""" WHY are Kalshi's non-Poly leagues so weak (-42% ROI, z-2.07, n=75)? Diagnostic. For the [0.05,0.12) mult-de-vig Kalshi strategy (bet at ask) on the poly_backfill set, break out per league and compare market-quality signals between the PROFITABLE Poly-covered leagues and the losing non-Poly ones: - WR vs implied(ask): is the loss a wrong-side / bad-fair problem (WR << implied) or just unlucky variance (WR ~ implied)? - pin vig (overround-1): is Pinnacle's fair noisier/wider in these leagues? - kalshi spread (ask-bid) and entry ask: worse execution / thinner books? - snapshots/fixture & distinct fixtures: data-density / staleness? - concentration: is the -42% one or two leagues, or broad? Read-only. """ 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 (gd, league, home, away), rows in fixtures.items(): 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 has_poly = any(_f(r.get('poly_home_price')) or _f(r.get('poly_away_price')) for _, r in pre) fx.append(dict(gd=gd, league=league, home=home, away=away, ftr=ftr, pre=[r for _, r in pre], npre=len(pre), has_poly=has_poly)) return fx def kalshi_bet(fx): """First in-band [0.05,0.12) Kalshi bet, both-sides dedupe to higher CLV. Returns dict w/ diagnostics or None.""" cands = [] 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 mid = _f(r.get(f'kalshi_{side}_mid')); ask = _f(r.get(f'kalshi_{side}_ask')) bid = _f(r.get(f'kalshi_{side}_bid')) if mid is None or ask is None or not (0 < mid < 1) or not (0 < ask < 1): continue 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') vig = (1/ph + 1/pd_ + 1/pa) - 1 spread = (ask - bid) if (bid is not None and bid > 0) else None cands.append(dict(clv=clv, ask=ask, won=won, vig=vig, spread=spread, side=side, pf=pf, mid=mid)); break if not cands: return None b = max(cands, key=lambda d: d['clv']) return b def agg(bets): n = len(bets) if not n: return None wins = sum(b['won'] for b in bets) net = sum(((1-b['ask'])/b['ask']) if b['won'] else -1.0 for b in bets) mu = sum(b['ask'] for b in bets); var = sum(b['ask']*(1-b['ask']) for b in bets) z = (wins-mu)/math.sqrt(var) if var > 0 else 0.0 sp = [b['spread'] for b in bets if b['spread'] is not None] return dict(n=n, wr=wins/n*100, iwr=mu/n*100, roi=net/n*100, z=z, vig=sum(b['vig'] for b in bets)/n*100, spread=(sum(sp)/len(sp)*100 if sp else float('nan')), ask=sum(b['ask'] for b in bets)/n, clv=sum(b['clv'] for b in bets)/n*100, nobid=100*sum(1 for b in bets if b['spread'] is None)/n) def main(): fx = load() bets = [] for f in fx: b = kalshi_bet(f) if b: b['league'] = f['league']; b['has_poly'] = f['has_poly'] b['npre'] = f['npre']; bets.append(b) poly_lg = {b['league'] for f in fx for b in [None] if False} # placeholder # leagues that ever carried a poly price poly_leagues = {f['league'] for f in fx if f['has_poly']} nonpoly = [b for b in bets if b['league'] not in poly_leagues] inpoly = [b for b in bets if b['league'] in poly_leagues] print("AGGREGATE comparison (Kalshi [0.05,0.12) at ask):") print(f" {'group':22s} {'n':>4} {'WR%':>5} {'impl%':>6} {'ROI%':>7} {'z':>6} " f"{'pinVig%':>8} {'kSprd%':>7} {'ask':>5} {'CLV%':>5} {'npre':>5}") for tag, grp in [('Poly-covered leagues', inpoly), ('NON-Poly leagues', nonpoly)]: a = agg(grp) npre = sum(b['npre'] for b in grp)/len(grp) if grp else 0 print(f" {tag:22s} {a['n']:4d} {a['wr']:5.1f} {a['iwr']:6.1f} {a['roi']:+7.1f} " f"{a['z']:+6.2f} {a['vig']:8.2f} {a['spread']:7.2f} {a['ask']:5.2f} " f"{a['clv']:5.1f} {npre:5.1f}") print("\nPer NON-Poly league:") print(f" {'lg':6s} {'n':>3} {'WR%':>5} {'impl%':>6} {'ROI%':>7} {'z':>6} " f"{'pinVig%':>8} {'kSprd%':>7} {'ask':>5} {'nobid%':>6}") byl = defaultdict(list) for b in nonpoly: byl[b['league']].append(b) for lg in sorted(byl, key=lambda l: agg(byl[l])['roi']): a = agg(byl[lg]) print(f" {lg:6s} {a['n']:3d} {a['wr']:5.1f} {a['iwr']:6.1f} {a['roi']:+7.1f} " f"{a['z']:+6.2f} {a['vig']:8.2f} {a['spread']:7.2f} {a['ask']:5.2f} {a['nobid']:6.1f}") print(f"\n non-Poly leagues present: {sorted(byl)}") print(f" Poly-covered leagues: {sorted(poly_leagues)}") if __name__ == '__main__': main()