"""Separate real market makers from patient directional traders, then rank them. (read-only; touches no account and places nothing) THE PROBLEM THIS SOLVES. `maker_share` only means "the order was resting when it got hit", which a whale building a position satisfies exactly as well as a market maker. That contaminated the earlier ranking: the top wallets by P&L were 1.00 legs / 0% all-three and made seven figures on a single match, which is a directional bet, not making. THE DISCRIMINATOR. A market maker works BOTH SIDES of the same token -- buying and selling it repeatedly and carrying little net exposure. A directional trader only ever accumulates. So per wallet, per token: flip = min(buy_qty, sell_qty) / max(buy_qty, sell_qty) ~1 maker, ~0 directional net_exp = |net position| / gross volume ~0 maker, ~1 directional P&L is settlement-aware: cash flow plus net position valued at gamma outcomePrices. That is robust to SPLIT/MERGE even though those never appear as trades -- a split's $1 cost equals the settlement liability of the phantom -1 Yes / -1 No position it creates, because p_yes + p_no = 1. (I previously claimed split/merge broke this accounting. It does not.) ONLY COMPLETE-HISTORY MATCHES are used: /trades refuses offset > ~10500, and a truncated market returns its most recent rows -- a settlement-adjacent slice of 0.999/0.001 cleanup that would make everyone look one-sided and profitable. STILL FILLS, NEVER QUOTES. Polymarket's CLOB is off-chain, so resting orders are public nowhere. "legs" means legs FILLED. """ import collections import csv import json import os import sys import time import requests GAMMA = 'https://gamma-api.polymarket.com' DATA = 'https://data-api.polymarket.com' HERE = os.path.dirname(os.path.abspath(__file__)) OUT = os.path.join(HERE, 'pm_maker_twosided.csv') PAGE = 500 OFFSET_CAP = 10000 def _get(url, params, tries=4): for i in range(tries): try: r = requests.get(url, params=params, timeout=30) except Exception: time.sleep(1.2 * (i + 1)); continue if r.status_code == 200: return r.json() if r.status_code == 400: return None time.sleep(1.2 * (i + 1)) return None def trades(cid, taker_only): rows, off = [], 0 while off < OFFSET_CAP: p = {'market': cid, 'limit': PAGE, 'offset': off} if not taker_only: p['takerOnly'] = 'false' d = _get(f'{DATA}/trades', p) if not d: break rows += d; off += len(d) if len(d) < PAGE: break time.sleep(0.1) return rows def complete_matches(n): out = [] for off in (0, 100, 200, 300, 400): evs = _get(f'{GAMMA}/events', {'limit': 100, 'offset': off, 'tag_slug': 'soccer', 'closed': 'true', 'order': 'volume', 'ascending': 'false'}) if not evs: break for e in evs: ms = e.get('markets') or [] if len(ms) != 3 or ' vs' not in (e.get('title') or '').lower(): continue if not all(m.get('outcomePrices') for m in ms): continue # complete history only if any(_get(f'{DATA}/trades', {'market': m.get('conditionId'), 'limit': 1, 'offset': OFFSET_CAP - 200, 'takerOnly': 'false'}) for m in ms): continue out.append(e) if len(out) >= n: return out return out def main(): n_matches = int(sys.argv[1]) if len(sys.argv) > 1 else 10 evs = complete_matches(n_matches) print(f'{len(evs)} complete-history soccer matches\n') W = collections.defaultdict(lambda: { 'buy': collections.defaultdict(float), 'sell': collections.defaultdict(float), 'cash': 0.0, 'pos': collections.defaultdict(float), 'settle_px': {}, 'mk': 0, 'tk': 0, 'legs': collections.defaultdict(set), 'matches': set()}) for e in evs: title = (e.get('title') or '').encode('ascii', 'replace').decode() print(f' {title[:50]:<50}', end='', flush=True) nr = 0 for m in e['markets']: cid = m.get('conditionId') try: prices = [float(x) for x in json.loads(m.get('outcomePrices'))] except Exception: continue allr = trades(cid, False) tk = {(t['transactionHash'], t['proxyWallet']) for t in trades(cid, True)} nr += len(allr) for t in allr: w = t['proxyWallet'] try: sz = float(t.get('size') or 0); px = float(t.get('price') or 0) oi = int(t.get('outcomeIndex') or 0) except (TypeError, ValueError): continue key = (cid, oi) s = W[w] if (t['transactionHash'], w) in tk: s['tk'] += 1 else: s['mk'] += 1 s['matches'].add(e.get('slug')) s['legs'][e.get('slug')].add(cid) s['settle_px'][key] = prices[oi] if oi < len(prices) else 0.0 if (t.get('side') or '').upper() == 'BUY': s['buy'][key] += sz; s['cash'] -= sz * px; s['pos'][key] += sz else: s['sell'][key] += sz; s['cash'] += sz * px; s['pos'][key] -= sz print(f' {nr:>6} rows') rows = [] for w, s in W.items(): tot = s['mk'] + s['tk'] if tot < 100 or s['mk'] / tot < 0.6: continue keys = set(s['buy']) | set(s['sell']) flips, nets, gross = [], [], 0.0 two_sided = 0 for k in keys: b, sl = s['buy'].get(k, 0.0), s['sell'].get(k, 0.0) g = b + sl gross += g if b > 0 and sl > 0: two_sided += 1 flips.append(min(b, sl) / max(b, sl) if max(b, sl) > 0 else 0.0) nets.append(abs(b - sl)) settle = sum(q * s['settle_px'].get(k, 0.0) for k, q in s['pos'].items()) nlegs = [len(v) for v in s['legs'].values()] rows.append({ 'wallet': w, 'pnl': round(s['cash'] + settle, 2), 'fills': tot, 'maker_share': round(s['mk'] / tot, 3), 'tokens': len(keys), 'two_sided_share': round(two_sided / len(keys), 3) if keys else 0, 'flip': round(sum(flips) / len(flips), 3) if flips else 0, 'net_exposure': round(sum(nets) / gross, 3) if gross else 1.0, 'gross_volume': round(gross), 'matches': len(s['matches']), 'mean_legs': round(sum(nlegs) / len(nlegs), 2) if nlegs else 0, 'all3_share': round(sum(1 for x in nlegs if x == 3) / len(nlegs), 2) if nlegs else 0, }) rows.sort(key=lambda r: -r['pnl']) with open(OUT, 'w', newline='', encoding='utf-8') as fh: wtr = csv.DictWriter(fh, fieldnames=list(rows[0].keys())) wtr.writeheader(); wtr.writerows(rows) print(f'\nwrote {len(rows)} wallets -> {OUT}\n') # A REAL maker works both sides and carries little net exposure. real = [r for r in rows if r['flip'] >= 0.3 and r['net_exposure'] <= 0.6] dirn = [r for r in rows if r['flip'] < 0.1] def summarise(rs, label): if not rs: print(f' {label}: none'); return p = [r for r in rs if r['pnl'] > 0] med = sorted(r['pnl'] for r in rs)[len(rs) // 2] print(' %-26s n=%-4d profitable=%3.0f%% median pnl=%9.0f legs=%.2f all3=%.0f%%' % (label, len(rs), 100 * len(p) / len(rs), med, sum(r['mean_legs'] for r in rs) / len(rs), 100 * sum(r['all3_share'] for r in rs) / len(rs))) print('POPULATION SPLIT') summarise(rows, 'all maker-dominant') summarise(real, 'TRUE MAKERS (2-sided)') summarise(dirn, 'directional (1-sided)') if real: print('\nTRUE MAKERS BY P&L') print(' %-44s %10s %6s %5s %6s %5s %s' % ('wallet', 'pnl', 'fills', 'flip', 'netexp', 'legs', 'all3')) for r in sorted(real, key=lambda r: -r['pnl'])[:12]: print(' %-44s %10.0f %6d %5.2f %6.2f %5.2f %4.0f%%' % (r['wallet'], r['pnl'], r['fills'], r['flip'], r['net_exposure'], r['mean_legs'], 100 * r['all3_share'])) win = [r for r in real if r['pnl'] > 0] los = [r for r in real if r['pnl'] <= 0] print('\nWHAT SEPARATES WINNERS FROM LOSERS (true makers only)') for label, rs in (('winners', win), ('losers', los)): if rs: print(' %-8s n=%-3d flip=%.2f netexp=%.2f legs=%.2f all3=%.0f%% gross=%.0f' % (label, len(rs), sum(r['flip'] for r in rs) / len(rs), sum(r['net_exposure'] for r in rs) / len(rs), sum(r['mean_legs'] for r in rs) / len(rs), 100 * sum(r['all3_share'] for r in rs) / len(rs), sum(r['gross_volume'] for r in rs) / len(rs))) return 0 if __name__ == '__main__': sys.exit(main())