meta / engine /simulator.py
BharathPESU's picture
Upload folder using huggingface_hub
9d82314 verified
Raw History Blame Contribute Delete
4.63 kB
"""Intraday trade simulator. Pure functions: no I/O.
Rules (see docs/STRATEGY.md sections 6-8):
- Entry at the open of the first bar at or after entry_time on entry_date.
- Stop is stop_loss_pct from the entry fill. Gap-through fills at the bar open.
- Forced exit at the close of the last bar at or before square_off_time.
- Slippage is applied against the trader on each fill.
- Gross P&L uses raw prices with no costs. Net P&L uses slipped prices minus fees.
- Max positions is applied per entry date. All entries share entry_time and all
positions close by square-off, so this is exact for intraday trading.
"""
from __future__ import annotations
from collections import Counter
from datetime import date
from typing import Any
import pandas as pd
from engine.config import BacktestConfig
from engine.costs import round_trip_cost
from engine.sizing import position_quantity
TRADE_COLUMNS = [
"symbol", "side", "entry_date", "entry_time", "entry_price",
"exit_time", "exit_price", "exit_reason", "qty",
"gross_pnl", "fees", "net_pnl", "surprise",
]
def _session_bars(bars: pd.DataFrame, day: date, cfg: BacktestConfig) -> pd.DataFrame | None:
day_bars = bars[bars.index.date == day]
if day_bars.empty:
return None
start_times = day_bars.index.time
end_times = (day_bars.index + pd.Timedelta(cfg.bar_interval)).time
sess = day_bars[(start_times >= cfg.entry_time) & (end_times <= cfg.square_off_time)]
return sess if not sess.empty else None
def simulate_trade(signal: dict[str, Any], bars: pd.DataFrame, cfg: BacktestConfig) -> dict[str, Any] | None:
"""Simulate one signal on one day. Returns a trade record, or None if not tradeable."""
sess = _session_bars(bars, signal["entry_date"], cfg)
if sess is None:
return None
is_long = signal["side"] == "long"
sign = 1 if is_long else -1
slip = cfg.costs.slippage_pct
raw_entry = float(sess.iloc[0].open)
entry_px = raw_entry * (1 + slip) if is_long else raw_entry * (1 - slip)
qty = position_quantity(cfg.capital, entry_px, cfg)
if qty <= 0:
return None
stop_px = entry_px * (1 - cfg.stop_loss_pct) if is_long else entry_px * (1 + cfg.stop_loss_pct)
exit_ts = sess.index[-1]
raw_exit = float(sess.iloc[-1].close)
reason = "square_off"
for ts, bar in sess.iterrows():
if is_long and bar.low <= stop_px:
raw_exit = min(float(bar.open), stop_px)
exit_ts, reason = ts, "stop"
break
if not is_long and bar.high >= stop_px:
raw_exit = max(float(bar.open), stop_px)
exit_ts, reason = ts, "stop"
break
exit_px = raw_exit * (1 - slip) if is_long else raw_exit * (1 + slip)
gross = sign * (raw_exit - raw_entry) * qty
slipped = sign * (exit_px - entry_px) * qty
buy_value = qty * (entry_px if is_long else exit_px)
sell_value = qty * (exit_px if is_long else entry_px)
fees = round_trip_cost(buy_value, sell_value, cfg.costs)
return {
"symbol": signal["symbol"],
"side": signal["side"],
"entry_date": signal["entry_date"],
"entry_time": sess.index[0].time(),
"entry_price": entry_px,
"exit_time": exit_ts.time(),
"exit_price": exit_px,
"exit_reason": reason,
"qty": qty,
"gross_pnl": gross,
"fees": fees,
"net_pnl": slipped - fees,
"surprise": signal.get("surprise"),
}
def run_backtest(
signals: pd.DataFrame,
bars_by_symbol: dict[str, pd.DataFrame],
cfg: BacktestConfig,
) -> pd.DataFrame:
"""Run all signals. Signals need columns: symbol, entry_date, side, announced_at, surprise."""
if signals.empty:
return pd.DataFrame(columns=TRADE_COLUMNS)
trades: list[dict[str, Any]] = []
per_day: Counter[date] = Counter()
seen: set[tuple[str, date]] = set()
ordered = signals.sort_values(["entry_date", "announced_at"])
for sig in ordered.to_dict("records"):
side_enabled = cfg.enable_long if sig["side"] == "long" else cfg.enable_short
if not side_enabled:
continue
key = (sig["symbol"], sig["entry_date"])
if key in seen or per_day[sig["entry_date"]] >= cfg.max_positions:
continue
bars = bars_by_symbol.get(sig["symbol"])
if bars is None:
continue
trade = simulate_trade(sig, bars, cfg)
if trade is None:
continue
seen.add(key)
per_day[sig["entry_date"]] += 1
trades.append(trade)
return pd.DataFrame(trades, columns=TRADE_COLUMNS)