H:%M'] def _pts(s): for f in _FMTS: try: return datetime.strptime(s, f) except (ValueError, TypeError): pass return None def _f(x): try: return float(x) except (TypeError, ValueError): return None def _result(rows): for r in reversed(rows): v = (r.get('ftr') or '').upper().strip() if v in ('H', 'D', 'A'): return v return None def fair_mult(o): r = [1.0/x for x in o]; s = sum(r); return [ri/s for ri in r] def load(): fixtures = defaultdict(list) for path in glob.glob(os.path.join(DIR, "*_polybackfill.csv")): with open(path, newline='', encoding='utf-8') as fh: for r in csv.DictReader(fh): fixtures[(r.get('game_date'), r.get('league'), r.get('home'), r.get('away'))].append(r) fx = [] for key, rows in fixtures.items(): gd, league, home, away = key if not gd: continue pre = sorted([(_pts(r['timestamp']), r) for r in rows if (r.get('status') or '').upper() == 'PRE' and _pts(r['timestamp'])], key=lambda t: t[0]) if not pre: continue ftr = _result(rows) if ftr is None: continue fx.append(dict(key=key, gd=gd, league=league, ftr=ftr, pre=[r for _, r in pre])) return fx def fire(fx, book): """Return (side, price_for_settlement, won, clv) of the book's bet, or None. book='kalshi' (price=ask) or 'poly' (price=mid).""" cands = [] have_price = False for side, sidx in (('home', 0), ('away', 2)): for r in fx['pre']: ph, pd_, pa = _f(r.get('pin_home')), _f(r.get('pin_draw')), _f(r.get('pin_away')) if None in (ph, pd_, pa) or not (ph > 1 and pd_ > 1 and pa > 1): continue if book == 'kalshi': mid = _f(r.get(f'kalshi_{side}_mid')); px = _f(r.get(f'kalshi_{side}_ask')) else: mid = _f(r.get(f'poly_{side}_price')); px = mid if mid is None or px is None or not (0 < mid < 1) or not (0 < px < 1): continue have_price = True pf = fair_mult((ph, pd_, pa))[sidx] clv = (pf - mid) / pf if pf else None if clv is None or not (BAND[0] <= clv < BAND[1]): continue won = (fx['ftr'] == 'H' and side == 'home') or (fx['ftr'] == 'A' and side == 'away') cands.append((clv, side, px, won)); break if not cands: return (None, have_price) clv, side, px, won = max(cands, key=lambda t: t[0]) return (dict(side=side, px=px, won=won, clv=clv), have_price) def roi(bets): n = len(bets) if not n: return "(0)" wins = sum(b['won'] for b in bets) net = sum(((1-b['px'])/b['px']) if b['won'] else -1.0 for b in bets) mu = sum(b['px'] for b in bets); var = sum(b['px']*(1-b['px']) for b in bets) z = (wins-mu)/math.sqrt(var) if var > 0 else 0.0 return f"n={n:3d} WR={wins/n*100:4.1f}% ROIturn={net/n*100:+6.1f}% z={z:+.2f}" def main(): fx = load() rows = [] for f in fx: k, k_has = fire(f, 'kalshi') p, p_has = fire(f, 'poly') rows.append(dict(f=f, k=k, p=p, k_has=k_has, p_has=p_has)) def report(universe, label): K = [r for r in universe if r['k']] P = [r for r in universe if r['p']] both = [r for r in universe if r['k'] and r['p']] konly = [r for r in universe if r['k'] and not r['p']] ponly = [r for r in universe if r['p'] and not r['k']] same = [r for r in both if r['k']['side'] == r['p']['side']] diff = [r for r in both if r['k']['side'] != r['p']['side']] union = [r for r in universe if r['k'] or r['p']] print(f"\n=== {label} (fixtures={len(universe)}) ===") print(f" Kalshi fired: {len(K):4d}") print(f" Poly fired: {len(P):4d}") print(f" UNION (either): {len(union):4d}") print(f" Both fired: {len(both):4d} " f"({100*len(both)/len(union):.0f}% of union)" if union else "") print(f" - same side: {len(same):4d}") print(f" - DIFFERENT side: {len(diff):4d}") print(f" Kalshi-only: {len(konly):4d}") print(f" Poly-only: {len(ponly):4d}") if K: jac = len(both)/len(union) if union else 0 print(f" Jaccard overlap (both/union): {jac:.2f}") print(f" Of Kalshi bets, also on Poly: {100*len(both)/len(K):.0f}%") if P: print(f" Of Poly bets, also on Kalshi: {100*len(both)/len(P):.0f}%") # do the UNIQUE bets carry edge? print(f"\n Performance by bucket:") print(f" Kalshi-only (Kalshi px): {roi([r['k'] for r in konly])}") print(f" Poly-only (Poly px): {roi([r['p'] for r in ponly])}") print(f" Both, Kalshi side==Poly: K {roi([r['k'] for r in same])}") print(f" Both, same: Poly side: P {roi([r['p'] for r in same])}") if diff: print(f" Both, DIFFERENT side: K {roi([r['k'] for r in diff])}") print(f" P {roi([r['p'] for r in diff])}") report(rows, "(a) ALL backfill fixtures") shared = [r for r in rows if r['k_has'] and r['p_has']] report(shared, "(b) SHARED-PRICE fixtures (both books quoted)") if __name__ == '__main__': main()