Gahow Wang
722cade974
Share point-in-time signals and a cash-funded order ledger across backtests and trading. Correct eligibility, missing-price fills, quality factors, regime inputs, warmup windows, and cash-injection accounting. Add reproducible registered-strategy comparisons and paired forward candidates with a 20% resize band. Validation: 125 Python tests, 23 web tests, TypeScript/Vite build, 87 strategy causality checks, and real-data target/order/equity parity for 10 strategies.
153 lines
7.1 KiB
Python
153 lines
7.1 KiB
Python
"""
|
|
Final check of the recommended US portfolio: three sleeves with different
|
|
return sources, equal capital, rebalanced monthly between sleeves.
|
|
|
|
momentum : overnight-return momentum + distance from 126d low, top 10, 10-day
|
|
rebalance, hold while rank <= 30 (the live overnight_recovery_top10)
|
|
defensive : sector-neutral quality (GP/A, CFO/A, ROE, accruals, leverage) +
|
|
low risk (idiosyncratic vol, beta), top 20, monthly, buffer 2
|
|
(price-only alternative: low risk alone)
|
|
trend : TrendRider V7 VT24 on leveraged ETFs (live trend_rider_v7_vt24)
|
|
|
|
PYTHONPATH=. uv run python -m research.lab.trend_rider_long # once: data/research/tr_long.pkl
|
|
PYTHONPATH=. uv run --with numba python -m research.lab.us_final
|
|
|
|
The live version is ``us_three_sleeve`` (strategies/sleeve_portfolio.py).
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
import os
|
|
import pickle
|
|
import warnings
|
|
|
|
import numpy as np
|
|
import pandas as pd
|
|
|
|
from research.lab import fundamentals as fu
|
|
from research.lab.engine import Costs
|
|
from research.lab.factor_scan import RESULTS, load
|
|
from research.lab.report import deflated_sharpe, stats
|
|
from research.lab.rotation import factor_ranks
|
|
from research.lab.us_sweep import _blend, run, sector_rank
|
|
|
|
warnings.filterwarnings("ignore")
|
|
|
|
SPAN = ("2011-01-03", "2026-09-24")
|
|
LONG = ("2006-09-27", "2026-09-24")
|
|
BLOCKS = [("2006-09-27", "2010-12-31"), ("2011-01-01", "2014-12-31"), ("2015-01-01", "2018-12-31"),
|
|
("2019-01-01", "2022-12-31"), ("2023-01-01", "2026-09-24")]
|
|
|
|
|
|
def sleeve_scores(p):
|
|
pr = factor_ranks(p, "us")
|
|
F = fu.load(p)
|
|
Q = _blend(*(sector_rank(F[k], p) for k in ("gp_a", "cfo_a", "roe", "accruals", "leverage")))
|
|
low = _blend(pr["ivol"], pr["low_beta"])
|
|
return {
|
|
"momentum": (_blend(pr["overnight_mom"], pr["rec_126"]), dict(top_n=10, buffer=2.0, every=10)),
|
|
"defensive": (_blend(Q, low), dict(top_n=20, buffer=2.0, every=21)),
|
|
"lowrisk": (low, dict(top_n=20, buffer=2.0, every=21)),
|
|
}
|
|
|
|
|
|
def monthly_blend(R: pd.DataFrame, weights: dict[str, float]) -> pd.Series:
|
|
"""Sleeves drift inside each month and are reset to ``weights`` at month
|
|
ends (the sleeves are separate accounts; only capital moves)."""
|
|
cols = list(weights)
|
|
w0 = np.array([weights[c] for c in cols])
|
|
out = []
|
|
for _, g in R[cols].groupby(R.index.to_period("M")):
|
|
grow = (1 + g).cumprod()
|
|
val = grow.mul(w0, axis=1).sum(axis=1)
|
|
prev = np.r_[1.0, val.values[:-1]]
|
|
out.append(pd.Series(val.values / prev - 1, index=g.index))
|
|
return pd.concat(out)
|
|
|
|
|
|
def describe(r: pd.Series, blocks=BLOCKS) -> dict:
|
|
s = stats((1 + r).cumprod())
|
|
bl = {f"{a[:4]}-{b[:4]}": stats((1 + r.loc[a:b]).cumprod())["sharpe"]
|
|
for a, b in blocks if len(r.loc[a:b]) > 200}
|
|
yrs = ((1 + r).groupby(r.index.year).prod() - 1)
|
|
return {**{k: float(v) for k, v in s.items()}, "blocks": bl, "yearly": {int(k): float(v) for k, v in yrs.items()}}
|
|
|
|
|
|
def line(name, d) -> str:
|
|
return (f"{name:44s} cagr {d['cagr']*100:5.1f}% sh {d['sharpe']:.2f} vol {d['vol']*100:4.1f}% "
|
|
f"mdd {d['mdd']*100:5.1f}% calmar {d['calmar']:.2f} worst yr {min(d['yearly'].values())*100:5.1f}% "
|
|
f"blocks [{' '.join(f'{v:.2f}' for v in d['blocks'].values())}]")
|
|
|
|
|
|
def main() -> None:
|
|
p, _ = load("us")
|
|
pf = p.slice()
|
|
pf.costs = Costs(p.costs.buy, p.costs.sell, None, 2.0, 10_000.0) # + $2 per order on a $10k sleeve
|
|
S = sleeve_scores(p)
|
|
R, Rfee = {}, {}
|
|
for name, (score, kw) in S.items():
|
|
span = LONG if name in ("momentum", "lowrisk") else SPAN
|
|
R[name] = run(p, score.where(p.member), [span], **kw)["_rets"]
|
|
Rfee[name] = run(pf, score.where(p.member), [span], **kw)["_rets"]
|
|
ew = p.close.notna().astype(float).where(p.member)
|
|
R["ew_sp500"] = run(p, ew, [LONG], top_n=10000, every=10)["_rets"]
|
|
with open(os.path.join("data", "research", "tr_long.pkl"), "rb") as fh:
|
|
tr = pickle.load(fh)["equity"]["trend_rider_v7_vt24"]
|
|
R["trend"] = Rfee["trend"] = tr.pct_change().reindex(p.dates).loc["2010-03-01":] # $2/order already
|
|
R["spy"] = p.benchmark.ffill().pct_change().reindex(p.dates)
|
|
R = pd.DataFrame(R).loc[LONG[0]:LONG[1]]
|
|
Rfee = pd.DataFrame(Rfee).loc[LONG[0]:LONG[1]]
|
|
|
|
out = {}
|
|
span = R.loc[SPAN[0]:SPAN[1], ["momentum", "defensive", "lowrisk", "trend", "spy"]].fillna(0.0)
|
|
monthly = (1 + span).resample("ME").prod() - 1
|
|
print("monthly return correlation 2011-2026:\n", monthly.corr().round(2).to_string(), "\n")
|
|
print("=== sleeves (10 bp costs; trend includes $2/order)")
|
|
for c in R.columns:
|
|
r = R[c].dropna()
|
|
out[c] = describe(r)
|
|
print(line(f"{c} {r.index[0].year}-", out[c]))
|
|
three = {"momentum": 1 / 3, "defensive": 1 / 3, "trend": 1 / 3}
|
|
three_px = {"momentum": 1 / 3, "lowrisk": 1 / 3, "trend": 1 / 3}
|
|
print("\n=== portfolios 2011-2026, monthly rebalanced between sleeves")
|
|
for label, w, src in [("3 sleeves (defensive = quality+lowrisk)", three, R),
|
|
("3 sleeves (defensive = lowrisk, price only)", three_px, R),
|
|
("3 sleeves, + $2/order on $10k sleeves", three, Rfee),
|
|
("momentum + defensive (no ETF sleeve)", {"momentum": .5, "defensive": .5}, R),
|
|
("momentum + trend", {"momentum": .5, "trend": .5}, R)]:
|
|
r = monthly_blend(src.loc[SPAN[0]:SPAN[1]].fillna(0.0), w)
|
|
out[label] = describe(r)
|
|
print(line(label, out[label]))
|
|
print("\n=== price-only walk-back 2006-2026 (includes 2008; trend sleeve from 2010-03, before that 50/50)")
|
|
rr = R.loc[LONG[0]:LONG[1]]
|
|
pre = monthly_blend(rr.loc[:"2010-02-28", ["momentum", "lowrisk"]].fillna(0.0), {"momentum": .5, "lowrisk": .5})
|
|
post = monthly_blend(rr.loc["2010-03-01":].fillna(0.0), three_px)
|
|
r = pd.concat([pre, post])
|
|
out["walkback price-only"] = describe(r)
|
|
print(line("momentum + lowrisk (+ trend from 2010)", out["walkback price-only"]))
|
|
for c in ("momentum", "lowrisk", "ew_sp500", "spy"):
|
|
print(line(f" {c} 2006-", describe(rr[c].dropna())))
|
|
|
|
# deflated Sharpe: every configuration this and the previous round tried
|
|
from research.lab.trials import count_trials
|
|
n = count_trials(RESULTS, "us")
|
|
if n < 1:
|
|
raise ValueError("A trial ledger is required for DSR")
|
|
r = monthly_blend(R.loc[SPAN[0]:SPAN[1]].fillna(0.0), three)
|
|
dsr = deflated_sharpe(r, n)
|
|
out["dsr"] = {"observed_trials_lower_bound": n, "role": "development", **{k: float(v) for k, v in dsr.items()}}
|
|
print(f"\nDSR of the 3-sleeve book: {dsr['dsr']:.3f} (Sharpe {dsr['sr_annual']:.2f} vs "
|
|
f"expected max of {n} unskilled trials {dsr['sr0_annual']:.2f})")
|
|
print("\nyearly returns (3 sleeves vs SPY):")
|
|
spy_y = out["spy"]["yearly"]
|
|
for y, v in out["3 sleeves (defensive = quality+lowrisk)"]["yearly"].items():
|
|
print(f" {y}: {v*100:6.1f}% SPY {spy_y.get(y, float('nan'))*100:6.1f}%")
|
|
with open(os.path.join(RESULTS, "us_final.json"), "w") as fh:
|
|
json.dump(out, fh, indent=1)
|
|
R.to_pickle(os.path.join(RESULTS, "us_final_sleeves.pkl"))
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|