027d73a504
Backtested the cross-venue funding arb on historical funding (Binance/OKX/Hyperliquid). Gross +21%/yr, spreads persist (capture 0.71), but NAIVE daily rebalance is cost-killed (net Sharpe -4.5, negative every month). HYSTERESIS (hold winners until spread decays, entry>10bp/exit>5bp) flips net to +10-14%/yr market-neutral (Sharpe +11-15, but inflated by ~1% vol + idealized fills; realistic ~3-6 / return ~10-14%). Turnover is the swing factor. First edge of the whole search to survive the net-of-cost horde -- market-neutral, persistent, no spot leg (solves hedgeability), operational not predictive. Caveats: 94d/one period, idealized fills, counterparty. Next: switch live harness to hysteresis, deepen history to full year, micro-live. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
85 lines
3.6 KiB
Python
85 lines
3.6 KiB
Python
#!/usr/bin/env python3
|
|
"""N-venue cross-venue funding backtest (Binance/OKX/Hyperliquid, deep history).
|
|
|
|
Per coin-day: spread = max-min daily funding across available venues (short max, long min).
|
|
Pick top-K by |spread|, book the REALIZED next-day max-min difference for the held pair, net of
|
|
turnover cost. Reports Sharpe/return by top-K, persistence, and PER-MONTH Sharpe (regime check).
|
|
"""
|
|
import json
|
|
import math
|
|
import os
|
|
import sys
|
|
|
|
import numpy as np
|
|
|
|
PANEL = "data/surfer/xvenue2/panel2.json"
|
|
COST_RT = 0.0010
|
|
HURDLE = 0.0005
|
|
|
|
|
|
def main():
|
|
panel = json.load(open(PANEL))
|
|
dates = sorted(set().union(*[set(v) for v in panel.values()]))
|
|
di = {d: i for i, d in enumerate(dates)}
|
|
coins = list(panel)
|
|
T, N = len(dates), len(coins)
|
|
# per coin-day: best (max) and worst (min) venue funding, and which venues
|
|
hi = np.full((T, N), np.nan); lo = np.full((T, N), np.nan)
|
|
for j, c in enumerate(coins):
|
|
for d, v in panel[c].items():
|
|
vals = list(v.values())
|
|
if len(vals) >= 2:
|
|
hi[di[d], j] = max(vals); lo[di[d], j] = min(vals)
|
|
spread = hi - lo # always >=0 (max-min)
|
|
print(f"N-venue backtest: {N} coins, {T} days ({dates[0]}..{dates[-1]})")
|
|
|
|
def sim(K, cost):
|
|
rets, prev = [], set()
|
|
for t in range(T - 1):
|
|
s_t, s_n = spread[t], spread[t + 1]
|
|
ok = np.isfinite(s_t) & np.isfinite(s_n) & (s_t > HURDLE)
|
|
idx = np.where(ok)[0]
|
|
if len(idx) == 0:
|
|
rets.append(0.0); continue
|
|
top = idx[np.argsort(-s_t[idx])[:K]]
|
|
realized = float(np.mean(s_n[top])) # next-day max-min still captured if pair persists
|
|
cur = set(coins[j] for j in top)
|
|
turn = len(cur ^ prev) / max(len(cur), 1)
|
|
rets.append(realized - turn * (cost / 2)); prev = cur
|
|
r = np.array(rets)
|
|
ann = r.mean() * 365; vol = r.std() * math.sqrt(365)
|
|
eq = np.cumprod(1 + r); dd = float((eq / np.maximum.accumulate(eq) - 1).min())
|
|
return ann, vol, (ann / vol if vol > 0 else float("nan")), dd, r
|
|
|
|
print(f"\n{'topK':>5} {'annNET%':>8} {'Sharpe':>7} {'maxDD%':>7} {'gross%':>7}")
|
|
R = {}
|
|
for K in [5, 10, 20]:
|
|
a, v, sh, dd, r = sim(K, COST_RT); g = sim(K, 0.0)[0]; R[K] = r
|
|
print(f"{K:>5} {100*a:>+8.1f} {sh:>+7.2f} {100*dd:>+7.1f} {100*g:>+7.0f}")
|
|
|
|
# persistence: top-spread coin still positive-spread direction next day. spread is max-min(>=0);
|
|
# the real persistence q = does the SAME venue stay highest? approximate via spread autocorr sign>0 always,
|
|
# so report realized/snapshot ratio = how much of today's spread you actually collect next day.
|
|
capt = []
|
|
for t in range(T - 1):
|
|
ok = np.isfinite(spread[t]) & np.isfinite(spread[t + 1]) & (spread[t] > HURDLE)
|
|
idx = np.where(ok)[0]
|
|
if len(idx):
|
|
top = idx[np.argsort(-spread[t][idx])[:10]]
|
|
capt.append(np.mean(spread[t + 1][top]) / max(np.mean(spread[t][top]), 1e-9))
|
|
print(f"\n capture ratio (next-day spread / today's spread, top-10): {np.mean(capt):.2f} (1.0=fully persists, ~0=collapses)")
|
|
|
|
# per-month Sharpe (regime check) on top-10
|
|
r10 = R[10]
|
|
mo = [dates[t][:7] for t in range(T - 1)]
|
|
print(" per-month Sharpe (top-10, net):")
|
|
for m in sorted(set(mo)):
|
|
seg = r10[np.array(mo) == m]
|
|
if len(seg) > 8 and seg.std() > 0:
|
|
print(f" {m}: {seg.mean()/seg.std()*math.sqrt(365):+.1f} (n={len(seg)})")
|
|
print("\n VERDICT: net Sharpe>1 across MONTHS + capture ratio>0.5 = robust deployable edge.")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|