moen0
1021 lines
39 KiB
Python
1021 lines
39 KiB
Python
import os
|
|
import sys
|
|
import json
|
|
import time
|
|
import random
|
|
from functools import lru_cache
|
|
from itertools import islice, product
|
|
from typing import Literal, Optional
|
|
from fastapi import FastAPI, HTTPException, Query
|
|
from fastapi.responses import StreamingResponse
|
|
from fastapi.middleware.cors import CORSMiddleware
|
|
from pydantic import BaseModel, Field
|
|
|
|
BACKEND_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
|
|
if BACKEND_DIR not in sys.path:
|
|
sys.path.insert(0, BACKEND_DIR)
|
|
|
|
from data.loader import load_candles, resample_candles
|
|
from engine.backtester import run_backtest, run_backtest_stream
|
|
from engine.monte_carlo import run_monte_carlo
|
|
from indicators.fvg import find_fvgs
|
|
from indicators.liquidity import find_liquidity_levels
|
|
from indicators.market_structure import detect_structure, find_swing_points
|
|
from indicators.order_blocks import find_order_blocks
|
|
from strategies.ict_strategy import ICTStrategy
|
|
|
|
DEFAULT_DATASET = "data.csv"
|
|
DATA_DIR = os.path.join(BACKEND_DIR, "data")
|
|
|
|
app = FastAPI()
|
|
|
|
app.add_middleware(
|
|
CORSMiddleware,
|
|
allow_origins=["*"],
|
|
allow_credentials=True,
|
|
allow_methods=["*"],
|
|
allow_headers=["*"],
|
|
)
|
|
|
|
|
|
def _list_csv_datasets():
|
|
if not os.path.isdir(DATA_DIR):
|
|
return []
|
|
return sorted([name for name in os.listdir(DATA_DIR) if name.endswith(".csv")])
|
|
|
|
|
|
def _resolve_dataset(dataset):
|
|
csvs = _list_csv_datasets()
|
|
if not csvs:
|
|
raise HTTPException(status_code=500, detail="No CSV datasets found in backend/data")
|
|
|
|
requested = dataset or DEFAULT_DATASET
|
|
if requested in csvs:
|
|
return requested
|
|
|
|
if requested == DEFAULT_DATASET:
|
|
return csvs[0]
|
|
|
|
raise HTTPException(status_code=400, detail=f"Dataset '{requested}' not found")
|
|
|
|
|
|
def _parse_day_filter(day_filter):
|
|
"""Parse comma separated day ints '0,1,2,3,4' into list. None or empty = all days."""
|
|
if not day_filter:
|
|
return None
|
|
try:
|
|
days = [int(d.strip()) for d in day_filter.split(",") if d.strip()]
|
|
valid = [d for d in days if 0 <= d <= 4]
|
|
return valid if valid else None
|
|
except ValueError:
|
|
return None
|
|
|
|
|
|
def _parse_int_list(value, fallback):
|
|
if not value:
|
|
return fallback
|
|
try:
|
|
parsed = [int(part.strip()) for part in value.split(",") if part.strip()]
|
|
except ValueError:
|
|
return fallback
|
|
return parsed or fallback
|
|
|
|
|
|
def _parse_float_list(value, fallback):
|
|
if not value:
|
|
return fallback
|
|
try:
|
|
parsed = [float(part.strip()) for part in value.split(",") if part.strip()]
|
|
except ValueError:
|
|
return fallback
|
|
return parsed or fallback
|
|
|
|
|
|
def _parse_bool_list(value, fallback):
|
|
if not value:
|
|
return fallback
|
|
parsed = []
|
|
for part in value.split(","):
|
|
token = part.strip().lower()
|
|
if token in {"true", "1", "yes", "y"}:
|
|
parsed.append(True)
|
|
elif token in {"false", "0", "no", "n"}:
|
|
parsed.append(False)
|
|
return parsed or fallback
|
|
|
|
|
|
def _parse_str_list(value, fallback):
|
|
if not value:
|
|
return fallback
|
|
parsed = [part.strip() for part in value.split(",") if part.strip()]
|
|
return parsed or fallback
|
|
|
|
|
|
def _dataset_path(dataset_id):
|
|
return os.path.join(DATA_DIR, dataset_id)
|
|
|
|
|
|
def _dataset_mtime(dataset_id):
|
|
path = _dataset_path(dataset_id)
|
|
try:
|
|
return os.path.getmtime(path)
|
|
except OSError as exc:
|
|
raise HTTPException(status_code=400, detail=f"Dataset '{dataset_id}' not found") from exc
|
|
|
|
|
|
@lru_cache(maxsize=32)
|
|
def _load_candles_cached(dataset_id, dataset_mtime):
|
|
candles = load_candles(f"data/{dataset_id}")
|
|
return tuple(candles)
|
|
|
|
|
|
@lru_cache(maxsize=128)
|
|
def _resample_candles_cached(dataset_id, timeframe, dataset_mtime):
|
|
candles_1m = _load_candles_cached(dataset_id, dataset_mtime)
|
|
candles = resample_candles(candles_1m, period=timeframe)
|
|
return tuple(candles)
|
|
|
|
|
|
def _get_candles_for_timeframe(dataset_id, timeframe):
|
|
return _resample_candles_cached(dataset_id, int(timeframe), _dataset_mtime(dataset_id))
|
|
|
|
|
|
def _build_strategy(
|
|
session, lookback, ob_age, atr_mult, use_fvg, use_ob,
|
|
proximity_pct, sweep, sweep_lookback,
|
|
min_gap_size, impulse_multiplier, require_unmitigated_fvg,
|
|
require_bos_confluence, min_ob_size, require_fvg_ob_confluence,
|
|
asian_sweep_only, day_filter,
|
|
use_break_even=False, be_trigger_rr=1.0,
|
|
use_partial_tp=False, partial_tp_rr=1.0, partial_tp_percent=50.0,
|
|
):
|
|
return ICTStrategy(
|
|
session=session,
|
|
lookback=lookback,
|
|
ob_max_age=ob_age,
|
|
atr_mult=atr_mult,
|
|
use_fvg=use_fvg,
|
|
use_ob=use_ob,
|
|
proximity_pct=proximity_pct,
|
|
use_liquidity_sweep=sweep,
|
|
sweep_lookback=sweep_lookback,
|
|
min_gap_size=min_gap_size,
|
|
impulse_multiplier=impulse_multiplier,
|
|
require_unmitigated_fvg=require_unmitigated_fvg,
|
|
require_bos_confluence=require_bos_confluence,
|
|
min_ob_size=min_ob_size,
|
|
require_fvg_ob_confluence=require_fvg_ob_confluence,
|
|
asian_sweep_only=asian_sweep_only,
|
|
day_filter=_parse_day_filter(day_filter),
|
|
use_break_even=use_break_even,
|
|
be_trigger_rr=be_trigger_rr,
|
|
use_partial_tp=use_partial_tp,
|
|
partial_tp_rr=partial_tp_rr,
|
|
partial_tp_percent=partial_tp_percent,
|
|
)
|
|
|
|
|
|
def _trade_payload(trade):
|
|
return {
|
|
"enter_time": trade.enter_time.isoformat(),
|
|
"exit_time": trade.exit_time.isoformat(),
|
|
"enter_price": trade.enter_price,
|
|
"exit_price": trade.exit_price,
|
|
"direction": trade.direction,
|
|
"pnl": trade.pnl,
|
|
"r_multiple": getattr(trade, "r_multiple", 0.0),
|
|
"partial_tp_taken": getattr(trade, "partial_tp_taken", False),
|
|
"partial_tp_realized_pnl": getattr(trade, "partial_tp_realized_pnl", 0.0),
|
|
}
|
|
|
|
|
|
def _stats_payload(trades, rr):
|
|
total_pnl = sum(t.pnl for t in trades)
|
|
winners = [t for t in trades if t.pnl > 0]
|
|
losers = [t for t in trades if t.pnl <= 0]
|
|
partial_tp_trades = [t for t in trades if getattr(t, "partial_tp_taken", False)]
|
|
partial_tp_realized_total = sum(float(getattr(t, "partial_tp_realized_pnl", 0.0) or 0.0) for t in partial_tp_trades)
|
|
return {
|
|
"total_trades": len(trades),
|
|
"winners": len(winners),
|
|
"losers": len(losers),
|
|
"win_rate": len(winners) / len(trades) * 100 if trades else 0,
|
|
"total_pnl": total_pnl,
|
|
"avg_win": sum(t.pnl for t in winners) / len(winners) if winners else 0,
|
|
"avg_loss": sum(t.pnl for t in losers) / len(losers) if losers else 0,
|
|
"risk_reward": rr,
|
|
"partial_tp_trades": len(partial_tp_trades),
|
|
"partial_tp_rate": (len(partial_tp_trades) / len(trades) * 100) if trades else 0,
|
|
"partial_tp_realized_total": partial_tp_realized_total,
|
|
"partial_tp_realized_avg": (partial_tp_realized_total / len(partial_tp_trades)) if partial_tp_trades else 0,
|
|
}
|
|
|
|
|
|
def _build_monte_carlo_payload(trade_r_multiples, *, runs, starting_balance, risk_per_trade_pct, sampling_method,
|
|
missed_trade_pct, pnl_variation_pct, price_noise_pct, slippage_per_trade,
|
|
spread_per_trade, ruin_drawdown_pct, seed=None, base=None):
|
|
monte_carlo = run_monte_carlo(
|
|
trade_r_multiples=trade_r_multiples,
|
|
runs=runs,
|
|
starting_balance=starting_balance,
|
|
risk_per_trade_pct=risk_per_trade_pct,
|
|
sampling_method=sampling_method,
|
|
missed_trade_pct=missed_trade_pct,
|
|
pnl_variation_pct=pnl_variation_pct,
|
|
price_noise_pct=price_noise_pct,
|
|
slippage_per_trade=slippage_per_trade,
|
|
spread_per_trade=spread_per_trade,
|
|
ruin_drawdown_pct=ruin_drawdown_pct,
|
|
seed=seed,
|
|
)
|
|
|
|
summary = dict(monte_carlo.get("summary", {}))
|
|
distribution = monte_carlo.get("distribution", [])
|
|
sample_runs = monte_carlo.get("sample_runs", distribution)
|
|
return {
|
|
**monte_carlo,
|
|
"summary": summary,
|
|
"distribution": distribution,
|
|
"sample_runs": sample_runs,
|
|
"base": base or None,
|
|
}
|
|
|
|
|
|
class MonteCarloRequest(BaseModel):
|
|
trade_r_multiples: list[float] = Field(default_factory=list, min_length=1)
|
|
runs: int = Field(default=500, ge=1, le=20000)
|
|
starting_balance: float = Field(default=10000.0, gt=0)
|
|
risk_per_trade_pct: float = Field(default=1.0, ge=0.0, le=100.0)
|
|
sampling_method: Literal["bootstrap", "shuffle"] = "bootstrap"
|
|
missed_trade_pct: float = Field(default=5.0, ge=0.0, le=100.0)
|
|
pnl_variation_pct: float = Field(default=15.0, ge=0.0, le=100.0)
|
|
price_noise_pct: float = Field(default=0.0, ge=0.0, le=100.0)
|
|
slippage_per_trade: float = Field(default=0.0, ge=0.0)
|
|
spread_per_trade: float = Field(default=0.0, ge=0.0)
|
|
ruin_drawdown_pct: float = Field(default=20.0, ge=0.0, le=100.0)
|
|
seed: Optional[int] = None
|
|
base_trade_count: Optional[int] = None
|
|
base_net_pnl: Optional[float] = None
|
|
base_win_rate: Optional[float] = None
|
|
base_profit_factor: Optional[float] = None
|
|
base_max_drawdown_pct: Optional[float] = None
|
|
|
|
|
|
def _build_equity_points(trades, starting_balance=10000.0):
|
|
equity = starting_balance
|
|
points = [starting_balance]
|
|
for trade in sorted(trades, key=lambda t: t.exit_time):
|
|
equity += trade.pnl
|
|
points.append(equity)
|
|
return points
|
|
|
|
|
|
def _risk_metrics(trades, starting_balance=10000.0):
|
|
if not trades:
|
|
return {
|
|
"net_pnl": 0.0,
|
|
"win_rate": 0.0,
|
|
"profit_factor": 0.0,
|
|
"sharpe_ratio": 0.0,
|
|
"sortino_ratio": 0.0,
|
|
"max_drawdown_pct": 0.0,
|
|
"calmar_ratio": 0.0,
|
|
"recovery_ratio": 0.0,
|
|
"trade_count": 0,
|
|
"trade_score": 0.0,
|
|
"pf_x_wr": 0.0,
|
|
"custom_fitness": 0.0,
|
|
}
|
|
|
|
pnls = [t.pnl for t in trades]
|
|
winners = [p for p in pnls if p > 0]
|
|
losers = [p for p in pnls if p < 0]
|
|
|
|
trade_count = len(trades)
|
|
net_pnl = sum(pnls)
|
|
win_rate = (len(winners) / trade_count) * 100 if trade_count else 0.0
|
|
|
|
gross_profit = sum(winners)
|
|
gross_loss = abs(sum(losers))
|
|
profit_factor = (gross_profit / gross_loss) if gross_loss > 0 else (999.0 if gross_profit > 0 else 0.0)
|
|
|
|
mean_pnl = net_pnl / trade_count if trade_count else 0.0
|
|
variance = sum((p - mean_pnl) ** 2 for p in pnls) / trade_count if trade_count else 0.0
|
|
std_dev = variance ** 0.5
|
|
sharpe = (mean_pnl / std_dev) * (trade_count ** 0.5) if std_dev > 0 else 0.0
|
|
|
|
downside = [min(0.0, p - mean_pnl) for p in pnls]
|
|
downside_variance = (sum(d * d for d in downside) / trade_count) if trade_count else 0.0
|
|
downside_dev = downside_variance ** 0.5
|
|
sortino = (mean_pnl / downside_dev) * (trade_count ** 0.5) if downside_dev > 0 else 0.0
|
|
|
|
equity_points = _build_equity_points(trades, starting_balance=starting_balance)
|
|
peak = equity_points[0]
|
|
max_drawdown_pct = 0.0
|
|
for value in equity_points:
|
|
if value > peak:
|
|
peak = value
|
|
drawdown_pct = ((peak - value) / peak) * 100 if peak > 0 else 0.0
|
|
if drawdown_pct > max_drawdown_pct:
|
|
max_drawdown_pct = drawdown_pct
|
|
|
|
calmar = (net_pnl / max_drawdown_pct) if max_drawdown_pct > 0 else 0.0
|
|
recovery = (net_pnl / max_drawdown_pct) if max_drawdown_pct > 0 else 0.0
|
|
|
|
if trade_count < 80:
|
|
trade_score = max(0.0, trade_count / 80)
|
|
elif trade_count > 400:
|
|
trade_score = max(0.0, 400 / trade_count)
|
|
else:
|
|
trade_score = 1.0
|
|
|
|
pf_x_wr = profit_factor * (win_rate / 100)
|
|
custom_fitness = pf_x_wr * trade_score
|
|
|
|
return {
|
|
"net_pnl": round(net_pnl, 6),
|
|
"win_rate": round(win_rate, 4),
|
|
"profit_factor": round(profit_factor, 6),
|
|
"sharpe_ratio": round(sharpe, 6),
|
|
"sortino_ratio": round(sortino, 6),
|
|
"max_drawdown_pct": round(max_drawdown_pct, 6),
|
|
"calmar_ratio": round(calmar, 6),
|
|
"recovery_ratio": round(recovery, 6),
|
|
"trade_count": trade_count,
|
|
"trade_score": round(trade_score, 6),
|
|
"pf_x_wr": round(pf_x_wr, 6),
|
|
"custom_fitness": round(custom_fitness, 6),
|
|
}
|
|
|
|
|
|
def _dominates(a, b):
|
|
maximize_keys = [
|
|
"net_pnl",
|
|
"sharpe_ratio",
|
|
"sortino_ratio",
|
|
"profit_factor",
|
|
"calmar_ratio",
|
|
"recovery_ratio",
|
|
"win_rate",
|
|
"trade_score",
|
|
"pf_x_wr",
|
|
"custom_fitness",
|
|
]
|
|
no_worse = all(a.get(key, 0) >= b.get(key, 0) for key in maximize_keys) and a.get("max_drawdown_pct", 0) <= b.get("max_drawdown_pct", 0)
|
|
strictly_better = any(a.get(key, 0) > b.get(key, 0) for key in maximize_keys) or a.get("max_drawdown_pct", 0) < b.get("max_drawdown_pct", 0)
|
|
return no_worse and strictly_better
|
|
|
|
|
|
def _pareto_front(items):
|
|
front = []
|
|
for candidate in items:
|
|
dominated = False
|
|
for other in items:
|
|
if other is candidate:
|
|
continue
|
|
if _dominates(other, candidate):
|
|
dominated = True
|
|
break
|
|
if not dominated:
|
|
front.append(candidate)
|
|
return front
|
|
|
|
|
|
class OptimizeRequest(BaseModel):
|
|
timeframe: int = Field(default=5, ge=1)
|
|
rr: float = Field(default=2.5, gt=0)
|
|
rr_values: list[float] = Field(default_factory=lambda: [2.5], min_length=1)
|
|
dataset: str = DEFAULT_DATASET
|
|
|
|
min_trades: int = Field(default=5, ge=1)
|
|
max_trades: int = Field(default=1000, ge=1)
|
|
max_combinations: int = Field(default=1000, ge=1, le=50000)
|
|
combo_sampling_mode: Literal["first", "random"] = "random"
|
|
combo_sampling_seed: Optional[int] = None
|
|
top_n: int = Field(default=10, ge=1)
|
|
|
|
sessions: list[str] = Field(default_factory=lambda: ["london", "new_york"], min_length=1)
|
|
lookback_values: list[int] = Field(default_factory=lambda: [3, 5, 7, 10], min_length=1)
|
|
ob_age_values: list[int] = Field(default_factory=lambda: [20, 50, 80], min_length=1)
|
|
atr_values: list[float] = Field(default_factory=lambda: [1.0, 1.5, 2.0, 2.5], min_length=1)
|
|
|
|
min_ob_size_values: list[float] = Field(default_factory=lambda: [0.00030, 0.00040, 0.00050, 0.00060, 0.00080], min_length=1)
|
|
min_gap_size_values: list[float] = Field(default_factory=lambda: [0.00015, 0.00020, 0.00025, 0.00030, 0.00035], min_length=1)
|
|
impulse_multiplier_values: list[float] = Field(default_factory=lambda: [1.15, 1.25, 1.35, 1.45, 1.60], min_length=1)
|
|
|
|
require_unmitigated_fvg_modes: list[bool] = Field(default_factory=lambda: [True, False], min_length=1)
|
|
require_fvg_ob_confluence_modes: list[bool] = Field(default_factory=lambda: [True, False], min_length=1)
|
|
require_bos_confluence_modes: list[bool] = Field(default_factory=lambda: [True, False], min_length=1)
|
|
sweep_modes: list[bool] = Field(default_factory=lambda: [True, False], min_length=1)
|
|
sweep_lb_values: list[int] = Field(default_factory=lambda: [5, 10, 15], min_length=1)
|
|
asian_sweep_only_modes: list[bool] = Field(default_factory=lambda: [True, False], min_length=1)
|
|
use_break_even_modes: list[bool] = Field(default_factory=lambda: [False, True], min_length=1)
|
|
be_trigger_rr_values: list[float] = Field(default_factory=lambda: [1.0, 1.5, 2.0], min_length=1)
|
|
use_partial_tp_modes: list[bool] = Field(default_factory=lambda: [False, True], min_length=1)
|
|
partial_tp_rr_values: list[float] = Field(default_factory=lambda: [1.0, 1.5, 2.0], min_length=1)
|
|
partial_tp_percent_values: list[float] = Field(default_factory=lambda: [50.0], min_length=1)
|
|
|
|
rank_objective: Literal[
|
|
"custom_fitness",
|
|
"net_pnl",
|
|
"sharpe_ratio",
|
|
"sortino_ratio",
|
|
"profit_factor",
|
|
"calmar_ratio",
|
|
"recovery_ratio",
|
|
"win_rate",
|
|
"pf_x_wr",
|
|
"trade_score",
|
|
"trade_count",
|
|
"max_drawdown_pct",
|
|
] = "custom_fitness"
|
|
|
|
|
|
@app.get("/api/datasets")
|
|
def get_datasets():
|
|
csvs = _list_csv_datasets()
|
|
default_id = DEFAULT_DATASET if DEFAULT_DATASET in csvs else (csvs[0] if csvs else "")
|
|
datasets = [
|
|
{
|
|
"id": csv,
|
|
"label": csv.replace(".csv", "").replace("_", " ").title(),
|
|
"default": csv == default_id,
|
|
}
|
|
for csv in csvs
|
|
]
|
|
return {"datasets": datasets}
|
|
|
|
|
|
@app.get("/api/candles")
|
|
def get_candles(timeframe: int = 5, dataset: str = DEFAULT_DATASET):
|
|
dataset_id = _resolve_dataset(dataset)
|
|
candles = _get_candles_for_timeframe(dataset_id, timeframe)
|
|
return {
|
|
"candles": [
|
|
{
|
|
"time": c.time_open.isoformat(),
|
|
"open": c.open,
|
|
"high": c.high,
|
|
"low": c.low,
|
|
"close": c.close,
|
|
}
|
|
for c in candles
|
|
]
|
|
}
|
|
|
|
|
|
@app.get("/api/indicators")
|
|
def get_indicators(timeframe: int = 5, dataset: str = DEFAULT_DATASET):
|
|
dataset_id = _resolve_dataset(dataset)
|
|
candles = _get_candles_for_timeframe(dataset_id, timeframe)
|
|
|
|
swings = find_swing_points(candles)
|
|
structure = detect_structure(swings)
|
|
levels = find_liquidity_levels(swings)
|
|
fvgs = find_fvgs(candles)
|
|
obs = find_order_blocks(candles, structure)
|
|
|
|
candle_times = [c.time_open.isoformat() for c in candles]
|
|
return {
|
|
"candle_times": candle_times,
|
|
"swings": swings,
|
|
"structure": structure,
|
|
"liquidity": levels,
|
|
"fvgs": fvgs,
|
|
"order_blocks": obs,
|
|
}
|
|
|
|
|
|
@app.get("/api/backtest")
|
|
def get_backtest(
|
|
timeframe: int = 5,
|
|
rr: float = 2.5,
|
|
lookback: int = 7,
|
|
ob_age: int = 50,
|
|
atr_mult: float = 2.5,
|
|
sweep: bool = True,
|
|
sweep_lookback: int = 5,
|
|
session: str = "london",
|
|
dataset: str = DEFAULT_DATASET,
|
|
use_fvg: bool = True,
|
|
use_ob: bool = True,
|
|
proximity_pct: float = 0.5,
|
|
# FVG Quality
|
|
min_gap_size: float = 0.0,
|
|
impulse_multiplier: float = 0.0,
|
|
require_unmitigated_fvg: bool = True,
|
|
require_bos_confluence: bool = False,
|
|
# Order Block
|
|
min_ob_size: float = 0.0,
|
|
require_fvg_ob_confluence: bool = False,
|
|
# Liquidity
|
|
asian_sweep_only: bool = False,
|
|
# Break-even
|
|
use_break_even: bool = False,
|
|
be_trigger_rr: float = 1.0,
|
|
# Partial TP
|
|
use_partial_tp: bool = False,
|
|
partial_tp_rr: float = 1.0,
|
|
partial_tp_percent: float = 50.0,
|
|
# Time
|
|
day_filter: Optional[str] = None,
|
|
# Risk Management
|
|
max_daily_loss: float = 0.0,
|
|
max_consecutive_losses: int = 0,
|
|
):
|
|
dataset_id = _resolve_dataset(dataset)
|
|
candles = _get_candles_for_timeframe(dataset_id, timeframe)
|
|
|
|
strategy = _build_strategy(
|
|
session=session, lookback=lookback, ob_age=ob_age, atr_mult=atr_mult,
|
|
use_fvg=use_fvg, use_ob=use_ob, proximity_pct=proximity_pct,
|
|
sweep=sweep, sweep_lookback=sweep_lookback,
|
|
min_gap_size=min_gap_size, impulse_multiplier=impulse_multiplier,
|
|
require_unmitigated_fvg=require_unmitigated_fvg,
|
|
require_bos_confluence=require_bos_confluence,
|
|
min_ob_size=min_ob_size, require_fvg_ob_confluence=require_fvg_ob_confluence,
|
|
asian_sweep_only=asian_sweep_only, day_filter=day_filter,
|
|
use_break_even=use_break_even, be_trigger_rr=be_trigger_rr,
|
|
use_partial_tp=use_partial_tp, partial_tp_rr=partial_tp_rr, partial_tp_percent=partial_tp_percent,
|
|
)
|
|
trades = run_backtest(
|
|
candles, strategy, 10000, risk_reward=rr,
|
|
max_daily_loss=max_daily_loss,
|
|
max_consecutive_losses=max_consecutive_losses,
|
|
)
|
|
|
|
return {
|
|
"trades": [_trade_payload(t) for t in trades],
|
|
"candle_times": [c.time_open.isoformat() for c in candles],
|
|
"stats": _stats_payload(trades, rr),
|
|
}
|
|
|
|
|
|
@app.post("/api/backtest/monte-carlo")
|
|
def backtest_monte_carlo(req: MonteCarloRequest):
|
|
base = {
|
|
"trade_count": req.base_trade_count if req.base_trade_count is not None else len(req.trade_r_multiples),
|
|
"net_pnl": req.base_net_pnl if req.base_net_pnl is not None else 0.0,
|
|
"win_rate": req.base_win_rate if req.base_win_rate is not None else 0.0,
|
|
"profit_factor": req.base_profit_factor if req.base_profit_factor is not None else 0.0,
|
|
"max_drawdown_pct": req.base_max_drawdown_pct if req.base_max_drawdown_pct is not None else 0.0,
|
|
}
|
|
|
|
return _build_monte_carlo_payload(
|
|
req.trade_r_multiples,
|
|
runs=req.runs,
|
|
starting_balance=req.starting_balance,
|
|
risk_per_trade_pct=req.risk_per_trade_pct,
|
|
sampling_method=req.sampling_method,
|
|
missed_trade_pct=req.missed_trade_pct,
|
|
pnl_variation_pct=req.pnl_variation_pct,
|
|
price_noise_pct=req.price_noise_pct,
|
|
slippage_per_trade=req.slippage_per_trade,
|
|
spread_per_trade=req.spread_per_trade,
|
|
ruin_drawdown_pct=req.ruin_drawdown_pct,
|
|
seed=req.seed,
|
|
base=base,
|
|
)
|
|
|
|
|
|
@app.post("/api/optimize")
|
|
def get_optimize(req: OptimizeRequest):
|
|
dataset_id = _resolve_dataset(req.dataset)
|
|
candles = _get_candles_for_timeframe(dataset_id, req.timeframe)
|
|
|
|
session_list = req.sessions
|
|
lookback_list = req.lookback_values
|
|
ob_age_list = req.ob_age_values
|
|
atr_list = req.atr_values
|
|
rr_list = [float(value) for value in req.rr_values if float(value) > 0]
|
|
if not rr_list:
|
|
rr_list = [req.rr]
|
|
min_ob_size_list = req.min_ob_size_values
|
|
min_gap_size_list = req.min_gap_size_values
|
|
impulse_multiplier_list = req.impulse_multiplier_values
|
|
require_unmitigated_fvg_list = req.require_unmitigated_fvg_modes
|
|
require_fvg_ob_confluence_list = req.require_fvg_ob_confluence_modes
|
|
require_bos_confluence_list = req.require_bos_confluence_modes
|
|
sweep_list = req.sweep_modes
|
|
sweep_lb_list = req.sweep_lb_values
|
|
asian_sweep_only_list = req.asian_sweep_only_modes
|
|
use_break_even_list = req.use_break_even_modes
|
|
be_trigger_rr_list = [float(value) for value in req.be_trigger_rr_values if float(value) > 0]
|
|
if not be_trigger_rr_list:
|
|
be_trigger_rr_list = [1.0]
|
|
use_partial_tp_list = req.use_partial_tp_modes
|
|
partial_tp_rr_list = [float(value) for value in req.partial_tp_rr_values if float(value) > 0]
|
|
if not partial_tp_rr_list:
|
|
partial_tp_rr_list = [1.0]
|
|
partial_tp_percent_list = [float(value) for value in req.partial_tp_percent_values if 0 < float(value) <= 100]
|
|
if not partial_tp_percent_list:
|
|
partial_tp_percent_list = [50.0]
|
|
|
|
ranking_key = req.rank_objective
|
|
|
|
def _rank_results(items):
|
|
# Lower drawdown is better; all other objectives are maximize.
|
|
if ranking_key == "max_drawdown_pct":
|
|
return sorted(items, key=lambda x: x.get(ranking_key, 0))
|
|
return sorted(items, key=lambda x: x.get(ranking_key, 0), reverse=True)
|
|
|
|
sample_mode = req.combo_sampling_mode
|
|
rng = random.Random(req.combo_sampling_seed) if sample_mode == "random" else None
|
|
|
|
combos = []
|
|
base_grid = {
|
|
"session": session_list,
|
|
"rr": rr_list,
|
|
"lookback": lookback_list,
|
|
"ob_age": ob_age_list,
|
|
"atr_mult": atr_list,
|
|
"min_ob_size": min_ob_size_list,
|
|
"min_gap_size": min_gap_size_list,
|
|
"impulse_multiplier": impulse_multiplier_list,
|
|
"require_unmitigated_fvg": require_unmitigated_fvg_list,
|
|
"require_fvg_ob_confluence": require_fvg_ob_confluence_list,
|
|
"require_bos_confluence": require_bos_confluence_list,
|
|
"sweep": sweep_list,
|
|
"asian_sweep_only": asian_sweep_only_list,
|
|
"use_break_even": use_break_even_list,
|
|
"be_trigger_rr": be_trigger_rr_list,
|
|
"use_partial_tp": use_partial_tp_list,
|
|
"partial_tp_rr": partial_tp_rr_list,
|
|
"partial_tp_percent": partial_tp_percent_list,
|
|
}
|
|
base_keys = tuple(base_grid.keys())
|
|
|
|
def _candidate_iter():
|
|
for values in product(*(base_grid[key] for key in base_keys)):
|
|
base_params = dict(zip(base_keys, values))
|
|
for sweep_lb in (sweep_lb_list if base_params["sweep"] else [0]):
|
|
yield {
|
|
**base_params,
|
|
"sweep_lookback": sweep_lb,
|
|
}
|
|
|
|
base_count_without_sweep = 1
|
|
for key in base_keys:
|
|
if key == "sweep":
|
|
continue
|
|
base_count_without_sweep *= len(base_grid[key])
|
|
|
|
true_sweep_count = sum(1 for enabled in sweep_list if enabled)
|
|
false_sweep_count = sum(1 for enabled in sweep_list if not enabled)
|
|
sweep_factor = (true_sweep_count * len(sweep_lb_list)) + false_sweep_count
|
|
total_generated_combinations = base_count_without_sweep * sweep_factor
|
|
|
|
if sample_mode == "first":
|
|
# Stop iteration as soon as we hit the cap (no ghost work on remaining permutations).
|
|
combos = list(islice(_candidate_iter(), req.max_combinations))
|
|
else:
|
|
# Reservoir sampling keeps an unbiased random sample without storing all combos.
|
|
for generated_idx, candidate in enumerate(_candidate_iter(), start=1):
|
|
if len(combos) < req.max_combinations:
|
|
if len(combos) < req.max_combinations:
|
|
combos.append(candidate)
|
|
else:
|
|
replace_index = rng.randint(0, generated_idx - 1)
|
|
if replace_index < req.max_combinations:
|
|
combos[replace_index] = candidate
|
|
|
|
executed_combinations = len(combos)
|
|
capped_by_max_combinations = total_generated_combinations > executed_combinations
|
|
|
|
def event_stream():
|
|
started = time.perf_counter()
|
|
results = []
|
|
|
|
# Emit immediately so the UI can move off 0% as soon as the stream opens.
|
|
yield f"data: {json.dumps({'type': 'progress', 'progress': 0, 'processed': 0, 'total_combinations': executed_combinations, 'generated_combinations': total_generated_combinations, 'executed_combinations': executed_combinations, 'max_combinations': req.max_combinations, 'capped_by_max_combinations': capped_by_max_combinations, 'combo_sampling_mode': sample_mode, 'combo_sampling_seed': req.combo_sampling_seed, 'valid_results': 0, 'top_results': []})}\n\n"
|
|
|
|
for i, params in enumerate(combos):
|
|
strategy = ICTStrategy(
|
|
session=params["session"],
|
|
lookback=params["lookback"],
|
|
ob_max_age=params["ob_age"],
|
|
atr_mult=params["atr_mult"],
|
|
use_liquidity_sweep=params["sweep"],
|
|
sweep_lookback=params["sweep_lookback"],
|
|
min_gap_size=params["min_gap_size"],
|
|
impulse_multiplier=params["impulse_multiplier"],
|
|
require_unmitigated_fvg=params["require_unmitigated_fvg"],
|
|
require_bos_confluence=params["require_bos_confluence"],
|
|
min_ob_size=params["min_ob_size"],
|
|
require_fvg_ob_confluence=params["require_fvg_ob_confluence"],
|
|
asian_sweep_only=params["asian_sweep_only"],
|
|
use_break_even=params["use_break_even"],
|
|
be_trigger_rr=params["be_trigger_rr"],
|
|
use_partial_tp=params["use_partial_tp"],
|
|
partial_tp_rr=params["partial_tp_rr"],
|
|
partial_tp_percent=params["partial_tp_percent"],
|
|
)
|
|
trades = run_backtest(candles, strategy, 10000, risk_reward=params["rr"])
|
|
|
|
if req.min_trades <= len(trades) <= req.max_trades:
|
|
metrics = _risk_metrics(trades, starting_balance=10000.0)
|
|
results.append({
|
|
"params": params,
|
|
"risk_reward": params["rr"],
|
|
"pnl": round(metrics["net_pnl"], 4),
|
|
"trades": metrics["trade_count"],
|
|
"win_rate": round(metrics["win_rate"], 2),
|
|
**metrics,
|
|
})
|
|
|
|
if (i + 1) % 4 == 0 or i == len(combos) - 1:
|
|
ranked = _rank_results(results)
|
|
pareto = _pareto_front(results)
|
|
progress = ((i + 1) / max(1, len(combos))) * 100
|
|
payload = {
|
|
"type": "progress",
|
|
"progress": progress,
|
|
"processed": i + 1,
|
|
"total_combinations": executed_combinations,
|
|
"generated_combinations": total_generated_combinations,
|
|
"executed_combinations": executed_combinations,
|
|
"max_combinations": req.max_combinations,
|
|
"capped_by_max_combinations": capped_by_max_combinations,
|
|
"combo_sampling_mode": sample_mode,
|
|
"combo_sampling_seed": req.combo_sampling_seed,
|
|
"valid_results": len(ranked),
|
|
"rank_objective": ranking_key,
|
|
"top_results": ranked[:max(1, req.top_n)],
|
|
"pareto_front": sorted(pareto, key=lambda x: x.get("custom_fitness", 0), reverse=True)[:max(1, req.top_n)],
|
|
}
|
|
yield f"data: {json.dumps(payload)}\n\n"
|
|
|
|
total_elapsed = time.perf_counter() - started
|
|
ranked = _rank_results(results)
|
|
pareto = _pareto_front(results)
|
|
best = ranked[0] if ranked else None
|
|
safe_elapsed = total_elapsed if total_elapsed > 0 else 0.0001
|
|
payload = {
|
|
"type": "finished",
|
|
"dataset": dataset_id,
|
|
"timeframe": req.timeframe,
|
|
"risk_reward_values": rr_list,
|
|
"min_trades": req.min_trades,
|
|
"max_trades": req.max_trades,
|
|
"rank_objective": ranking_key,
|
|
"all_results": ranked,
|
|
"top_results": ranked[:max(1, req.top_n)],
|
|
"pareto_front": sorted(pareto, key=lambda x: x.get("custom_fitness", 0), reverse=True),
|
|
"best": best,
|
|
"best_result": best,
|
|
"elapsed_seconds": round(total_elapsed, 3),
|
|
"combos_per_second": round(len(combos) / safe_elapsed, 2),
|
|
"total_combinations": executed_combinations,
|
|
"generated_combinations": total_generated_combinations,
|
|
"executed_combinations": executed_combinations,
|
|
"max_combinations": req.max_combinations,
|
|
"capped_by_max_combinations": capped_by_max_combinations,
|
|
"combo_sampling_mode": sample_mode,
|
|
"combo_sampling_seed": req.combo_sampling_seed,
|
|
"valid_results": len(ranked),
|
|
}
|
|
yield f"data: {json.dumps(payload)}\n\n"
|
|
|
|
return StreamingResponse(
|
|
event_stream(),
|
|
media_type="text/event-stream",
|
|
headers={"Cache-Control": "no-cache", "Connection": "keep-alive", "X-Accel-Buffering": "no"},
|
|
)
|
|
|
|
|
|
@app.get("/api/optimize/monte-carlo")
|
|
def get_optimize_monte_carlo(
|
|
timeframe: int = 5,
|
|
rr: float = 2.5,
|
|
dataset: str = DEFAULT_DATASET,
|
|
session: str = "london",
|
|
lookback: int = 7,
|
|
ob_age: int = 50,
|
|
atr_mult: float = 2.5,
|
|
min_ob_size: float = 0.0005,
|
|
min_gap_size: float = 0.00025,
|
|
impulse_multiplier: float = 1.35,
|
|
require_unmitigated_fvg: bool = True,
|
|
require_fvg_ob_confluence: bool = False,
|
|
require_bos_confluence: bool = False,
|
|
sweep: bool = True,
|
|
sweep_lookback: int = 5,
|
|
asian_sweep_only: bool = False,
|
|
use_break_even: bool = False,
|
|
be_trigger_rr: float = 1.0,
|
|
use_partial_tp: bool = False,
|
|
partial_tp_rr: float = 1.0,
|
|
partial_tp_percent: float = 50.0,
|
|
runs: int = Query(default=500, ge=1, le=20000),
|
|
shuffle_trades: bool = True,
|
|
pnl_variation_pct: float = Query(default=15.0, ge=0.0, le=100.0),
|
|
price_noise_pct: float = Query(default=0.0, ge=0.0, le=100.0),
|
|
slippage_per_trade: float = Query(default=0.0, ge=0.0),
|
|
spread_per_trade: float = Query(default=0.0, ge=0.0),
|
|
ruin_drawdown_pct: float = Query(default=20.0, ge=0.0, le=100.0),
|
|
):
|
|
dataset_id = _resolve_dataset(dataset)
|
|
candles = _get_candles_for_timeframe(dataset_id, timeframe)
|
|
|
|
strategy = _build_strategy(
|
|
session=session,
|
|
lookback=lookback,
|
|
ob_age=ob_age,
|
|
atr_mult=atr_mult,
|
|
use_fvg=True,
|
|
use_ob=True,
|
|
proximity_pct=0.5,
|
|
sweep=sweep,
|
|
sweep_lookback=sweep_lookback,
|
|
min_gap_size=min_gap_size,
|
|
impulse_multiplier=impulse_multiplier,
|
|
require_unmitigated_fvg=require_unmitigated_fvg,
|
|
require_bos_confluence=require_bos_confluence,
|
|
min_ob_size=min_ob_size,
|
|
require_fvg_ob_confluence=require_fvg_ob_confluence,
|
|
asian_sweep_only=asian_sweep_only,
|
|
use_break_even=use_break_even,
|
|
be_trigger_rr=be_trigger_rr,
|
|
use_partial_tp=use_partial_tp,
|
|
partial_tp_rr=partial_tp_rr,
|
|
partial_tp_percent=partial_tp_percent,
|
|
day_filter=None,
|
|
)
|
|
trades = run_backtest(candles, strategy, 10000, risk_reward=rr)
|
|
trade_r_multiples = [getattr(t, "r_multiple", 0.0) for t in trades]
|
|
|
|
base_metrics = _risk_metrics(trades, starting_balance=10000.0)
|
|
return {
|
|
"dataset": dataset_id,
|
|
"timeframe": timeframe,
|
|
"risk_reward": rr,
|
|
"best_params": {
|
|
"session": session,
|
|
"lookback": lookback,
|
|
"ob_age": ob_age,
|
|
"atr_mult": atr_mult,
|
|
"min_ob_size": min_ob_size,
|
|
"min_gap_size": min_gap_size,
|
|
"impulse_multiplier": impulse_multiplier,
|
|
"require_unmitigated_fvg": require_unmitigated_fvg,
|
|
"require_fvg_ob_confluence": require_fvg_ob_confluence,
|
|
"require_bos_confluence": require_bos_confluence,
|
|
"sweep": sweep,
|
|
"sweep_lookback": sweep_lookback,
|
|
"asian_sweep_only": asian_sweep_only,
|
|
"use_break_even": use_break_even,
|
|
"be_trigger_rr": be_trigger_rr,
|
|
"use_partial_tp": use_partial_tp,
|
|
"partial_tp_rr": partial_tp_rr,
|
|
"partial_tp_percent": partial_tp_percent,
|
|
},
|
|
"base": {
|
|
"trade_count": len(trades),
|
|
"net_pnl": base_metrics["net_pnl"],
|
|
"win_rate": base_metrics["win_rate"],
|
|
"profit_factor": base_metrics["profit_factor"],
|
|
"max_drawdown_pct": base_metrics["max_drawdown_pct"],
|
|
},
|
|
"config": {
|
|
"runs": runs,
|
|
"risk_per_trade_pct": 1.0,
|
|
"shuffle_trades": shuffle_trades,
|
|
"pnl_variation_pct": pnl_variation_pct,
|
|
"price_noise_pct": price_noise_pct,
|
|
"slippage_per_trade": slippage_per_trade,
|
|
"spread_per_trade": spread_per_trade,
|
|
"ruin_drawdown_pct": ruin_drawdown_pct,
|
|
},
|
|
**_build_monte_carlo_payload(
|
|
trade_r_multiples,
|
|
runs=runs,
|
|
starting_balance=10000.0,
|
|
risk_per_trade_pct=1.0,
|
|
sampling_method="shuffle" if shuffle_trades else "bootstrap",
|
|
missed_trade_pct=0.0,
|
|
pnl_variation_pct=pnl_variation_pct,
|
|
price_noise_pct=price_noise_pct,
|
|
slippage_per_trade=slippage_per_trade,
|
|
spread_per_trade=spread_per_trade,
|
|
ruin_drawdown_pct=ruin_drawdown_pct,
|
|
seed=None,
|
|
base={
|
|
"trade_count": len(trades),
|
|
"net_pnl": base_metrics["net_pnl"],
|
|
"win_rate": base_metrics["win_rate"],
|
|
"profit_factor": base_metrics["profit_factor"],
|
|
"max_drawdown_pct": base_metrics["max_drawdown_pct"],
|
|
},
|
|
),
|
|
}
|
|
|
|
|
|
@app.get("/api/backtest/stream")
|
|
def stream_backtest(
|
|
timeframe: int = 5,
|
|
rr: float = 2.5,
|
|
lookback: int = 7,
|
|
ob_age: int = 50,
|
|
atr_mult: float = 2.5,
|
|
sweep: bool = True,
|
|
sweep_lookback: int = 5,
|
|
session: str = "london",
|
|
dataset: str = DEFAULT_DATASET,
|
|
use_fvg: bool = True,
|
|
use_ob: bool = True,
|
|
proximity_pct: float = 0.5,
|
|
# FVG Quality
|
|
min_gap_size: float = 0.0,
|
|
impulse_multiplier: float = 0.0,
|
|
require_unmitigated_fvg: bool = True,
|
|
require_bos_confluence: bool = False,
|
|
# Order Block
|
|
min_ob_size: float = 0.0,
|
|
require_fvg_ob_confluence: bool = False,
|
|
# Liquidity
|
|
asian_sweep_only: bool = False,
|
|
# Break-even
|
|
use_break_even: bool = False,
|
|
be_trigger_rr: float = 1.0,
|
|
# Partial TP
|
|
use_partial_tp: bool = False,
|
|
partial_tp_rr: float = 1.0,
|
|
partial_tp_percent: float = 50.0,
|
|
# Time
|
|
day_filter: Optional[str] = None,
|
|
# Risk Management
|
|
max_daily_loss: float = 0.0,
|
|
max_consecutive_losses: int = 0,
|
|
):
|
|
dataset_id = _resolve_dataset(dataset)
|
|
candles = _get_candles_for_timeframe(dataset_id, timeframe)
|
|
|
|
strategy = _build_strategy(
|
|
session=session, lookback=lookback, ob_age=ob_age, atr_mult=atr_mult,
|
|
use_fvg=use_fvg, use_ob=use_ob, proximity_pct=proximity_pct,
|
|
sweep=sweep, sweep_lookback=sweep_lookback,
|
|
min_gap_size=min_gap_size, impulse_multiplier=impulse_multiplier,
|
|
require_unmitigated_fvg=require_unmitigated_fvg,
|
|
require_bos_confluence=require_bos_confluence,
|
|
min_ob_size=min_ob_size, require_fvg_ob_confluence=require_fvg_ob_confluence,
|
|
asian_sweep_only=asian_sweep_only, day_filter=day_filter,
|
|
use_break_even=use_break_even, be_trigger_rr=be_trigger_rr,
|
|
use_partial_tp=use_partial_tp, partial_tp_rr=partial_tp_rr, partial_tp_percent=partial_tp_percent,
|
|
)
|
|
|
|
def _sse(data):
|
|
return f"data: {json.dumps(data)}\n\n"
|
|
|
|
def event_stream():
|
|
started = time.perf_counter()
|
|
streamed_trades = []
|
|
try:
|
|
for event in run_backtest_stream(
|
|
candles, strategy, 10000, risk_reward=rr,
|
|
max_daily_loss=max_daily_loss,
|
|
max_consecutive_losses=max_consecutive_losses,
|
|
):
|
|
kind = event["type"]
|
|
if kind == "start":
|
|
yield _sse({"type": "start", "total_candles": event["total_candles"]})
|
|
elif kind == "progress":
|
|
total = max(1, event["total_candles"])
|
|
progress_pct = int((event["processed_candles"] / total) * 100)
|
|
yield _sse({
|
|
"type": "progress",
|
|
"processed_candles": event["processed_candles"],
|
|
"total_candles": event["total_candles"],
|
|
"progress_pct": progress_pct,
|
|
})
|
|
elif kind == "trade":
|
|
trade = event["trade"]
|
|
streamed_trades.append(trade)
|
|
yield _sse({
|
|
"type": "trade",
|
|
"trade": _trade_payload(trade),
|
|
"stats": _stats_payload(streamed_trades, rr),
|
|
"processed_candles": event["processed_candles"],
|
|
"total_candles": event["total_candles"],
|
|
})
|
|
elif kind == "done":
|
|
duration_ms = (time.perf_counter() - started) * 1000
|
|
yield _sse({
|
|
"type": "done",
|
|
"trades": [_trade_payload(t) for t in streamed_trades],
|
|
"stats": _stats_payload(streamed_trades, rr),
|
|
"duration_ms": round(duration_ms, 1),
|
|
"candle_times": [c.time_open.isoformat() for c in candles],
|
|
})
|
|
except Exception as exc:
|
|
yield _sse({"type": "error", "message": str(exc)})
|
|
|
|
return StreamingResponse(
|
|
event_stream(),
|
|
media_type="text/event-stream",
|
|
headers={
|
|
"Cache-Control": "no-cache",
|
|
"Connection": "keep-alive",
|
|
"X-Accel-Buffering": "no",
|
|
},
|
|
)
|