Menu
Expert guidance for algorithmic trading and quantitative research. Use when developing trading strategies, backtesting, building trading systems, or conducting quantitative analysis.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Comprehensive machine learning guidance covering ML algorithms, deep learning, model training, evaluation, deployment, MLOps, and production best practices. Use when building ML models, training neural networks, implementing ML pipelines, or deploying ML systems to production.
Expert Python programming guidance covering modern best practices, design patterns, testing, type hints, and performance optimization. Use when writing Python code, refactoring, debugging, or implementing Python projects.
Comprehensive software engineering practices covering system design, architecture patterns, SOLID principles, clean code, testing strategies, CI/CD, and development workflows. Use when designing systems, reviewing code, implementing best practices, or planning software architecture.
Basé sur la classification professionnelle SOC
| name | quantitative-finance |
| description | Expert guidance for algorithmic trading and quantitative research. Use when developing trading strategies, backtesting, building trading systems, or conducting quantitative analysis. |
Expert guidance for senior quantitative developers, researchers, and systematic traders building sophisticated trading systems and conducting quantitative research.
from typing import Dict, List, Tuple
import pandas as pd
import numpy as np
from datetime import datetime
class AlphaResearch:
"""Framework for systematic alpha research."""
def __init__(self, universe: List[str], start_date: str, end_date: str):
self.universe = universe
self.start_date = pd.to_datetime(start_date)
self.end_date = pd.to_datetime(end_date)
self.data = None
self.signals = None
def load_data(self) -> pd.DataFrame:
"""Load and clean market data."""
# Load price, volume, fundamental data
data = self._fetch_data()
data = self._clean_data(data)
data = self._calculate_returns(data)
self.data = data
return data
def generate_alpha(self, params: Dict) -> pd.DataFrame:
"""Generate alpha signals based on strategy logic."""
signals = pd.DataFrame(index=self.data.index)
# Example: Mean reversion strategy
lookback = params.get('lookback', 20)
zscore = params.get('zscore_threshold', 2.0)
for ticker in self.universe:
price = self.data[ticker]['close']
ma = price.rolling(lookback).mean()
std = price.rolling(lookback).std()
zscore_val = (price - ma) / std
signals[ticker] = -np.sign(zscore_val) * (np.abs(zscore_val) > zscore)
self.signals = signals
return signals
def backtest(self, signals: pd.DataFrame,
transaction_costs: float = 0.001) -> Dict:
"""Backtest strategy with realistic assumptions."""
returns = self.data.xs('returns', level=1, axis=1)
# Position changes (trades)
positions = signals.shift(1).fillna(0)
trades = positions.diff().fillna(positions)
# Calculate returns
strategy_returns = (positions * returns).sum(axis=1)
# Apply transaction costs
tc = np.abs(trades).sum(axis=1) * transaction_costs
strategy_returns = strategy_returns - tc
# Calculate metrics
metrics = self._calculate_metrics(strategy_returns)
return {
'returns': strategy_returns,
'positions': positions,
'metrics': metrics
}
def _calculate_metrics(self, returns: pd.Series) -> Dict:
"""Calculate performance metrics."""
total_return = (1 + returns).prod() - 1
annual_return = (1 + total_return) ** (252 / len(returns)) - 1
annual_vol = returns.std() * np.sqrt(252)
sharpe = annual_return / annual_vol if annual_vol > 0 else 0
# Calculate max drawdown
cum_returns = (1 + returns).cumprod()
running_max = cum_returns.expanding().max()
drawdown = (cum_returns - running_max) / running_max
max_drawdown = drawdown.min()
return {
'total_return': total_return,
'annual_return': annual_return,
'annual_volatility': annual_vol,
'sharpe_ratio': sharpe,
'max_drawdown': max_drawdown,
'win_rate': (returns > 0).sum() / len(returns)
}
Hypothesis Formation
Data Collection & Cleaning
Feature Engineering
Backtesting
Risk Management
from abc import ABC, abstractmethod
from dataclasses import dataclass
from typing import Optional, List
import asyncio
from datetime import datetime
import numpy as np
@dataclass
class Order:
"""Order representation."""
symbol: str
side: str # 'buy' or 'sell'
quantity: int
order_type: str # 'market', 'limit', 'stop'
price: Optional[float] = None
timestamp: datetime = None
order_id: Optional[str] = None
status: str = 'pending'
@dataclass
class Position:
"""Position representation."""
symbol: str
quantity: int
entry_price: float
current_price: float
unrealized_pnl: float
realized_pnl: float = 0.0
class TradingStrategy(ABC):
"""Base class for trading strategies."""
@abstractmethod
async def on_market_data(self, data: Dict) -> List[Order]:
"""Process market data and generate orders."""
pass
@abstractmethod
async def on_order_update(self, order: Order) -> None:
"""Handle order status updates."""
pass
@abstractmethod
async def on_position_update(self, position: Position) -> None:
"""Handle position updates."""
pass
class OrderManagementSystem:
"""Order management and execution system."""
def __init__(self):
self.orders: Dict[str, Order] = {}
self.positions: Dict[str, Position] = {}
self.fills: List[Dict] = []
async def submit_order(self, order: Order) -> str:
"""Submit order to exchange."""
order.timestamp = datetime.now()
order.order_id = self._generate_order_id()
# Validate order
if not self._validate_order(order):
order.status = 'rejected'
return order.order_id
# Risk checks
if not self._risk_check(order):
order.status = 'rejected'
return order.order_id
# Submit to exchange
order.status = 'submitted'
self.orders[order.order_id] = order
# Simulate async execution
asyncio.create_task(self._execute_order(order))
return order.order_id
async def _execute_order(self, order: Order) -> None:
"""Simulate order execution."""
await asyncio.sleep(0.1) # Simulate network latency
# Fill order
fill_price = order.price if order.order_type == 'limit' else self._get_market_price(order.symbol)
fill = {
'order_id': order.order_id,
'symbol': order.symbol,
'quantity': order.quantity,
'price': fill_price,
'timestamp': datetime.now()
}
self.fills.append(fill)
order.status = 'filled'
# Update positions
self._update_position(fill)
def _update_position(self, fill: Dict) -> None:
"""Update position based on fill."""
symbol = fill['symbol']
if symbol not in self.positions:
self.positions[symbol] = Position(
symbol=symbol,
quantity=fill['quantity'],
entry_price=fill['price'],
current_price=fill['price'],
unrealized_pnl=0.0
)
else:
pos = self.positions[symbol]
# Update position (simplified)
total_cost = pos.quantity * pos.entry_price + fill['quantity'] * fill['price']
pos.quantity += fill['quantity']
pos.entry_price = total_cost / pos.quantity if pos.quantity != 0 else 0
def _validate_order(self, order: Order) -> bool:
"""Validate order parameters."""
if order.quantity <= 0:
return False
if order.order_type == 'limit' and order.price is None:
return False
return True
def _risk_check(self, order: Order) -> bool:
"""Pre-trade risk checks."""
# Check position limits
# Check notional limits
# Check leverage
# Check buying power
return True
def _generate_order_id(self) -> str:
"""Generate unique order ID."""
return f"ORD_{datetime.now().strftime('%Y%m%d%H%M%S%f')}"
def _get_market_price(self, symbol: str) -> float:
"""Get current market price."""
# In production, fetch from market data feed
return 100.0
class RiskManager:
"""Real-time risk management."""
def __init__(self, max_position_size: float, max_portfolio_var: float):
self.max_position_size = max_position_size
self.max_portfolio_var = max_portfolio_var
def check_position_limit(self, symbol: str, quantity: int,
current_position: int) -> bool:
"""Check if new position would exceed limits."""
new_position = current_position + quantity
return abs(new_position) <= self.max_position_size
def check_portfolio_risk(self, positions: Dict[str, Position]) -> bool:
"""Check portfolio-level risk metrics."""
# Calculate portfolio VaR
portfolio_var = self._calculate_portfolio_var(positions)
return portfolio_var <= self.max_portfolio_var
def calculate_position_size(self, signal_strength: float,
volatility: float,
account_value: float) -> int:
"""Calculate optimal position size using Kelly criterion."""
# Simplified Kelly criterion
win_rate = 0.55 # Historical win rate
win_loss_ratio = 1.5 # Average win / average loss
kelly_fraction = (win_rate * win_loss_ratio - (1 - win_rate)) / win_loss_ratio
kelly_fraction = min(kelly_fraction, 0.25) # Cap at 25%
# Adjust for signal strength and volatility
position_size = kelly_fraction * account_value * signal_strength / volatility
return int(position_size)
def _calculate_portfolio_var(self, positions: Dict[str, Position],
confidence_level: float = 0.95) -> float:
"""Calculate portfolio Value at Risk."""
# Simplified VaR calculation
# In production, use historical simulation or Monte Carlo
portfolio_value = sum(p.quantity * p.current_price for p in positions.values())
portfolio_volatility = 0.02 # Daily volatility
z_score = 1.645 if confidence_level == 0.95 else 2.326
var = portfolio_value * portfolio_volatility * z_score
return var
import pandas as pd
import numpy as np
from statsmodels.tsa.stattools import adfuller, acf, pacf
from statsmodels.tsa.arima.model import ARIMA
from arch import arch_model
class TimeSeriesAnalysis:
"""Time series analysis for trading strategies."""
@staticmethod
def test_stationarity(series: pd.Series) -> Dict:
"""Augmented Dickey-Fuller test for stationarity."""
result = adfuller(series.dropna())
return {
'adf_statistic': result[0],
'p_value': result[1],
'is_stationary': result[1] < 0.05,
'critical_values': result[4]
}
@staticmethod
def calculate_half_life(series: pd.Series) -> float:
"""Calculate mean reversion half-life."""
lag = series.shift(1)
delta = series - lag
# Run OLS regression
from sklearn.linear_model import LinearRegression
model = LinearRegression()
model.fit(lag.dropna().values.reshape(-1, 1),
delta.dropna().values)
lambda_coef = model.coef_[0]
half_life = -np.log(2) / lambda_coef
return half_life
@staticmethod
def fit_garch(returns: pd.Series, p: int = 1, q: int = 1) -> Dict:
"""Fit GARCH model for volatility forecasting."""
model = arch_model(returns, vol='Garch', p=p, q=q)
result = model.fit(disp='off')
# Forecast next period volatility
forecast = result.forecast(horizon=1)
next_vol = np.sqrt(forecast.variance.values[-1, 0])
return {
'model': result,
'forecast_volatility': next_vol,
'params': result.params
}
@staticmethod
def cointegration_test(series1: pd.Series, series2: pd.Series) -> Dict:
"""Test for cointegration between two series."""
from statsmodels.tsa.stattools import coint
score, pvalue, _ = coint(series1, series2)
return {
'test_statistic': score,
'p_value': pvalue,
'is_cointegrated': pvalue < 0.05
}
import pandas as pd
import numpy as np
from typing import List, Dict
class FactorModel:
"""Multi-factor model for return attribution."""
def __init__(self, factors: List[str]):
self.factors = factors
self.factor_returns = None
self.factor_loadings = None
def fit(self, returns: pd.DataFrame, factor_data: pd.DataFrame) -> None:
"""Fit factor model to historical data."""
from sklearn.linear_model import LinearRegression
self.factor_returns = factor_data[self.factors]
self.factor_loadings = pd.DataFrame(index=returns.columns,
columns=self.factors)
for asset in returns.columns:
model = LinearRegression()
model.fit(self.factor_returns.values, returns[asset].values)
self.factor_loadings.loc[asset] = model.coef_
def calculate_factor_returns(self, date: str) -> pd.Series:
"""Calculate factor returns for a given date."""
return self.factor_returns.loc[date]
def attribute_returns(self, portfolio_weights: pd.Series,
date: str) -> Dict:
"""Attribute portfolio returns to factors."""
factor_rets = self.calculate_factor_returns(date)
# Calculate factor contribution
factor_contrib = {}
for factor in self.factors:
loadings = self.factor_loadings[factor]
weighted_loadings = (loadings * portfolio_weights).sum()
factor_contrib[factor] = weighted_loadings * factor_rets[factor]
return factor_contrib
import pandas as pd
import numpy as np
from typing import List
class FeatureEngineering:
"""Feature engineering for ML trading strategies."""
@staticmethod
def create_technical_features(df: pd.DataFrame) -> pd.DataFrame:
"""Create technical analysis features."""
features = df.copy()
# Returns
features['returns'] = df['close'].pct_change()
features['log_returns'] = np.log(df['close'] / df['close'].shift(1))
# Moving averages
for window in [5, 10, 20, 50, 200]:
features[f'sma_{window}'] = df['close'].rolling(window).mean()
features[f'ema_{window}'] = df['close'].ewm(span=window).mean()
# Volatility
features['volatility_20'] = df['returns'].rolling(20).std()
features['atr_14'] = self._calculate_atr(df, 14)
# Momentum
features['rsi_14'] = self._calculate_rsi(df['close'], 14)
features['macd'] = self._calculate_macd(df['close'])
# Volume
features['volume_sma_20'] = df['volume'].rolling(20).mean()
features['volume_ratio'] = df['volume'] / features['volume_sma_20']
return features
@staticmethod
def create_microstructure_features(df: pd.DataFrame) -> pd.DataFrame:
"""Create market microstructure features."""
features = df.copy()
# Bid-ask spread
features['spread'] = df['ask'] - df['bid']
features['spread_bps'] = features['spread'] / df['mid'] * 10000
# Order flow imbalance
features['order_imbalance'] = (df['bid_size'] - df['ask_size']) / (df['bid_size'] + df['ask_size'])
# VWAP
features['vwap'] = (df['close'] * df['volume']).cumsum() / df['volume'].cumsum()
features['vwap_distance'] = (df['close'] - features['vwap']) / features['vwap']
return features
@staticmethod
def _calculate_rsi(prices: pd.Series, period: int = 14) -> pd.Series:
"""Calculate Relative Strength Index."""
delta = prices.diff()
gain = (delta.where(delta > 0, 0)).rolling(window=period).mean()
loss = (-delta.where(delta < 0, 0)).rolling(window=period).mean()
rs = gain / loss
rsi = 100 - (100 / (1 + rs))
return rsi
@staticmethod
def _calculate_macd(prices: pd.Series,
fast: int = 12, slow: int = 26) -> pd.Series:
"""Calculate MACD."""
ema_fast = prices.ewm(span=fast).mean()
ema_slow = prices.ewm(span=slow).mean()
macd = ema_fast - ema_slow
return macd
@staticmethod
def _calculate_atr(df: pd.DataFrame, period: int = 14) -> pd.Series:
"""Calculate Average True Range."""
high_low = df['high'] - df['low']
high_close = np.abs(df['high'] - df['close'].shift())
low_close = np.abs(df['low'] - df['close'].shift())
ranges = pd.concat([high_low, high_close, low_close], axis=1)
true_range = ranges.max(axis=1)
atr = true_range.rolling(period).mean()
return atr
class MLTradingModel:
"""Machine learning model for trading."""
def __init__(self, model_type: str = 'xgboost'):
self.model_type = model_type
self.model = None
self.feature_importance = None
def train(self, X: pd.DataFrame, y: pd.Series,
validation_split: float = 0.2) -> Dict:
"""Train ML model with proper validation."""
from sklearn.model_selection import TimeSeriesSplit
# Time series cross-validation
tscv = TimeSeriesSplit(n_splits=5)
if self.model_type == 'xgboost':
from xgboost import XGBClassifier
self.model = XGBClassifier(
n_estimators=100,
max_depth=5,
learning_rate=0.1,
random_state=42
)
# Train with cross-validation
scores = []
for train_idx, val_idx in tscv.split(X):
X_train, X_val = X.iloc[train_idx], X.iloc[val_idx]
y_train, y_val = y.iloc[train_idx], y.iloc[val_idx]
self.model.fit(X_train, y_train)
score = self.model.score(X_val, y_val)
scores.append(score)
# Get feature importance
self.feature_importance = pd.Series(
self.model.feature_importances_,
index=X.columns
).sort_values(ascending=False)
return {
'cv_scores': scores,
'mean_cv_score': np.mean(scores),
'feature_importance': self.feature_importance
}
def predict_signal(self, X: pd.DataFrame) -> np.ndarray:
"""Generate trading signals."""
probabilities = self.model.predict_proba(X)
# Convert probabilities to signals (-1, 0, 1)
signals = np.where(probabilities[:, 1] > 0.55, 1,
np.where(probabilities[:, 1] < 0.45, -1, 0))
return signals
import pandas as pd
import numpy as np
from typing import Dict, List, Callable
from dataclasses import dataclass
@dataclass
class BacktestConfig:
"""Backtesting configuration."""
initial_capital: float = 1000000
commission: float = 0.001 # 10 bps
slippage: float = 0.0005 # 5 bps
margin_requirement: float = 0.25
position_limit: float = 0.1 # 10% per position
class Backtester:
"""Professional backtesting engine."""
def __init__(self, config: BacktestConfig):
self.config = config
self.trades: List[Dict] = []
self.portfolio_values: List[float] = []
self.positions: Dict[str, int] = {}
def run(self, strategy: Callable, data: pd.DataFrame) -> Dict:
"""Run backtest."""
cash = self.config.initial_capital
portfolio_value = cash
for date, row in data.iterrows():
# Generate signals
signals = strategy(date, row, self.positions)
# Execute trades
for symbol, signal in signals.items():
if signal != 0:
trade = self._execute_trade(
symbol, signal, row[symbol]['close'],
date, cash, portfolio_value
)
if trade:
self.trades.append(trade)
cash -= trade['cost']
# Update portfolio value
positions_value = sum(
self.positions.get(symbol, 0) * row[symbol]['close']
for symbol in self.positions
)
portfolio_value = cash + positions_value
self.portfolio_values.append(portfolio_value)
# Calculate metrics
metrics = self._calculate_metrics()
return {
'trades': pd.DataFrame(self.trades),
'portfolio_values': pd.Series(self.portfolio_values, index=data.index),
'metrics': metrics
}
def _execute_trade(self, symbol: str, signal: int, price: float,
date: pd.Timestamp, cash: float,
portfolio_value: float) -> Dict:
"""Execute trade with realistic assumptions."""
# Calculate position size
max_position_value = portfolio_value * self.config.position_limit
if signal > 0: # Buy
shares = int(max_position_value / price)
else: # Sell
shares = -self.positions.get(symbol, 0)
if shares == 0:
return None
# Apply slippage
execution_price = price * (1 + self.config.slippage * np.sign(shares))
# Calculate costs
notional = abs(shares * execution_price)
commission = notional * self.config.commission
total_cost = shares * execution_price + commission
# Check if we have enough cash
if total_cost > cash:
return None
# Update position
self.positions[symbol] = self.positions.get(symbol, 0) + shares
return {
'date': date,
'symbol': symbol,
'shares': shares,
'price': execution_price,
'commission': commission,
'cost': total_cost
}
def _calculate_metrics(self) -> Dict:
"""Calculate comprehensive performance metrics."""
returns = pd.Series(self.portfolio_values).pct_change().dropna()
# Basic metrics
total_return = (self.portfolio_values[-1] / self.config.initial_capital) - 1
annual_return = (1 + total_return) ** (252 / len(returns)) - 1
annual_vol = returns.std() * np.sqrt(252)
sharpe = annual_return / annual_vol if annual_vol > 0 else 0
# Drawdown analysis
cum_returns = (1 + returns).cumprod()
running_max = cum_returns.expanding().max()
drawdown = (cum_returns - running_max) / running_max
max_drawdown = drawdown.min()
# Calculate Calmar ratio
calmar = annual_return / abs(max_drawdown) if max_drawdown != 0 else 0
# Trade analysis
trades_df = pd.DataFrame(self.trades)
if len(trades_df) > 0:
avg_trade_size = trades_df['cost'].abs().mean()
total_commission = trades_df['commission'].sum()
else:
avg_trade_size = 0
total_commission = 0
return {
'total_return': total_return,
'annual_return': annual_return,
'annual_volatility': annual_vol,
'sharpe_ratio': sharpe,
'max_drawdown': max_drawdown,
'calmar_ratio': calmar,
'num_trades': len(self.trades),
'avg_trade_size': avg_trade_size,
'total_commission': total_commission,
'win_rate': (returns > 0).sum() / len(returns)
}
import pandas as pd
import numpy as np
from typing import Dict, List
class OrderBook:
"""Limit order book analysis."""
def __init__(self):
self.bids: List[Dict] = []
self.asks: List[Dict] = []
def update(self, bids: List[Dict], asks: List[Dict]) -> None:
"""Update order book."""
self.bids = sorted(bids, key=lambda x: x['price'], reverse=True)
self.asks = sorted(asks, key=lambda x: x['price'])
def get_best_bid_ask(self) -> tuple:
"""Get best bid and ask."""
best_bid = self.bids[0]['price'] if self.bids else None
best_ask = self.asks[0]['price'] if self.asks else None
return best_bid, best_ask
def get_mid_price(self) -> float:
"""Calculate mid price."""
best_bid, best_ask = self.get_best_bid_ask()
if best_bid and best_ask:
return (best_bid + best_ask) / 2
return None
def get_spread(self) -> float:
"""Calculate bid-ask spread."""
best_bid, best_ask = self.get_best_bid_ask()
if best_bid and best_ask:
return best_ask - best_bid
return None
def calculate_imbalance(self, levels: int = 5) -> float:
"""Calculate order book imbalance."""
bid_volume = sum(b['size'] for b in self.bids[:levels])
ask_volume = sum(a['size'] for a in self.asks[:levels])
if bid_volume + ask_volume > 0:
return (bid_volume - ask_volume) / (bid_volume + ask_volume)
return 0
def estimate_market_impact(self, order_size: int,
side: str) -> float:
"""Estimate market impact of an order."""
if side == 'buy':
levels = self.asks
else:
levels = self.bids
remaining_size = order_size
total_cost = 0
for level in levels:
if remaining_size <= 0:
break
fill_size = min(remaining_size, level['size'])
total_cost += fill_size * level['price']
remaining_size -= fill_size
if remaining_size > 0:
# Order cannot be fully filled
return float('inf')
avg_price = total_cost / order_size
mid_price = self.get_mid_price()
return (avg_price - mid_price) / mid_price
class ExecutionOptimizer:
"""Optimal execution strategies."""
@staticmethod
def twap_schedule(total_size: int, duration_minutes: int,
interval_minutes: int = 1) -> List[int]:
"""Time-Weighted Average Price execution schedule."""
num_intervals = duration_minutes // interval_minutes
size_per_interval = total_size // num_intervals
schedule = [size_per_interval] * num_intervals
# Add remainder to last interval
schedule[-1] += total_size - sum(schedule)
return schedule
@staticmethod
def vwap_schedule(total_size: int,
historical_volume_profile: pd.Series) -> List[int]:
"""Volume-Weighted Average Price execution schedule."""
total_hist_volume = historical_volume_profile.sum()
schedule = []
for vol in historical_volume_profile:
size = int(total_size * (vol / total_hist_volume))
schedule.append(size)
# Adjust for rounding
schedule[-1] += total_size - sum(schedule)
return schedule
@staticmethod
def implementation_shortfall(arrival_price: float,
execution_prices: List[float],
sizes: List[int],
commission: float = 0.001) -> Dict:
"""Calculate implementation shortfall."""
total_size = sum(sizes)
total_cost = sum(p * s for p, s in zip(execution_prices, sizes))
avg_price = total_cost / total_size
# Slippage
slippage = (avg_price - arrival_price) / arrival_price
# Commission
commission_cost = total_cost * commission
# Total shortfall
total_shortfall = (avg_price - arrival_price) * total_size + commission_cost
return {
'avg_execution_price': avg_price,
'slippage': slippage,
'commission_cost': commission_cost,
'total_shortfall': total_shortfall,
'shortfall_bps': (total_shortfall / (arrival_price * total_size)) * 10000
}
import numpy as np
import pandas as pd
from scipy.optimize import minimize
from typing import Dict, Optional
class PortfolioOptimizer:
"""Portfolio construction and optimization."""
def __init__(self, returns: pd.DataFrame):
self.returns = returns
self.mean_returns = returns.mean()
self.cov_matrix = returns.cov()
self.num_assets = len(returns.columns)
def max_sharpe_ratio(self, risk_free_rate: float = 0.02) -> Dict:
"""Maximize Sharpe ratio."""
def neg_sharpe(weights):
portfolio_return = np.sum(self.mean_returns * weights) * 252
portfolio_vol = np.sqrt(
np.dot(weights.T, np.dot(self.cov_matrix * 252, weights))
)
sharpe = (portfolio_return - risk_free_rate) / portfolio_vol
return -sharpe
constraints = {'type': 'eq', 'fun': lambda x: np.sum(x) - 1}
bounds = tuple((0, 1) for _ in range(self.num_assets))
initial_guess = np.array([1/self.num_assets] * self.num_assets)
result = minimize(
neg_sharpe, initial_guess,
method='SLSQP', bounds=bounds, constraints=constraints
)
optimal_weights = result.x
portfolio_return = np.sum(self.mean_returns * optimal_weights) * 252
portfolio_vol = np.sqrt(
np.dot(optimal_weights.T, np.dot(self.cov_matrix * 252, optimal_weights))
)
return {
'weights': pd.Series(optimal_weights, index=self.returns.columns),
'return': portfolio_return,
'volatility': portfolio_vol,
'sharpe': (portfolio_return - risk_free_rate) / portfolio_vol
}
def min_variance(self) -> Dict:
"""Minimum variance portfolio."""
def portfolio_variance(weights):
return np.dot(weights.T, np.dot(self.cov_matrix * 252, weights))
constraints = {'type': 'eq', 'fun': lambda x: np.sum(x) - 1}
bounds = tuple((0, 1) for _ in range(self.num_assets))
initial_guess = np.array([1/self.num_assets] * self.num_assets)
result = minimize(
portfolio_variance, initial_guess,
method='SLSQP', bounds=bounds, constraints=constraints
)
optimal_weights = result.x
portfolio_return = np.sum(self.mean_returns * optimal_weights) * 252
portfolio_vol = np.sqrt(portfolio_variance(optimal_weights))
return {
'weights': pd.Series(optimal_weights, index=self.returns.columns),
'return': portfolio_return,
'volatility': portfolio_vol
}
def risk_parity(self) -> Dict:
"""Risk parity portfolio."""
def risk_contribution(weights):
portfolio_vol = np.sqrt(
np.dot(weights.T, np.dot(self.cov_matrix * 252, weights))
)
marginal_contrib = np.dot(self.cov_matrix * 252, weights)
contrib = weights * marginal_contrib / portfolio_vol
return contrib
def risk_parity_objective(weights):
contrib = risk_contribution(weights)
target_contrib = portfolio_vol / self.num_assets
return np.sum((contrib - target_contrib) ** 2)
constraints = {'type': 'eq', 'fun': lambda x: np.sum(x) - 1}
bounds = tuple((0, 1) for _ in range(self.num_assets))
initial_guess = np.array([1/self.num_assets] * self.num_assets)
result = minimize(
risk_parity_objective, initial_guess,
method='SLSQP', bounds=bounds, constraints=constraints
)
optimal_weights = result.x
portfolio_return = np.sum(self.mean_returns * optimal_weights) * 252
portfolio_vol = np.sqrt(
np.dot(optimal_weights.T, np.dot(self.cov_matrix * 252, optimal_weights))
)
return {
'weights': pd.Series(optimal_weights, index=self.returns.columns),
'return': portfolio_return,
'volatility': portfolio_vol,
'risk_contributions': risk_contribution(optimal_weights)
}
def black_litterman(self, market_caps: pd.Series,
views: Optional[Dict] = None) -> Dict:
"""Black-Litterman portfolio optimization."""
# Calculate market implied returns
market_weights = market_caps / market_caps.sum()
risk_aversion = 2.5
market_returns = risk_aversion * np.dot(
self.cov_matrix * 252, market_weights
)
if views is None:
# No views, return market portfolio
return {
'weights': market_weights,
'expected_returns': market_returns
}
# Incorporate views (simplified)
# In production, implement full Black-Litterman model
return {
'weights': market_weights,
'expected_returns': market_returns
}
import logging
from typing import Dict, Callable
from datetime import datetime
import smtplib
from email.mime.text import MIMEText
class TradingSystemMonitor:
"""Monitor trading system health."""
def __init__(self, alert_email: str):
self.alert_email = alert_email
self.logger = logging.getLogger(__name__)
self.metrics: Dict = {}
def check_system_health(self) -> bool:
"""Perform system health checks."""
checks = [
self._check_market_data_feed(),
self._check_order_execution(),
self._check_risk_limits(),
self._check_position_reconciliation()
]
return all(checks)
def _check_market_data_feed(self) -> bool:
"""Check if market data feed is active."""
# Check last update timestamp
# Check data quality
# Check for stale data
return True
def _check_order_execution(self) -> bool:
"""Check order execution system."""
# Check pending orders
# Check order latency
# Check fill rates
return True
def _check_risk_limits(self) -> bool:
"""Check risk limit breaches."""
# Check position limits
# Check VaR limits
# Check drawdown limits
return True
def _check_position_reconciliation(self) -> bool:
"""Reconcile positions with broker."""
# Compare internal positions with broker
# Check for discrepancies
return True
def send_alert(self, message: str, severity: str = 'WARNING') -> None:
"""Send alert notification."""
self.logger.log(
logging.WARNING if severity == 'WARNING' else logging.ERROR,
message
)
# Send email alert
self._send_email_alert(message, severity)
def _send_email_alert(self, message: str, severity: str) -> None:
"""Send email alert."""
subject = f"[{severity}] Trading System Alert"
msg = MIMEText(f"{message}\n\nTimestamp: {datetime.now()}")
msg['Subject'] = subject
msg['From'] = 'trading-system@example.com'
msg['To'] = self.alert_email
# In production, configure SMTP server
# smtplib.SMTP().send_message(msg)
# Wrong - optimizing on same data used for testing
best_params = grid_search(data, params)
backtest(data, best_params)
# Correct - walk-forward optimization
for train_period, test_period in walk_forward_splits(data):
params = optimize(train_period)
results.append(backtest(test_period, params))
# Wrong - using future data
df['signal'] = df['close'].shift(-1) > df['close']
# Correct - only use past data
df['signal'] = df['close'].shift(1) > df['close'].shift(2)
# Wrong - only using currently listed stocks
universe = get_current_stock_universe()
# Correct - use point-in-time universe
universe = get_historical_universe(date)
# Wrong - ignoring costs
returns = signal * price_change
# Correct - include realistic costs
returns = signal * price_change - abs(signal.diff()) * (commission + slippage)