jashdoshi77's picture
addes more section , improvments , scalability
c81e8a5
"""
Backtesting Engine.
Event-driven backtester that simulates strategy execution over historical data:
- Iterates daily bars
- Applies strategy entry/exit logic
- Tracks positions and portfolio value
- Accounts for transaction costs and slippage
- Produces equity curve, trade log, and comprehensive metrics
"""
from __future__ import annotations
import json
import logging
from datetime import date, timedelta
from typing import Any, Dict, List, Optional
import numpy as np
import pandas as pd
from app.config import get_settings
from app.services.data_ingestion.yahoo import yahoo_adapter
from app.services.feature_engineering.pipeline import feature_pipeline
logger = logging.getLogger(__name__)
_settings = get_settings()
RISK_FREE_RATE = _settings.risk_free_rate
TRADING_DAYS = _settings.trading_days_per_year
class BacktestEngine:
"""Event-driven portfolio backtester."""
async def run_backtest(
self,
strategy_config: Dict[str, Any],
start_date: date,
end_date: date,
initial_capital: float = 1_000_000.0,
commission_pct: float = 0.001,
slippage_pct: float = 0.0005,
benchmark_ticker: str = "SPY",
rebalance_frequency: str = "monthly",
) -> Dict[str, Any]:
"""
Run a complete backtest simulation.
Returns:
Dict with metrics, equity curve, trades, and monthly returns.
"""
universe = strategy_config.get("universe", [])
if not universe:
return {"status": "failed", "error": "Empty universe"}
# 1. Fetch historical data
price_frames: Dict[str, pd.DataFrame] = {}
for ticker in universe:
df = await yahoo_adapter.get_price_dataframe(
ticker, period="max"
)
if not df.empty:
# Filter by date range
df.index = pd.to_datetime(df.index)
mask = (df.index >= pd.Timestamp(start_date)) & (
df.index <= pd.Timestamp(end_date)
)
filtered = df.loc[mask]
if not filtered.empty:
price_frames[ticker] = filtered
if not price_frames:
return {"status": "failed", "error": "No price data available for universe"}
# Fetch benchmark
bench_df = await yahoo_adapter.get_price_dataframe(benchmark_ticker, period="max")
bench_prices = None
if not bench_df.empty:
bench_df.index = pd.to_datetime(bench_df.index)
mask = (bench_df.index >= pd.Timestamp(start_date)) & (
bench_df.index <= pd.Timestamp(end_date)
)
bench_filtered = bench_df.loc[mask]
if not bench_filtered.empty:
bench_prices = bench_filtered["Close"]
# 2. Build aligned price matrix
close_prices = pd.DataFrame(
{t: df["Close"] for t, df in price_frames.items()}
).dropna()
if close_prices.empty:
return {"status": "failed", "error": "No overlapping price data"}
# 3. Compute features for signal generation
featured_data: Dict[str, pd.DataFrame] = {}
for ticker, df in price_frames.items():
featured_data[ticker] = feature_pipeline.compute_all_features(df)
# 4. Run simulation
dates = close_prices.index.tolist()
portfolio_value = initial_capital
cash = initial_capital
positions: Dict[str, float] = {} # ticker -> shares
weights: Dict[str, float] = {}
equity_curve: List[Dict[str, Any]] = []
trades: List[Dict[str, Any]] = []
total_commission = 0.0
total_slippage = 0.0
# Determine rebalance dates
rebal_dates = self._get_rebalance_dates(dates, rebalance_frequency)
# Strategy parameters
entry_rules = strategy_config.get("entry_rules", [])
exit_rules = strategy_config.get("exit_rules", [])
sizing_method = strategy_config.get("position_sizing", {}).get("method", "equal_weight")
max_position = strategy_config.get("position_sizing", {}).get("max_position_pct", 0.15)
stop_loss = strategy_config.get("risk_management", {}).get("stop_loss_pct", 0.05)
take_profit = strategy_config.get("risk_management", {}).get("take_profit_pct", 0.20)
entry_prices: Dict[str, float] = {}
for i, dt in enumerate(dates):
day_prices = close_prices.loc[dt]
# Update portfolio value
position_value = sum(
positions.get(t, 0) * day_prices.get(t, 0)
for t in positions
if t in day_prices.index
)
portfolio_value = cash + position_value
# Record equity curve point
prev_value = equity_curve[-1]["portfolio_value"] if equity_curve else initial_capital
daily_ret = (portfolio_value / prev_value - 1) if prev_value > 0 else 0
bench_val = None
if bench_prices is not None and dt in bench_prices.index:
bench_val = float(bench_prices.loc[dt])
equity_curve.append({
"date": dt.strftime("%Y-%m-%d"),
"portfolio_value": round(portfolio_value, 2),
"benchmark_value": bench_val,
"daily_return": round(daily_ret, 6),
})
# Check stop loss / take profit for existing positions
for ticker in list(positions.keys()):
if ticker not in day_prices.index or positions[ticker] == 0:
continue
current_price = day_prices[ticker]
entry_price = entry_prices.get(ticker, current_price)
pnl_pct = (current_price / entry_price - 1) if entry_price > 0 else 0
should_exit = False
exit_reason = ""
if pnl_pct <= -stop_loss:
should_exit = True
exit_reason = "stop_loss"
elif pnl_pct >= take_profit:
should_exit = True
exit_reason = "take_profit"
if should_exit:
shares = positions[ticker]
trade_value = shares * current_price
comm = trade_value * commission_pct
slip = trade_value * slippage_pct
total_commission += comm
total_slippage += slip
cash += trade_value - comm - slip
pnl = (current_price - entry_price) * shares
trades.append({
"date": dt.strftime("%Y-%m-%d"),
"ticker": ticker,
"action": "sell",
"quantity": round(shares, 2),
"price": round(current_price, 2),
"commission": round(comm, 2),
"slippage": round(slip, 2),
"pnl": round(pnl, 2),
"reason": exit_reason,
})
del positions[ticker]
del entry_prices[ticker]
# Rebalance on schedule
if dt in rebal_dates:
# Generate new target weights
target_weights = self._compute_target_weights(
universe, day_prices, featured_data, dt, sizing_method, max_position
)
# Execute trades to reach target
for ticker, target_weight in target_weights.items():
if ticker not in day_prices.index:
continue
price = day_prices[ticker]
if price <= 0:
continue
target_value = portfolio_value * target_weight
current_shares = positions.get(ticker, 0)
current_value = current_shares * price
delta_value = target_value - current_value
if abs(delta_value) < portfolio_value * 0.005:
continue # Skip tiny trades
delta_shares = delta_value / price
trade_value = abs(delta_value)
comm = trade_value * commission_pct
slip = trade_value * slippage_pct
total_commission += comm
total_slippage += slip
if delta_shares > 0:
cash -= (trade_value + comm + slip)
action = "buy"
entry_prices[ticker] = price
else:
cash += (trade_value - comm - slip)
action = "sell"
new_shares = current_shares + delta_shares
if abs(new_shares) < 0.01:
positions.pop(ticker, None)
entry_prices.pop(ticker, None)
else:
positions[ticker] = new_shares
trades.append({
"date": dt.strftime("%Y-%m-%d"),
"ticker": ticker,
"action": action,
"quantity": round(abs(delta_shares), 2),
"price": round(price, 2),
"commission": round(comm, 2),
"slippage": round(slip, 2),
"pnl": None,
})
# 5. Compute final metrics
equity_values = [p["portfolio_value"] for p in equity_curve]
daily_returns = [p["daily_return"] for p in equity_curve if p["daily_return"] is not None]
metrics = self._compute_backtest_metrics(
equity_values,
daily_returns,
trades,
initial_capital,
total_commission,
total_slippage,
bench_prices,
)
# Monthly returns
monthly_returns = self._compute_monthly_returns(equity_curve)
# Drawdown series — O(n) running peak
running_peak = 0.0
for point in equity_curve:
running_peak = max(running_peak, point["portfolio_value"])
point["drawdown"] = round(
(point["portfolio_value"] - running_peak) / running_peak, 6
) if running_peak > 0 else 0
return {
"status": "completed",
"start_date": start_date.isoformat(),
"end_date": end_date.isoformat(),
"initial_capital": initial_capital,
"final_value": round(equity_values[-1], 2) if equity_values else initial_capital,
"metrics": metrics,
"equity_curve": equity_curve,
"trades": trades,
"monthly_returns": monthly_returns,
}
def _compute_target_weights(
self,
universe: List[str],
prices: pd.Series,
featured_data: Dict[str, pd.DataFrame],
current_date: pd.Timestamp,
method: str,
max_weight: float,
) -> Dict[str, float]:
"""Compute target portfolio weights for a rebalance date."""
valid_tickers = [t for t in universe if t in prices.index and prices[t] > 0]
if not valid_tickers:
return {}
n = len(valid_tickers)
if method == "equal_weight":
w = min(1.0 / n, max_weight)
return {t: w for t in valid_tickers}
elif method == "score_weighted":
# Use momentum as weight proxy
scores = {}
for t in valid_tickers:
if t in featured_data and "momentum_10" in featured_data[t].columns:
feat_df = featured_data[t]
mask = feat_df.index <= current_date
if mask.any():
mom = feat_df.loc[mask, "momentum_10"].iloc[-1]
scores[t] = max(float(mom) if pd.notna(mom) else 0, 0)
else:
scores[t] = 0
else:
scores[t] = 0
total = sum(scores.values()) or 1.0
return {t: min(s / total, max_weight) for t, s in scores.items()}
else:
w = min(1.0 / n, max_weight)
return {t: w for t in valid_tickers}
@staticmethod
def _get_rebalance_dates(
dates: List[pd.Timestamp],
frequency: str,
) -> set:
"""Determine which dates are rebalance dates."""
if not dates:
return set()
rebal = set()
rebal.add(dates[0]) # Always rebalance on first date
if frequency == "daily":
return set(dates)
elif frequency == "weekly":
for d in dates:
if d.weekday() == 0: # Monday
rebal.add(d)
elif frequency == "monthly":
current_month = dates[0].month
for d in dates:
if d.month != current_month:
rebal.add(d)
current_month = d.month
elif frequency == "quarterly":
current_quarter = (dates[0].month - 1) // 3
for d in dates:
q = (d.month - 1) // 3
if q != current_quarter:
rebal.add(d)
current_quarter = q
return rebal
@staticmethod
def _compute_backtest_metrics(
equity_values: List[float],
daily_returns: List[float],
trades: List[Dict],
initial_capital: float,
total_commission: float,
total_slippage: float,
bench_prices: Optional[pd.Series],
) -> Dict[str, Any]:
"""Compute comprehensive backtest metrics."""
if not equity_values or not daily_returns:
return {}
returns = np.array(daily_returns)
final = equity_values[-1]
n_days = len(returns)
n_years = n_days / TRADING_DAYS
total_return = (final / initial_capital - 1)
ann_return = (1 + total_return) ** (1 / n_years) - 1 if n_years > 0 else 0
ann_vol = float(np.std(returns, ddof=1) * np.sqrt(TRADING_DAYS)) if len(returns) > 1 else 0
sharpe = (ann_return - RISK_FREE_RATE) / ann_vol if ann_vol > 0 else 0
# Sortino
downside = returns[returns < 0]
down_dev = float(np.std(downside, ddof=1) * np.sqrt(TRADING_DAYS)) if len(downside) > 1 else ann_vol
sortino = (ann_return - RISK_FREE_RATE) / down_dev if down_dev > 0 else 0
# Max drawdown
cum = np.cumprod(1 + returns)
peak = np.maximum.accumulate(cum)
dd = (cum - peak) / peak
max_dd = float(np.min(dd)) if len(dd) > 0 else 0
# Calmar
calmar = ann_return / abs(max_dd) if max_dd != 0 else 0
# Win rate and profit factor
trade_pnls = [t.get("pnl", 0) for t in trades if t.get("pnl") is not None]
wins = [p for p in trade_pnls if p > 0]
losses = [p for p in trade_pnls if p < 0]
win_rate = len(wins) / len(trade_pnls) if trade_pnls else 0
profit_factor = (
sum(wins) / abs(sum(losses)) if losses else float("inf") if wins else 0
)
avg_trade_return = float(np.mean(trade_pnls)) if trade_pnls else 0
metrics = {
"total_return": round(total_return, 4),
"annualized_return": round(ann_return, 4),
"sharpe_ratio": round(sharpe, 4),
"sortino_ratio": round(sortino, 4),
"max_drawdown": round(max_dd, 4),
"volatility": round(ann_vol, 4),
"calmar_ratio": round(calmar, 4),
"win_rate": round(win_rate, 4),
"profit_factor": round(profit_factor, 4) if profit_factor != float("inf") else None,
"total_trades": len(trades),
"avg_trade_return": round(avg_trade_return, 2),
"total_commission": round(total_commission, 2),
"total_slippage": round(total_slippage, 2),
}
# Alpha/beta vs benchmark
if bench_prices is not None and len(bench_prices) > 10:
bench_ret = bench_prices.pct_change().dropna().values
min_len = min(len(returns), len(bench_ret))
if min_len > 10:
r = returns[:min_len]
b = bench_ret[:min_len]
cov_rb = np.cov(r, b)[0, 1]
var_b = np.var(b, ddof=1)
beta = cov_rb / var_b if var_b > 0 else 1.0
alpha = ann_return - beta * float(np.mean(b) * TRADING_DAYS)
metrics["beta"] = round(float(beta), 4)
metrics["alpha"] = round(float(alpha), 4)
return metrics
@staticmethod
def _compute_monthly_returns(
equity_curve: List[Dict[str, Any]],
) -> Dict[str, float]:
"""Compute monthly returns from equity curve."""
if not equity_curve:
return {}
monthly: Dict[str, float] = {}
prev_value = equity_curve[0]["portfolio_value"]
current_month = equity_curve[0]["date"][:7]
for point in equity_curve:
month = point["date"][:7]
if month != current_month:
monthly[current_month] = round(
(point["portfolio_value"] / prev_value - 1), 4
)
prev_value = point["portfolio_value"]
current_month = month
# Final month
if equity_curve:
monthly[current_month] = round(
(equity_curve[-1]["portfolio_value"] / prev_value - 1), 4
)
return monthly
backtest_engine = BacktestEngine()