ICT Turtle Soup
Backtest

"""
Turtle Soup ICT — Parameter Optimisation + Out-of-Sample Test
==============================================================
1. Load EURUSD daily data (real via yfinance or synthetic)
2. Split: Years 1-2 = IN-SAMPLE | Year 3 = OUT-OF-SAMPLE
3. Grid search over lookback, R:R ratio, ATR multiplier
4. Rank by Sharpe ratio on in-sample data
5. Validate best params on out-of-sample data
6. Measure overfitting via Sharpe decay
"""

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from itertools import product

# ── SIGNAL LOGIC ─────────────────────────────────────────
def prepare(df, lookback=20):
    d = df.copy()
    d['turtle_high'] = d['High'].shift(1).rolling(lookback).max()
    d['turtle_low']  = d['Low'].shift(1).rolling(lookback).min()
    hl  = d['High'] - d['Low']
    hpc = (d['High'] - d['Close'].shift(1)).abs()
    lpc = (d['Low']  - d['Close'].shift(1)).abs()
    d['ATR'] = pd.concat([hl, hpc, lpc], axis=1).max(axis=1).rolling(14).mean()
    # Turtle Soup: sweep the level then close back inside
    d['long_signal']  = (d['Low']  < d['turtle_low'])  & (d['Close'] > d['turtle_low'])
    d['short_signal'] = (d['High'] > d['turtle_high']) & (d['Close'] < d['turtle_high'])
    return d

# ── BACKTEST ENGINE ───────────────────────────────────────
def run_backtest(df, rr_ratio=2.0, atr_mult=1.0, risk=0.01, capital=10_000):
    rows = df.reset_index()
    equity = [capital]; trades = []; in_trade = False
    entry_price = sl = tp = direction = None

    for i in range(1, len(rows)):
        row = rows.iloc[i]; prev = rows.iloc[i - 1]

        if in_trade:
            hit_tp = hit_sl = False
            if direction == 'long':
                if row['Low']  <= sl: hit_sl = True
                if row['High'] >= tp: hit_tp = True
            else:
                if row['High'] >= sl: hit_sl = True
                if row['Low']  <= tp: hit_tp = True
            if hit_tp or hit_sl:
                exit_px = tp if hit_tp else sl
                pnl_pips = ((exit_px - entry_price) if direction == 'long'
                            else (entry_price - exit_px)) * 10_000
                stop_pips = abs(entry_price - sl) * 10_000
                if stop_pips > 0:
                    capital *= (1 + risk * (pnl_pips / stop_pips))
                trades.append({'result': 'win' if hit_tp else 'loss',
                               'pnl_pips': round(pnl_pips, 1)})
                in_trade = False

        if not in_trade:
            atr = prev['ATR']
            if pd.isna(atr) or atr == 0:
                equity.append(capital); continue
            stop_dist = atr * atr_mult
            if prev['long_signal']:
                direction = 'long'; entry_price = row['Open']
                sl = entry_price - stop_dist; tp = entry_price + stop_dist * rr_ratio
                in_trade = True
            elif prev['short_signal']:
                direction = 'short'; entry_price = row['Open']
                sl = entry_price + stop_dist; tp = entry_price - stop_dist * rr_ratio
                in_trade = True
        equity.append(capital)

    eq = pd.Series(equity, index=df.index)
    td = pd.DataFrame(trades) if trades else pd.DataFrame(columns=['result','pnl_pips'])
    return eq, td