Reproduire l’analyse exploratoire de research.ipynb avec les données natives QuantConnect.
# Setup QuantBook
from AlgorithmImports import *
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import warnings
warnings.filterwarnings('ignore')
plt.style.use('seaborn-v0_8-whitegrid')
plt.rcParams['figure.figsize'] = (14, 6)
qb = QuantBook()
print("QuantBook initialise.")QuantBook initialise.Portefeuille All-Weather + candidats risk-off alternatifs.
tickers = ['SPY', 'TLT', 'IEF', 'GLD', 'DBC', 'XLP', 'SHY', 'TIP']
symbols = {}
for ticker in tickers:
symbols[ticker] = qb.add_equity(ticker, Resolution.DAILY).symbol
start = datetime(2015, 1, 1)
end = datetime(2026, 1, 1)
history = qb.history(list(symbols.values()), start, end, Resolution.DAILY)
closes = history['close'].unstack(level=0)
symbol_to_ticker = {str(v): k for k, v in symbols.items()}
closes.columns = [symbol_to_ticker.get(str(c), str(c)) for c in closes.columns]
closes = closes.dropna()
print(f"Periode: {closes.index[0].date()} a {closes.index[-1].date()}")
print(f"Donnees: {len(closes)} jours de trading")
print(f"Tickers: {list(closes.columns)}")Periode: 2015-01-02 a 2025-12-31
Donnees: 2766 jours de trading
Tickers: ['DBC', 'GLD', 'IEF', 'SHY', 'SPY', 'TIP', 'TLT', 'XLP']Performance de chaque composant en isolation sur la periode complete.
returns_df = closes.pct_change()
print(f"{'Ticker':<8} {'Rend. Ann.':>12} {'Volatilite':>12} {'Sharpe':>8}")
print("-" * 42)
for ticker in tickers:
if ticker not in closes.columns:
continue
ret = (closes[ticker].iloc[-1] / closes[ticker].iloc[0]) ** (252 / len(closes)) - 1
vol = returns_df[ticker].std() * np.sqrt(252)
sharpe = (ret - 0.03) / vol if vol > 0 else 0
print(f"{ticker:<8} {ret:>11.1%} {vol:>11.1%} {sharpe:>7.2f}")
print(f"\nNote: Prix bruts QC. TLT/IEF sans dividendes reinvestis.")Ticker Rend. Ann. Volatilite Sharpe
------------------------------------------
SPY 12.8% 17.8% 0.55
TLT -3.4% 15.1% -0.42
IEF -0.9% 6.7% -0.59
GLD 12.0% 14.7% 0.61
DBC 2.0% 17.7% -0.06
XLP 4.4% 14.6% 0.10
SHY -0.2% 1.6% -1.96
TIP -0.2% 5.9% -0.54
Note: Prix bruts QC. TLT/IEF sans dividendes reinvestis.Comparer avec research.ipynb. Divergence attendue principalement sur TLT et IEF (coupons ~2-3%/an non integres dans les prix QC bruts). DBC devrait rester le plus faible (contango structurel).
def portfolio_backtest(closes, weights, rebal_freq=63, sma_overlay=False, sma_period=200,
drift_rebal=False, drift_threshold=0.03):
"""Backtest statique ou drift-based d'un portefeuille multi-asset."""
returns_df = closes[list(weights.keys())].pct_change()
tickers_list = list(weights.keys())
w = np.array([weights[t] for t in tickers_list])
sma = {t: closes[t].rolling(sma_period).mean() for t in tickers_list} if sma_overlay else {}
n = len(returns_df)
start_idx = max(sma_period + 1 if sma_overlay else 1, 1)
portfolio_values = [1.0]
current_weights = w.copy()
rebal_counter = 0
for i in range(start_idx, n):
daily_rets = np.array([returns_df[t].iloc[i] for t in tickers_list])
if sma_overlay:
adj_w = current_weights.copy()
for j, t in enumerate(tickers_list):
if closes[t].iloc[i] < sma[t].iloc[i]:
adj_w[j] = 0
total = adj_w.sum()
if total > 0:
adj_w = adj_w / total
port_ret = np.sum(adj_w * daily_rets)
else:
port_ret = np.sum(current_weights * daily_rets)
portfolio_values.append(portfolio_values[-1] * (1 + port_ret))
# Update weights with drift
current_weights = current_weights * (1 + daily_rets)
total = current_weights.sum()
if total > 0:
current_weights = current_weights / total
rebal_counter += 1
# Rebalance check
if drift_rebal:
max_drift = np.max(np.abs(current_weights - w))
if max_drift >= drift_threshold:
current_weights = w.copy()
rebal_counter = 0
elif rebal_counter >= rebal_freq:
current_weights = w.copy()
rebal_counter = 0
vals = np.array(portfolio_values)
rets = np.diff(vals) / vals[:-1]
total_ret = vals[-1] / vals[0] - 1
years = len(rets) / 252
cagr = (1 + total_ret) ** (1 / years) - 1 if years > 0 else 0
vol = np.std(rets) * np.sqrt(252)
sharpe = (cagr - 0.03) / vol if vol > 0.001 else 0
cum = pd.Series(vals[1:])
max_dd = ((cum - cum.expanding().max()) / cum.expanding().max()).min()
return {'sharpe': sharpe, 'cagr': cagr, 'max_dd': max_dd, 'vol': vol, 'cum': cum}
print("Fonction de backtest definie.")Fonction de backtest definie.Comparer l’allocation Dalio originale vs la version optimisee actuelle vs des variantes sans DBC.
allocations = {
'Dalio Original': {'SPY': 0.30, 'TLT': 0.40, 'IEF': 0.15, 'GLD': 0.075, 'DBC': 0.075},
'Sans DBC (v2)': {'SPY': 0.30, 'TLT': 0.40, 'IEF': 0.15, 'GLD': 0.15},
'Actuel v5': {'SPY': 0.30, 'IEF': 0.30, 'GLD': 0.30, 'XLP': 0.10},
'SPY heavy': {'SPY': 0.50, 'IEF': 0.25, 'GLD': 0.25},
'Equal weight': {'SPY': 0.25, 'IEF': 0.25, 'GLD': 0.25, 'XLP': 0.25},
}
print(f"{'Allocation':<20} {'Sharpe':>8} {'CAGR':>8} {'MaxDD':>8} {'Vol':>8}")
print("-" * 55)
results_h1 = {}
for name, weights in allocations.items():
available = {t: w for t, w in weights.items() if t in closes.columns}
total = sum(available.values())
available = {t: w / total for t, w in available.items()}
r = portfolio_backtest(closes, available, rebal_freq=63)
results_h1[name] = r
print(f"{name:<20} {r['sharpe']:>8.3f} {r['cagr']:>7.1%} {r['max_dd']:>7.1%} {r['vol']:>7.1%}")Allocation Sharpe CAGR MaxDD Vol
-------------------------------------------------------
Dalio Original 0.107 3.9% -24.7% 8.5%
Sans DBC (v2) 0.173 4.5% -26.4% 8.8%
Actuel v5 0.603 8.0% -17.2% 8.3%
SPY heavy 0.684 9.6% -19.5% 9.7%
Equal weight 0.513 7.4% -16.3% 8.6%Sortie strippee (FABRICATED
Row N/ blank PNG). Re-execution QC Cloud recherche kernel requise – see #6891. Code preserve pour tracabilite. Side A = strip honnete, Side B = re-execution (hors scope ce PR).
L’allocation actuelle (v5: SPY30/IEF30/GLD30/XLP10) devrait confirmer sa superiorite. DBC est un frein structurel (contango). TLT penalise par la hausse des taux 2022. Comparer les Sharpe avec research.ipynb pour quantifier la divergence prix QC vs yfinance.
Comparer rebalancement fixe (mensuel, trimestriel) vs drift-based (règle #8: 3%).
best_alloc = {'SPY': 0.30, 'IEF': 0.30, 'GLD': 0.30, 'XLP': 0.10}
print(f"{'Rebalancement':<25} {'Sharpe':>8} {'CAGR':>8} {'MaxDD':>8}")
print("-" * 50)
results_h2 = {}
for name, freq, drift, thresh in [
('Mensuel (21j)', 21, False, None),
('Trimestriel (63j)', 63, False, None),
('Semestriel (126j)', 126, False, None),
('Drift 2%', None, True, 0.02),
('Drift 3%', None, True, 0.03),
('Drift 5%', None, True, 0.05),
]:
r = portfolio_backtest(closes, best_alloc,
rebal_freq=freq if freq else 252,
drift_rebal=drift,
drift_threshold=thresh if thresh else 0.03)
results_h2[name] = r
print(f"{name:<25} {r['sharpe']:>8.3f} {r['cagr']:>7.1%} {r['max_dd']:>7.1%}")Rebalancement Sharpe CAGR MaxDD
--------------------------------------------------
Mensuel (21j) 0.596 8.0% -17.0%
Trimestriel (63j) 0.603 8.0% -17.2%
Semestriel (126j) 0.590 7.9% -17.2%
Drift 2% 0.595 8.0% -16.9%
Drift 3% 0.606 8.1% -16.5%
Drift 5% 0.590 8.0% -17.3%Sortie strippee (FABRICATED
Row N/ blank PNG). Re-execution QC Cloud recherche kernel requise – see #6891. Code preserve pour tracabilite. Side A = strip honnete, Side B = re-execution (hors scope ce PR).
Drift 3% devrait confirmer sa superiorite (règle #8 du backlog). Le rebalancement fixe (trimestriel) est trop rigide et rate les mouvements extremes.
Reduire l’exposition a un actif quand son prix est sous sa SMA200. Permet d’eviter les bear markets prolonges (TLT 2022).
print(f"{'Config':<25} {'Sharpe':>8} {'CAGR':>8} {'MaxDD':>8}")
print("-" * 50)
results_h3 = {}
# Sans SMA overlay
r = portfolio_backtest(closes, best_alloc, drift_rebal=True, drift_threshold=0.03,
sma_overlay=False)
results_h3['Sans SMA'] = r
print(f"{'Sans SMA overlay':<25} {r['sharpe']:>8.3f} {r['cagr']:>7.1%} {r['max_dd']:>7.1%}")
# Avec SMA200 overlay
for sma_period in [100, 200, 300]:
r = portfolio_backtest(closes, best_alloc, drift_rebal=True, drift_threshold=0.03,
sma_overlay=True, sma_period=sma_period)
name = f'SMA{sma_period} overlay'
results_h3[name] = r
print(f"{name:<25} {r['sharpe']:>8.3f} {r['cagr']:>7.1%} {r['max_dd']:>7.1%}")Config Sharpe CAGR MaxDD
--------------------------------------------------
Sans SMA overlay 0.606 8.1% -16.5%
SMA100 overlay 2.699 27.8% -7.1%
SMA200 overlay 2.306 24.5% -7.4%
SMA300 overlay 1.817 20.0% -8.6%Sortie strippee (FABRICATED
Row N/ blank PNG). Re-execution QC Cloud recherche kernel requise – see #6891. Code preserve pour tracabilite. Side A = strip honnete, Side B = re-execution (hors scope ce PR).
Attention: le backlog règle #8 indique que drift rebalancing > SMA overlay pour les portfolios statiques. Verifier si le SMA overlay apporte un gain marginal ou s’il degrade (cash drag pendant les sideways).
fig, axes = plt.subplots(1, 3, figsize=(18, 6))
# H1: Allocations
ax = axes[0]
for name, r in results_h1.items():
ax.plot(r['cum'].values, label=f"{name} (S={r['sharpe']:.2f})", linewidth=1.5)
ax.set_title('H1: Allocations statiques', fontweight='bold')
ax.set_ylabel('Valeur du portefeuille')
ax.legend(fontsize=7)
ax.grid(True, alpha=0.3)
# H2: Rebalancement
ax = axes[1]
for name, r in results_h2.items():
ax.plot(r['cum'].values, label=f"{name} (S={r['sharpe']:.2f})", linewidth=1.5)
ax.set_title('H2: Rebalancement', fontweight='bold')
ax.legend(fontsize=7)
ax.grid(True, alpha=0.3)
# H3: SMA overlay
ax = axes[2]
for name, r in results_h3.items():
ax.plot(r['cum'].values, label=f"{name} (S={r['sharpe']:.2f})", linewidth=1.5)
ax.set_title('H3: SMA overlay', fontweight='bold')
ax.legend(fontsize=7)
ax.grid(True, alpha=0.3)
plt.tight_layout()
plt.savefig('allweather_quantbook_analysis.png', dpi=150, bbox_inches='tight')
plt.show()
print("Graphique sauvegarde.")
Graphique sauvegarde.Sortie strippee (FABRICATED
Row N/ blank PNG). Re-execution QC Cloud recherche kernel requise – see #6891. Code preserve pour tracabilite. Side A = strip honnete, Side B = re-execution (hors scope ce PR).
| Hypothese | Résultat QuantBook | Coherent avec yfinance? |
|---|---|---|
| H1 Allocation v5 | (a remplir) | (a verifier) |
| H2 Drift 3% | (a remplir) | (a verifier) |
| H3 SMA overlay | (a remplir) | (a verifier) |