Modified evaluation scripts
This commit is contained in:
Quanliang Liu
parent
13b826a31d
commit
b512027574
|
|
@ -9,3 +9,4 @@ eval_results/
|
|||
eval_data/
|
||||
*.egg-info/
|
||||
.env
|
||||
.history/
|
||||
|
|
@ -0,0 +1,66 @@
|
|||
from .baseline_strategies import (
|
||||
BuyAndHoldStrategy,
|
||||
MACDStrategy,
|
||||
KDJRSIStrategy,
|
||||
ZMRStrategy,
|
||||
SMAStrategy,
|
||||
get_all_baseline_strategies
|
||||
)
|
||||
|
||||
from .metrics import (
|
||||
calculate_cumulative_return,
|
||||
calculate_annualized_return,
|
||||
calculate_sharpe_ratio,
|
||||
calculate_maximum_drawdown,
|
||||
calculate_all_metrics,
|
||||
create_comparison_table
|
||||
)
|
||||
|
||||
from .backtest import (
|
||||
BacktestEngine,
|
||||
TradingAgentsBacktester,
|
||||
load_stock_data
|
||||
)
|
||||
|
||||
from .visualize import (
|
||||
plot_cumulative_returns,
|
||||
plot_transaction_history,
|
||||
plot_metrics_comparison,
|
||||
plot_drawdown,
|
||||
create_summary_report
|
||||
)
|
||||
|
||||
from .run_evaluation import run_evaluation
|
||||
|
||||
__all__ = [
|
||||
# Strategies
|
||||
'BuyAndHoldStrategy',
|
||||
'MACDStrategy',
|
||||
'KDJRSIStrategy',
|
||||
'ZMRStrategy',
|
||||
'SMAStrategy',
|
||||
'get_all_baseline_strategies',
|
||||
|
||||
# Metrics
|
||||
'calculate_cumulative_return',
|
||||
'calculate_annualized_return',
|
||||
'calculate_sharpe_ratio',
|
||||
'calculate_maximum_drawdown',
|
||||
'calculate_all_metrics',
|
||||
'create_comparison_table',
|
||||
|
||||
# Backtesting
|
||||
'BacktestEngine',
|
||||
'TradingAgentsBacktester',
|
||||
'load_stock_data',
|
||||
|
||||
# Visualization
|
||||
'plot_cumulative_returns',
|
||||
'plot_transaction_history',
|
||||
'plot_metrics_comparison',
|
||||
'plot_drawdown',
|
||||
'create_summary_report',
|
||||
|
||||
# Main evaluation
|
||||
'run_evaluation',
|
||||
]
|
||||
|
|
@ -0,0 +1,221 @@
|
|||
"""
|
||||
Backtesting engine for TradingAgents and baseline strategies.
|
||||
"""
|
||||
|
||||
import pandas as pd
|
||||
from typing import Dict, List
|
||||
from pathlib import Path
|
||||
import json
|
||||
|
||||
|
||||
STD_FIELDS = {"Open", "High", "Low", "Close", "Adj Close", "Volume"}
|
||||
|
||||
|
||||
class TradingAgentsBacktester:
|
||||
"""Backtest engine for TradingAgents framework."""
|
||||
|
||||
def __init__(self, trading_agents_graph, initial_capital=100000):
|
||||
self.graph = trading_agents_graph
|
||||
self.initial_capital = float(initial_capital)
|
||||
self.name = "TradingAgents"
|
||||
|
||||
def backtest(self, ticker: str, start_date: str, end_date: str, data: pd.DataFrame) -> pd.DataFrame:
|
||||
# Restrict to window
|
||||
df = data.loc[start_date:end_date].copy()
|
||||
portfolio = pd.DataFrame(index=df.index)
|
||||
portfolio["close"] = df["Close"]
|
||||
if "Volume" in df.columns:
|
||||
portfolio["Volume"] = df["Volume"]
|
||||
|
||||
portfolio["signal"] = 0
|
||||
portfolio["position"] = 0.0
|
||||
portfolio["cash"] = self.initial_capital
|
||||
portfolio["shares"] = 0.0
|
||||
portfolio["portfolio_value"] = self.initial_capital
|
||||
|
||||
decisions: List[Dict] = []
|
||||
|
||||
print(f"\nRunning TradingAgents backtest on {ticker} from {start_date} to {end_date}")
|
||||
print(f"Total trading days: {len(df)}")
|
||||
print("-" * 80)
|
||||
|
||||
for i, (date, row) in enumerate(df.iterrows()):
|
||||
date_str = date.strftime("%Y-%m-%d")
|
||||
price = float(row["Close"])
|
||||
|
||||
# Get decision
|
||||
try:
|
||||
print(f"\n[{i+1}/{len(df)}] {date_str} ... ", end="")
|
||||
final_state, decision = self.graph.propagate(ticker, date_str)
|
||||
print(f"Decision: {decision}")
|
||||
signal = self._parse_decision(decision)
|
||||
decisions.append({"date": date_str, "decision": decision, "signal": signal, "price": price})
|
||||
except Exception as e:
|
||||
print(f"Error: {e}")
|
||||
signal = 0
|
||||
decisions.append({"date": date_str, "decision": "ERROR", "signal": 0, "price": price, "error": str(e)})
|
||||
|
||||
# Previous day state
|
||||
if i > 0:
|
||||
prev_cash = float(portfolio["cash"].iloc[i - 1])
|
||||
prev_shares = float(portfolio["shares"].iloc[i - 1])
|
||||
prev_pos = float(portfolio["position"].iloc[i - 1])
|
||||
else:
|
||||
prev_cash = self.initial_capital
|
||||
prev_shares = 0.0
|
||||
prev_pos = 0.0
|
||||
|
||||
cash, shares, position = prev_cash, prev_shares, prev_pos
|
||||
|
||||
# Execute: BUY opens/keeps long with all cash; SELL closes to cash; HOLD keeps.
|
||||
if signal == 1 and prev_pos <= 0:
|
||||
# Go long full notional
|
||||
shares = cash / price if price > 0 else 0.0
|
||||
cash = 0.0
|
||||
position = 1.0
|
||||
elif signal == -1 and prev_pos > 0:
|
||||
# Exit long to cash (no shorting here; paper's figs show short arrows,
|
||||
# but transactions table is still long/flat in our public code)
|
||||
cash = shares * price
|
||||
shares = 0.0
|
||||
position = 0.0
|
||||
else:
|
||||
# Hold current stance
|
||||
position = prev_pos
|
||||
|
||||
portval = cash + shares * price
|
||||
|
||||
portfolio.loc[date, "signal"] = signal
|
||||
portfolio.loc[date, "position"] = position
|
||||
portfolio.loc[date, "cash"] = cash
|
||||
portfolio.loc[date, "shares"] = shares
|
||||
portfolio.loc[date, "portfolio_value"] = portval
|
||||
|
||||
# Returns
|
||||
portfolio["market_return"] = portfolio["close"].pct_change().fillna(0.0)
|
||||
portfolio["portfolio_return"] = portfolio["portfolio_value"].pct_change().fillna(0.0)
|
||||
portfolio["strategy_return"] = portfolio["portfolio_return"]
|
||||
portfolio["cumulative_return"] = (1.0 + portfolio["strategy_return"]).cumprod()
|
||||
|
||||
self._save_decisions_log(ticker, decisions, start_date, end_date)
|
||||
return portfolio
|
||||
|
||||
def _parse_decision(self, decision: str) -> int:
|
||||
"""
|
||||
Parse decision to signal.
|
||||
We interpret:
|
||||
- contains 'BUY' or 'LONG' -> 1
|
||||
- contains 'SELL' or 'EXIT' -> -1 (we use -1 as 'close to cash' here)
|
||||
- otherwise HOLD -> 0
|
||||
"""
|
||||
d = str(decision).upper()
|
||||
if "BUY" in d or "LONG" in d:
|
||||
return 1
|
||||
if "SELL" in d or "EXIT" in d or "CLOSE" in d:
|
||||
return -1
|
||||
return 0
|
||||
|
||||
def _save_decisions_log(self, ticker: str, decisions: List[Dict], start_date: str, end_date: str):
|
||||
out = Path(f"eval_results/{ticker}/TradingAgents_backtest")
|
||||
out.mkdir(parents=True, exist_ok=True)
|
||||
fp = out / f"decisions_{start_date}_to_{end_date}.json"
|
||||
with open(fp, "w") as f:
|
||||
json.dump(decisions, f, indent=2)
|
||||
print(f"\nDecisions log saved to: {fp}")
|
||||
|
||||
|
||||
class BacktestEngine:
|
||||
"""Engine to run and compare multiple strategies."""
|
||||
|
||||
def __init__(self, data: pd.DataFrame, initial_capital: float = 100000):
|
||||
self.data = data
|
||||
self.initial_capital = float(initial_capital)
|
||||
self.results: Dict[str, pd.DataFrame] = {}
|
||||
|
||||
def run_strategy(self, strategy, start_date: str = None, end_date: str = None, label = None) -> pd.DataFrame:
|
||||
data_filtered = self.data.loc[start_date:end_date] if (start_date and end_date) else self.data
|
||||
print(f"\nRunning {strategy.name}...")
|
||||
portfolio = strategy.backtest(data_filtered)
|
||||
self.results[label or strategy.name] = portfolio
|
||||
return portfolio
|
||||
|
||||
def run_all_strategies(self, strategies: Dict, start_date: str = None, end_date: str = None):
|
||||
for name, strategy in strategies.items():
|
||||
try:
|
||||
self.run_strategy(strategy, start_date, end_date)
|
||||
print(f"✓ {name} completed")
|
||||
except Exception as e:
|
||||
print(f"✗ {name} failed: {e}")
|
||||
|
||||
def get_results(self) -> Dict[str, pd.DataFrame]:
|
||||
return self.results
|
||||
|
||||
|
||||
def load_stock_data(ticker: str, start_date: str, end_date: str) -> pd.DataFrame:
|
||||
try:
|
||||
import yfinance as yf
|
||||
# Normalize accidental ('A','A','P','L') / ['A','A','P','L']
|
||||
if isinstance(ticker, (list, tuple)) and all(isinstance(c, str) and len(c) == 1 for c in ticker):
|
||||
ticker = "".join(ticker)
|
||||
|
||||
if not isinstance(ticker, str):
|
||||
raise ValueError("Pass a single ticker symbol as a string, e.g., 'AAPL'.")
|
||||
|
||||
df = yf.download(ticker, start=start_date, end=end_date, progress=False)
|
||||
if df.empty:
|
||||
raise ValueError(f"No data found for {ticker}")
|
||||
return df
|
||||
|
||||
except Exception as e:
|
||||
print(f"Error loading data: {e}")
|
||||
raise
|
||||
|
||||
def standardize_single_ticker(df: pd.DataFrame, ticker: str | None = None) -> pd.DataFrame:
|
||||
"""Return a single-ticker OHLCV DataFrame with simple columns.
|
||||
Works with yfinance single or multi-ticker outputs.
|
||||
"""
|
||||
df = df.copy()
|
||||
|
||||
# If columns are MultiIndex (common with multi-ticker yfinance)
|
||||
if isinstance(df.columns, pd.MultiIndex):
|
||||
# Figure out which level is the field (Open/High/...) and which is ticker
|
||||
lvl0 = set(map(str, df.columns.get_level_values(0)))
|
||||
lvl1 = set(map(str, df.columns.get_level_values(1)))
|
||||
if len(STD_FIELDS & lvl0) > 0:
|
||||
field_level, ticker_level = 0, 1
|
||||
elif len(STD_FIELDS & lvl1) > 0:
|
||||
field_level, ticker_level = 1, 0
|
||||
else:
|
||||
raise ValueError("Cannot detect OHLCV field level in MultiIndex columns.")
|
||||
|
||||
available = list(pd.Index(df.columns.get_level_values(ticker_level)).unique())
|
||||
|
||||
# Normalize weird ticker inputs like ('A','A','P','L') -> 'AAPL'
|
||||
if isinstance(ticker, (list, tuple)) and all(isinstance(c, str) and len(c) == 1 for c in ticker):
|
||||
ticker = "".join(ticker)
|
||||
if ticker is None:
|
||||
if len(available) != 1:
|
||||
raise ValueError(f"Multi-ticker DataFrame. Pick one with ticker=..., available={available}")
|
||||
ticker = available[0]
|
||||
if str(ticker) not in map(str, available):
|
||||
raise ValueError(f"Ticker {ticker!r} not in columns. Available: {available}")
|
||||
|
||||
# Slice to that ticker and drop the ticker level
|
||||
df = df.xs(ticker, axis=1, level=ticker_level)
|
||||
|
||||
# Map Adj Close -> Close if Close missing
|
||||
if "Close" not in df.columns and "Adj Close" in df.columns:
|
||||
df = df.rename(columns={"Adj Close": "Close"})
|
||||
|
||||
# Final sanity
|
||||
req = ["Open", "High", "Low", "Close"]
|
||||
missing = [c for c in req if c not in df.columns]
|
||||
if missing:
|
||||
raise ValueError(f"Data missing columns: {missing}")
|
||||
|
||||
# Ensure 'Close' is a Series (not 1-col DataFrame)
|
||||
close = df["Close"]
|
||||
if isinstance(close, pd.DataFrame) and close.shape[1] == 1:
|
||||
df["Close"] = close.iloc[:, 0]
|
||||
|
||||
return df
|
||||
|
|
@ -0,0 +1,198 @@
|
|||
"""
|
||||
Baseline trading strategies for comparison.
|
||||
Implements: Buy&Hold, MACD, KDJ+RSI, ZMR, SMA
|
||||
"""
|
||||
|
||||
import pandas as pd
|
||||
import numpy as np
|
||||
from abc import ABC, abstractmethod
|
||||
|
||||
|
||||
class BaseStrategy(ABC):
|
||||
"""Base class for trading strategies."""
|
||||
|
||||
def __init__(self, initial_capital=100000):
|
||||
self.initial_capital = float(initial_capital)
|
||||
self.name = self.__class__.__name__
|
||||
|
||||
def _close_series(self, data: pd.DataFrame) -> pd.Series:
|
||||
close = data["Close"]
|
||||
if isinstance(close, pd.DataFrame):
|
||||
if close.shape[1] == 1:
|
||||
close = close.iloc[:, 0]
|
||||
else:
|
||||
raise ValueError("Multiple 'Close' columns detected. Pass single-ticker data.")
|
||||
return pd.to_numeric(close, errors="coerce")
|
||||
|
||||
@abstractmethod
|
||||
def generate_signals(self, data: pd.DataFrame) -> pd.Series:
|
||||
"""Generate *target* position by date (1 long, -1 short, 0 flat)."""
|
||||
pass
|
||||
|
||||
def _prep_ohlcv(self, data: pd.DataFrame) -> pd.DataFrame:
|
||||
req = ["Open", "High", "Low", "Close"]
|
||||
for col in req:
|
||||
if col not in data.columns:
|
||||
raise ValueError(f"Data missing column '{col}'")
|
||||
return data.copy()
|
||||
|
||||
def backtest(self, data: pd.DataFrame) -> pd.DataFrame:
|
||||
df = self._prep_ohlcv(data)
|
||||
signals = self.generate_signals(df).astype(float)
|
||||
signals = signals.clip(lower=-1, upper=1).reindex(df.index).fillna(0)
|
||||
|
||||
# ONE place for hold semantics (Option A: 0 = no new signal)
|
||||
position = signals.replace(0, np.nan).ffill().fillna(0)
|
||||
|
||||
close = self._close_series(df)
|
||||
market_ret = close.pct_change().fillna(0.0)
|
||||
exposure = position.shift(1).fillna(0.0)
|
||||
strat_ret = (exposure * market_ret).astype(float)
|
||||
|
||||
cumret = (1.0 + strat_ret).cumprod()
|
||||
portval = self.initial_capital * cumret
|
||||
|
||||
portfolio = pd.DataFrame(index=df.index)
|
||||
portfolio["signal"] = signals
|
||||
portfolio["position"] = position
|
||||
portfolio["close"] = close
|
||||
if "Volume" in df.columns:
|
||||
vol = df["Volume"]
|
||||
if isinstance(vol, pd.DataFrame) and vol.shape[1] == 1:
|
||||
vol = vol.iloc[:, 0]
|
||||
if isinstance(vol, pd.Series):
|
||||
portfolio["Volume"] = vol
|
||||
portfolio["market_return"] = market_ret
|
||||
portfolio["strategy_return"] = strat_ret
|
||||
portfolio["cumulative_return"] = cumret
|
||||
portfolio["portfolio_value"] = portval
|
||||
portfolio["trade"] = portfolio["position"].diff().fillna(0.0)
|
||||
return portfolio
|
||||
|
||||
|
||||
class BuyAndHoldStrategy(BaseStrategy):
|
||||
"""Buy at start and hold long the whole period (no short)."""
|
||||
|
||||
def generate_signals(self, data: pd.DataFrame) -> pd.Series:
|
||||
s = pd.Series(1.0, index=data.index)
|
||||
return s
|
||||
|
||||
|
||||
class MACDStrategy(BaseStrategy):
|
||||
"""
|
||||
MACD Strategy.
|
||||
Long when MACD > signal, Short when MACD < signal.
|
||||
"""
|
||||
|
||||
def generate_signals(self, data: pd.DataFrame) -> pd.Series:
|
||||
df = data.copy()
|
||||
if "macd" not in df.columns or "macds" not in df.columns:
|
||||
df = self._calculate_macd(df)
|
||||
macd_diff = (df["macd"] - df["macds"]).fillna(0.0)
|
||||
sig = pd.Series(0.0, index=df.index)
|
||||
sig[macd_diff > 0] = 1.0
|
||||
sig[macd_diff < 0] = -1.0
|
||||
return sig
|
||||
|
||||
def _calculate_macd(self, data: pd.DataFrame, fast=12, slow=26, signal=9):
|
||||
exp1 = data["Close"].ewm(span=fast, adjust=False).mean()
|
||||
exp2 = data["Close"].ewm(span=slow, adjust=False).mean()
|
||||
macd = exp1 - exp2
|
||||
macds = macd.ewm(span=signal, adjust=False).mean()
|
||||
data["macd"] = macd
|
||||
data["macds"] = macds
|
||||
data["macdh"] = macd - macds
|
||||
return data
|
||||
|
||||
|
||||
class KDJRSIStrategy(BaseStrategy):
|
||||
"""
|
||||
KDJ & RSI Strategy (classic oversold/overbought gating).
|
||||
Long when RSI<30 & K<20; Short when RSI>70 & K>80.
|
||||
"""
|
||||
|
||||
def generate_signals(self, data: pd.DataFrame) -> pd.Series:
|
||||
df = data.copy()
|
||||
if "rsi" not in df.columns:
|
||||
df = self._calculate_rsi(df)
|
||||
if "kdj_k" not in df.columns:
|
||||
df = self._calculate_kdj(df)
|
||||
|
||||
sig = pd.Series(0.0, index=df.index)
|
||||
sig[(df["rsi"] < 30) & (df["kdj_k"] < 20)] = 1.0
|
||||
sig[(df["rsi"] > 70) & (df["kdj_k"] > 80)] = -1.0
|
||||
return sig
|
||||
|
||||
def _calculate_rsi(self, data: pd.DataFrame, period=14):
|
||||
# Wilder's smoothing approximation via EMA improves stability
|
||||
delta = data["Close"].diff()
|
||||
up = delta.clip(lower=0)
|
||||
down = -delta.clip(upper=0)
|
||||
roll_up = up.ewm(alpha=1/period, adjust=False).mean()
|
||||
roll_down = down.ewm(alpha=1/period, adjust=False).mean()
|
||||
rs = roll_up / roll_down.replace(0, np.nan)
|
||||
data["rsi"] = 100 - (100 / (1 + rs))
|
||||
return data
|
||||
|
||||
def _calculate_kdj(self, data: pd.DataFrame, period=9):
|
||||
low_min = data["Low"].rolling(window=period, min_periods=period).min()
|
||||
high_max = data["High"].rolling(window=period, min_periods=period).max()
|
||||
den = (high_max - low_min).replace(0, np.nan)
|
||||
rsv = 100 * (data["Close"] - low_min) / den
|
||||
k = rsv.ewm(com=2, adjust=False, min_periods=1).mean()
|
||||
d = k.ewm(com=2, adjust=False, min_periods=1).mean()
|
||||
j = 3 * k - 2 * d
|
||||
data["kdj_k"], data["kdj_d"], data["kdj_j"] = k, d, j
|
||||
return data
|
||||
|
||||
|
||||
class ZMRStrategy(BaseStrategy):
|
||||
"""
|
||||
Zero-mean reversion on z-score of Close vs rolling mean.
|
||||
"""
|
||||
|
||||
def __init__(self, initial_capital=100000, lookback=20, threshold=1.0):
|
||||
super().__init__(initial_capital)
|
||||
self.lookback = int(lookback)
|
||||
self.threshold = float(threshold)
|
||||
|
||||
def generate_signals(self, data: pd.DataFrame) -> pd.Series:
|
||||
close = self._close_series(data)
|
||||
rm = close.rolling(window=self.lookback, min_periods=self.lookback).mean()
|
||||
rs = close.rolling(window=self.lookback, min_periods=self.lookback).std()
|
||||
z = (close - rm) / rs.replace(0, pd.NA)
|
||||
sig = pd.Series(0.0, index=data.index)
|
||||
sig[z < -self.threshold] = 1.0
|
||||
sig[z > self.threshold] = -1.0
|
||||
return sig
|
||||
|
||||
|
||||
class SMAStrategy(BaseStrategy):
|
||||
"""
|
||||
SMA crossover (50/200 by default).
|
||||
"""
|
||||
|
||||
def __init__(self, initial_capital=100000, short_window=50, long_window=200):
|
||||
super().__init__(initial_capital)
|
||||
self.short_window = int(short_window)
|
||||
self.long_window = int(long_window)
|
||||
|
||||
def generate_signals(self, data: pd.DataFrame) -> pd.Series:
|
||||
close = self._close_series(data)
|
||||
short = close.rolling(window=self.short_window, min_periods=self.short_window).mean()
|
||||
long_ = close.rolling(window=self.long_window, min_periods=self.long_window).mean()
|
||||
sig = pd.Series(0.0, index=data.index)
|
||||
sig[short > long_] = 1.0
|
||||
sig[short < long_] = -1.0
|
||||
return sig
|
||||
|
||||
|
||||
def get_all_baseline_strategies(initial_capital=100000):
|
||||
"""Get all baseline strategies for comparison."""
|
||||
return {
|
||||
"BuyAndHold": BuyAndHoldStrategy(initial_capital),
|
||||
"MACD": MACDStrategy(initial_capital),
|
||||
"KDJ&RSI": KDJRSIStrategy(initial_capital),
|
||||
"ZMR": ZMRStrategy(initial_capital),
|
||||
"SMA": SMAStrategy(initial_capital),
|
||||
}
|
||||
|
|
@ -0,0 +1,116 @@
|
|||
"""
|
||||
Evaluation metrics for trading strategies.
|
||||
Implements: Cumulative Return, Annualized Return, Sharpe Ratio, Maximum Drawdown
|
||||
"""
|
||||
|
||||
import pandas as pd
|
||||
import numpy as np
|
||||
from typing import Dict
|
||||
|
||||
|
||||
def _require_cols(df: pd.DataFrame, cols):
|
||||
missing = [c for c in cols if c not in df.columns]
|
||||
if missing:
|
||||
raise ValueError(f"Portfolio missing columns: {missing}")
|
||||
|
||||
|
||||
def calculate_cumulative_return(portfolio: pd.DataFrame) -> float:
|
||||
"""CR% = (V_end / V_start - 1) * 100"""
|
||||
_require_cols(portfolio, ["portfolio_value"])
|
||||
v_start = float(portfolio["portfolio_value"].iloc[0])
|
||||
v_end = float(portfolio["portfolio_value"].iloc[-1])
|
||||
if v_start <= 0:
|
||||
return 0.0
|
||||
return (v_end / v_start - 1.0) * 100.0
|
||||
|
||||
|
||||
def calculate_annualized_return(portfolio: pd.DataFrame, trading_days: int | None = None) -> float:
|
||||
"""AR% = ((V_end / V_start) ** (1/years) - 1) * 100 with 252 trading days/year."""
|
||||
_require_cols(portfolio, ["portfolio_value"])
|
||||
v_start = float(portfolio["portfolio_value"].iloc[0])
|
||||
v_end = float(portfolio["portfolio_value"].iloc[-1])
|
||||
if v_start <= 0 or v_end <= 0:
|
||||
return 0.0
|
||||
if trading_days is None:
|
||||
trading_days = len(portfolio)
|
||||
years = trading_days / 252.0
|
||||
if years <= 0:
|
||||
return 0.0
|
||||
return ((v_end / v_start) ** (1.0 / years) - 1.0) * 100.0
|
||||
|
||||
|
||||
def calculate_sharpe_ratio(portfolio: pd.DataFrame, risk_free_rate: float = 0.02) -> float:
|
||||
"""
|
||||
SR = (E[r] - r_f) / stdev(r), where r are *daily* strategy returns,
|
||||
annualized using 252 trading days (paper S1.2.3).
|
||||
"""
|
||||
_require_cols(portfolio, ["strategy_return"])
|
||||
r = portfolio["strategy_return"].dropna().astype(float)
|
||||
if len(r) < 2 or r.std() == 0:
|
||||
return 0.0
|
||||
mean_ann = r.mean() * 252.0
|
||||
std_ann = r.std(ddof=1) * np.sqrt(252.0)
|
||||
if std_ann == 0:
|
||||
return 0.0
|
||||
return (mean_ann - risk_free_rate) / std_ann
|
||||
|
||||
|
||||
def calculate_maximum_drawdown(portfolio: pd.DataFrame) -> float:
|
||||
"""MDD% = max drawdown on portfolio_value (peak->trough) * 100"""
|
||||
_require_cols(portfolio, ["portfolio_value"])
|
||||
values = portfolio["portfolio_value"].astype(float)
|
||||
running_max = values.cummax()
|
||||
drawdown = (values - running_max) / running_max
|
||||
return float(drawdown.min() * -100.0)
|
||||
|
||||
|
||||
def calculate_win_rate(portfolio: pd.DataFrame) -> float:
|
||||
"""% days where strategy_return > 0"""
|
||||
_require_cols(portfolio, ["strategy_return"])
|
||||
r = portfolio["strategy_return"].dropna()
|
||||
if len(r) == 0:
|
||||
return 0.0
|
||||
return 100.0 * (r > 0).sum() / len(r)
|
||||
|
||||
|
||||
def calculate_profit_factor(portfolio: pd.DataFrame) -> float:
|
||||
"""Gross profit / gross loss on daily returns (informative extra metric)."""
|
||||
_require_cols(portfolio, ["strategy_return"])
|
||||
r = portfolio["strategy_return"].dropna()
|
||||
gp = r[r > 0].sum()
|
||||
gl = -r[r < 0].sum()
|
||||
if gl == 0:
|
||||
return float("inf") if gp > 0 else 0.0
|
||||
return float(gp / gl)
|
||||
|
||||
|
||||
def calculate_all_metrics(portfolio: pd.DataFrame, risk_free_rate: float = 0.02) -> Dict[str, float]:
|
||||
return {
|
||||
"Cumulative Return (%)": calculate_cumulative_return(portfolio),
|
||||
"Annualized Return (%)": calculate_annualized_return(portfolio),
|
||||
"Sharpe Ratio": calculate_sharpe_ratio(portfolio, risk_free_rate),
|
||||
"Maximum Drawdown (%)": calculate_maximum_drawdown(portfolio),
|
||||
# Extras (not in table but handy)
|
||||
"Win Rate (%)": calculate_win_rate(portfolio),
|
||||
"Profit Factor": calculate_profit_factor(portfolio),
|
||||
}
|
||||
|
||||
|
||||
def print_metrics(metrics: Dict[str, float], strategy_name: str = "Strategy"):
|
||||
print(f"\n{'='*60}")
|
||||
print(f"{strategy_name} Performance Metrics")
|
||||
print(f"{'='*60}")
|
||||
for k, v in metrics.items():
|
||||
if "Ratio" in k or "Factor" in k:
|
||||
print(f"{k:30s}: {v:8.2f}")
|
||||
else:
|
||||
print(f"{k:30s}: {v:8.2f}%")
|
||||
print(f"{'='*60}\n")
|
||||
|
||||
|
||||
def create_comparison_table(all_metrics: Dict[str, Dict[str, float]]) -> pd.DataFrame:
|
||||
df = pd.DataFrame(all_metrics).T
|
||||
df = df.round(2)
|
||||
if "Sharpe Ratio" in df.columns:
|
||||
df = df.sort_values("Sharpe Ratio", ascending=False)
|
||||
return df
|
||||
|
|
@ -0,0 +1,209 @@
|
|||
"""
|
||||
Main evaluation script to run backtesting and generate results.
|
||||
Evaluates TradingAgents against baseline strategies for a single ticker.
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import sys
|
||||
from pathlib import Path
|
||||
from datetime import datetime
|
||||
import pandas as pd
|
||||
|
||||
# Add parent directory to path
|
||||
sys.path.insert(0, str(Path(__file__).parent.parent))
|
||||
|
||||
from evaluation.baseline_strategies import get_all_baseline_strategies
|
||||
from evaluation.backtest import BacktestEngine, TradingAgentsBacktester, load_stock_data, standardize_single_ticker
|
||||
from evaluation.metrics import calculate_all_metrics, create_comparison_table, print_metrics
|
||||
from evaluation.visualize import create_summary_report
|
||||
|
||||
from tradingagents.graph.trading_graph import TradingAgentsGraph
|
||||
from tradingagents.default_config import DEFAULT_CONFIG
|
||||
|
||||
def is_debugging() -> bool:
|
||||
try:
|
||||
import debugpy
|
||||
return debugpy.is_client_connected()
|
||||
except Exception:
|
||||
return False
|
||||
|
||||
|
||||
def run_evaluation(
|
||||
ticker: str,
|
||||
start_date: str,
|
||||
end_date: str,
|
||||
initial_capital: float = 100000,
|
||||
include_tradingagents: bool = True,
|
||||
output_dir: str = None,
|
||||
config: dict = None
|
||||
):
|
||||
"""
|
||||
Run complete evaluation: baselines + TradingAgents for a single ticker.
|
||||
"""
|
||||
print(f"\n{'='*80}")
|
||||
print(f"EVALUATION: {ticker} from {start_date} to {end_date}")
|
||||
print(f"Initial Capital: ${initial_capital:,.2f}")
|
||||
print(f"{'='*80}\n")
|
||||
|
||||
# Output dir
|
||||
if output_dir is None:
|
||||
output_dir = f"eval_results/{ticker}/{datetime.now().strftime('%Y%m%d_%H%M%S')}"
|
||||
out = Path(output_dir)
|
||||
out.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
# Load data
|
||||
print("\n" + "="*80)
|
||||
print("STEP 1: Loading Stock Data")
|
||||
print("="*80)
|
||||
data = load_stock_data(ticker, start_date, end_date)
|
||||
data = standardize_single_ticker(data, ticker)
|
||||
|
||||
# Backtest engine
|
||||
engine = BacktestEngine(data, initial_capital)
|
||||
|
||||
# Baselines
|
||||
print("\n" + "="*80)
|
||||
print("STEP 2: Running Baseline Strategies")
|
||||
print("="*80)
|
||||
baselines = get_all_baseline_strategies(initial_capital)
|
||||
|
||||
for name, strategy in baselines.items():
|
||||
try:
|
||||
print(f"\nRunning {name}...", end=" ")
|
||||
portfolio = engine.run_strategy(strategy, start_date, end_date)
|
||||
print("✓ Complete")
|
||||
except Exception as e:
|
||||
print(f"✗ Failed: {e}")
|
||||
|
||||
# TradingAgents
|
||||
if include_tradingagents:
|
||||
print("\n" + "="*80)
|
||||
print("STEP 3: Running TradingAgents")
|
||||
print("="*80)
|
||||
try:
|
||||
cfg = (config or DEFAULT_CONFIG).copy()
|
||||
# Fast eval defaults (you can override from CLI)
|
||||
cfg["deep_think_llm"] = cfg.get("deep_think_llm", "gpt-4o-mini")
|
||||
cfg["quick_think_llm"] = cfg.get("quick_think_llm", "gpt-4o-mini")
|
||||
cfg["max_debate_rounds"] = cfg.get("max_debate_rounds", 1)
|
||||
cfg["max_risk_discuss_rounds"] = cfg.get("max_risk_discuss_rounds", 1)
|
||||
# Deterministic-ish decoding for reproducibility
|
||||
cfg.setdefault("llm_params", {}).update({"temperature": 0, "top_p": 1.0, "seed": 42})
|
||||
|
||||
print(f"\nInitializing TradingAgents...")
|
||||
print(f" Deep Thinking LLM: {cfg['deep_think_llm']}")
|
||||
print(f" Quick Thinking LLM: {cfg['quick_think_llm']}")
|
||||
print(f" Debate Rounds: {cfg['max_debate_rounds']}")
|
||||
|
||||
graph = TradingAgentsGraph(
|
||||
selected_analysts=["market", "social", "news", "fundamentals"],
|
||||
debug=False,
|
||||
config=cfg
|
||||
)
|
||||
ta_backtester = TradingAgentsBacktester(graph, initial_capital)
|
||||
ta_portfolio = ta_backtester.backtest(ticker, start_date, end_date, data)
|
||||
|
||||
engine.results["TradingAgents"] = ta_portfolio
|
||||
print("\n✓ TradingAgents backtest complete")
|
||||
|
||||
except Exception as e:
|
||||
print(f"\n✗ TradingAgents failed: {e}")
|
||||
import traceback
|
||||
traceback.print_exc()
|
||||
|
||||
# Metrics
|
||||
print("\n" + "="*80)
|
||||
print("STEP 4: Calculating Performance Metrics")
|
||||
print("="*80)
|
||||
all_metrics = {}
|
||||
for name, portfolio in engine.results.items():
|
||||
metrics = calculate_all_metrics(portfolio)
|
||||
all_metrics[name] = metrics
|
||||
print_metrics(metrics, name)
|
||||
|
||||
comparison_df = create_comparison_table(all_metrics)
|
||||
|
||||
print("\n" + "="*80)
|
||||
print("PERFORMANCE COMPARISON TABLE")
|
||||
print("="*80)
|
||||
print(comparison_df.to_string())
|
||||
print("\n")
|
||||
|
||||
comparison_df.to_csv(out / f"{ticker}_comparison.csv")
|
||||
print(f"Comparison table saved to: {out / f'{ticker}_comparison.csv'}")
|
||||
|
||||
# Visuals
|
||||
print("\n" + "="*80)
|
||||
print("STEP 5: Generating Visualizations")
|
||||
print("="*80)
|
||||
create_summary_report(ticker, engine.results, comparison_df, output_dir)
|
||||
|
||||
print("\n" + "="*80)
|
||||
print("EVALUATION COMPLETE")
|
||||
print("="*80)
|
||||
print(f"\nResults saved to: {out}")
|
||||
print(f" - Comparison table: {ticker}_comparison.csv")
|
||||
print(f" - Cumulative returns plot: {ticker}_cumulative_returns.png")
|
||||
print(f" - Metrics comparison: {ticker}_metrics_comparison.png")
|
||||
if include_tradingagents and "TradingAgents" in engine.results:
|
||||
print(f" - Transaction history: {ticker}_TradingAgents_transactions.png")
|
||||
print(f" - Drawdown analysis: {ticker}_drawdown.png")
|
||||
|
||||
return engine.results, comparison_df
|
||||
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(description="Run TradingAgents evaluation with baseline comparisons")
|
||||
parser.add_argument("ticker", type=str, help="Stock ticker symbol (e.g., AAPL)")
|
||||
parser.add_argument("--start-date", type=str, required=True, help="Start date (YYYY-MM-DD)")
|
||||
parser.add_argument("--end-date", type=str, required=True, help="End date (YYYY-MM-DD)")
|
||||
parser.add_argument("--capital", type=float, default=100000, help="Initial capital (default: 100000)")
|
||||
parser.add_argument("--no-tradingagents", action="store_true", help="Skip TradingAgents")
|
||||
parser.add_argument("--output-dir", type=str, default=None, help="Output directory for results")
|
||||
parser.add_argument("--deep-llm", type=str, default="gpt-4o-mini", help="Deep thinking LLM model")
|
||||
parser.add_argument("--quick-llm", type=str, default="gpt-4o-mini", help="Quick thinking LLM model")
|
||||
parser.add_argument("--debate-rounds", type=int, default=1, help="Number of debate rounds (default: 1)")
|
||||
|
||||
# Used for debugging
|
||||
|
||||
if is_debugging():
|
||||
config = DEFAULT_CONFIG.copy()
|
||||
config.update({
|
||||
"deep_think_llm": "gpt-4o-mini",
|
||||
"quick_think_llm": "gpt-4o-mini",
|
||||
"max_debate_rounds": 1,
|
||||
"max_risk_discuss_rounds": 1,
|
||||
"llm_params": {"temperature": 0, "top_p": 1.0, "seed": 42},
|
||||
})
|
||||
run_evaluation(
|
||||
ticker="AAPL",
|
||||
start_date="2024-01-01",
|
||||
end_date="2024-03-30",
|
||||
initial_capital=1000,
|
||||
include_tradingagents=False,
|
||||
output_dir="./evaluation/results",
|
||||
config=config
|
||||
)
|
||||
return
|
||||
|
||||
# Build config
|
||||
args = parser.parse_args()
|
||||
config = DEFAULT_CONFIG.copy()
|
||||
config["deep_think_llm"] = args.deep_llm
|
||||
config["quick_think_llm"] = args.quick_llm
|
||||
config["max_debate_rounds"] = args.debate_rounds
|
||||
config["max_risk_discuss_rounds"] = args.debate_rounds
|
||||
config.setdefault("llm_params", {}).update({"temperature": 0, "top_p": 1.0, "seed": 42})
|
||||
|
||||
run_evaluation(
|
||||
ticker=args.ticker,
|
||||
start_date=args.start_date,
|
||||
end_date=args.end_date,
|
||||
initial_capital=args.capital,
|
||||
include_tradingagents=not args.no_tradingagents,
|
||||
output_dir=args.output_dir,
|
||||
config=config
|
||||
)
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
|
@ -0,0 +1,401 @@
|
|||
"""
|
||||
Visualization tools for trading strategy evaluation.
|
||||
Generates plots and reports for comparing TradingAgents with baseline strategies.
|
||||
"""
|
||||
|
||||
import pandas as pd
|
||||
import numpy as np
|
||||
import matplotlib.pyplot as plt
|
||||
from pathlib import Path
|
||||
from typing import Dict
|
||||
import warnings
|
||||
|
||||
warnings.filterwarnings('ignore')
|
||||
|
||||
# Try to import seaborn for better styling (optional)
|
||||
try:
|
||||
import seaborn as sns
|
||||
plt.style.use('seaborn-v0_8-darkgrid')
|
||||
sns.set_palette("husl")
|
||||
HAS_SEABORN = True
|
||||
except ImportError:
|
||||
HAS_SEABORN = False
|
||||
# Use default matplotlib styling
|
||||
plt.rcParams['figure.facecolor'] = 'white'
|
||||
plt.rcParams['axes.facecolor'] = 'white'
|
||||
plt.rcParams['axes.grid'] = True
|
||||
|
||||
|
||||
def plot_cumulative_returns(
|
||||
results: Dict[str, pd.DataFrame],
|
||||
ticker: str,
|
||||
output_path: str = None,
|
||||
figsize: tuple = (14, 8)
|
||||
) -> plt.Figure:
|
||||
"""
|
||||
Plot cumulative returns comparison for all strategies.
|
||||
|
||||
Args:
|
||||
results: Dictionary mapping strategy name to portfolio DataFrame
|
||||
ticker: Stock ticker symbol
|
||||
output_path: Path to save the figure (optional)
|
||||
figsize: Figure size (width, height)
|
||||
|
||||
Returns:
|
||||
matplotlib Figure object
|
||||
"""
|
||||
fig, ax = plt.subplots(figsize=figsize)
|
||||
|
||||
for name, portfolio in results.items():
|
||||
if "cumulative_return" in portfolio.columns:
|
||||
cumulative = (portfolio["cumulative_return"] - 1) * 100 # Convert to percentage
|
||||
ax.plot(portfolio.index, cumulative, label=name, linewidth=2, alpha=0.8)
|
||||
|
||||
ax.set_xlabel('Date', fontsize=12, fontweight='bold')
|
||||
ax.set_ylabel('Cumulative Return (%)', fontsize=12, fontweight='bold')
|
||||
ax.set_title(f'{ticker} - Cumulative Returns Comparison', fontsize=14, fontweight='bold')
|
||||
ax.legend(loc='best', fontsize=10, framealpha=0.9)
|
||||
ax.grid(True, alpha=0.3)
|
||||
ax.axhline(y=0, color='black', linestyle='--', linewidth=1, alpha=0.5)
|
||||
|
||||
# Format y-axis as percentage
|
||||
ax.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'{y:.1f}%'))
|
||||
|
||||
plt.tight_layout()
|
||||
|
||||
if output_path:
|
||||
fig.savefig(output_path, dpi=300, bbox_inches='tight')
|
||||
print(f"✓ Saved cumulative returns plot to: {output_path}")
|
||||
|
||||
return fig
|
||||
|
||||
|
||||
def plot_transaction_history(
|
||||
portfolio: pd.DataFrame,
|
||||
ticker: str,
|
||||
strategy_name: str = "TradingAgents",
|
||||
output_path: str = None,
|
||||
figsize: tuple = (14, 10)
|
||||
) -> plt.Figure:
|
||||
"""
|
||||
Plot transaction history with buy/sell signals overlaid on price chart.
|
||||
|
||||
Args:
|
||||
portfolio: Portfolio DataFrame with 'signal' and 'close' columns
|
||||
ticker: Stock ticker symbol
|
||||
strategy_name: Name of the strategy
|
||||
output_path: Path to save the figure (optional)
|
||||
figsize: Figure size (width, height)
|
||||
|
||||
Returns:
|
||||
matplotlib Figure object
|
||||
"""
|
||||
fig, (ax1, ax2) = plt.subplots(2, 1, figsize=figsize, height_ratios=[2, 1])
|
||||
|
||||
# Price chart with signals
|
||||
ax1.plot(portfolio.index, portfolio["close"], label='Close Price',
|
||||
color='blue', linewidth=1.5, alpha=0.7)
|
||||
|
||||
# Buy signals (signal == 1 and previous signal != 1)
|
||||
signals = portfolio["signal"].copy()
|
||||
buy_signals = (signals == 1) & (signals.shift(1) != 1)
|
||||
sell_signals = (signals == -1) & (signals.shift(1) != -1)
|
||||
|
||||
# Plot buy/sell markers
|
||||
if buy_signals.any():
|
||||
ax1.scatter(portfolio.index[buy_signals],
|
||||
portfolio.loc[buy_signals, "close"],
|
||||
marker='^', color='green', s=100, label='Buy',
|
||||
zorder=5, alpha=0.8)
|
||||
|
||||
if sell_signals.any():
|
||||
ax1.scatter(portfolio.index[sell_signals],
|
||||
portfolio.loc[sell_signals, "close"],
|
||||
marker='v', color='red', s=100, label='Sell',
|
||||
zorder=5, alpha=0.8)
|
||||
|
||||
ax1.set_ylabel('Price ($)', fontsize=12, fontweight='bold')
|
||||
ax1.set_title(f'{ticker} - {strategy_name} Transaction History',
|
||||
fontsize=14, fontweight='bold')
|
||||
ax1.legend(loc='best', fontsize=10)
|
||||
ax1.grid(True, alpha=0.3)
|
||||
|
||||
# Portfolio value
|
||||
ax2.plot(portfolio.index, portfolio["portfolio_value"],
|
||||
label='Portfolio Value', color='purple', linewidth=2)
|
||||
ax2.fill_between(portfolio.index, portfolio["portfolio_value"],
|
||||
alpha=0.3, color='purple')
|
||||
ax2.set_xlabel('Date', fontsize=12, fontweight='bold')
|
||||
ax2.set_ylabel('Portfolio Value ($)', fontsize=12, fontweight='bold')
|
||||
ax2.legend(loc='best', fontsize=10)
|
||||
ax2.grid(True, alpha=0.3)
|
||||
|
||||
# Format y-axis as currency
|
||||
ax2.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'${y:,.0f}'))
|
||||
|
||||
plt.tight_layout()
|
||||
|
||||
if output_path:
|
||||
fig.savefig(output_path, dpi=300, bbox_inches='tight')
|
||||
print(f"✓ Saved transaction history plot to: {output_path}")
|
||||
|
||||
return fig
|
||||
|
||||
|
||||
def plot_metrics_comparison(
|
||||
comparison_df: pd.DataFrame,
|
||||
ticker: str,
|
||||
output_path: str = None,
|
||||
figsize: tuple = (16, 10)
|
||||
) -> plt.Figure:
|
||||
"""
|
||||
Create bar charts comparing key metrics across strategies.
|
||||
|
||||
Args:
|
||||
comparison_df: DataFrame with strategies as rows and metrics as columns
|
||||
ticker: Stock ticker symbol
|
||||
output_path: Path to save the figure (optional)
|
||||
figsize: Figure size (width, height)
|
||||
|
||||
Returns:
|
||||
matplotlib Figure object
|
||||
"""
|
||||
# Select key metrics (matching paper's Table 1)
|
||||
metrics_to_plot = [
|
||||
"Cumulative Return (%)",
|
||||
"Annualized Return (%)",
|
||||
"Sharpe Ratio",
|
||||
"Maximum Drawdown (%)"
|
||||
]
|
||||
|
||||
# Filter to available metrics
|
||||
available_metrics = [m for m in metrics_to_plot if m in comparison_df.columns]
|
||||
|
||||
if not available_metrics:
|
||||
raise ValueError("No matching metrics found in comparison DataFrame")
|
||||
|
||||
n_metrics = len(available_metrics)
|
||||
fig, axes = plt.subplots(2, 2, figsize=figsize)
|
||||
axes = axes.flatten()
|
||||
|
||||
for idx, metric in enumerate(available_metrics):
|
||||
ax = axes[idx]
|
||||
data = comparison_df[metric].sort_values(ascending=False)
|
||||
|
||||
# Color code: TradingAgents in different color
|
||||
colors = ['#FF6B6B' if name == 'TradingAgents' else '#4ECDC4'
|
||||
for name in data.index]
|
||||
|
||||
bars = ax.barh(range(len(data)), data.values, color=colors, alpha=0.8)
|
||||
ax.set_yticks(range(len(data)))
|
||||
ax.set_yticklabels(data.index, fontsize=10)
|
||||
ax.set_xlabel(metric, fontsize=11, fontweight='bold')
|
||||
ax.set_title(metric, fontsize=12, fontweight='bold')
|
||||
ax.grid(True, alpha=0.3, axis='x')
|
||||
|
||||
# Add value labels on bars
|
||||
for i, (bar, value) in enumerate(zip(bars, data.values)):
|
||||
if "Ratio" in metric:
|
||||
label = f'{value:.2f}'
|
||||
else:
|
||||
label = f'{value:.1f}%'
|
||||
ax.text(value, bar.get_y() + bar.get_height()/2,
|
||||
f' {label}', va='center', fontsize=9)
|
||||
|
||||
# Hide unused subplots
|
||||
for idx in range(n_metrics, 4):
|
||||
axes[idx].axis('off')
|
||||
|
||||
fig.suptitle(f'{ticker} - Performance Metrics Comparison',
|
||||
fontsize=16, fontweight='bold', y=0.995)
|
||||
plt.tight_layout()
|
||||
|
||||
if output_path:
|
||||
fig.savefig(output_path, dpi=300, bbox_inches='tight')
|
||||
print(f"✓ Saved metrics comparison plot to: {output_path}")
|
||||
|
||||
return fig
|
||||
|
||||
|
||||
def plot_drawdown(
|
||||
results: Dict[str, pd.DataFrame],
|
||||
ticker: str,
|
||||
output_path: str = None,
|
||||
figsize: tuple = (14, 8)
|
||||
) -> plt.Figure:
|
||||
"""
|
||||
Plot drawdown analysis for all strategies.
|
||||
|
||||
Args:
|
||||
results: Dictionary mapping strategy name to portfolio DataFrame
|
||||
ticker: Stock ticker symbol
|
||||
output_path: Path to save the figure (optional)
|
||||
figsize: Figure size (width, height)
|
||||
|
||||
Returns:
|
||||
matplotlib Figure object
|
||||
"""
|
||||
fig, ax = plt.subplots(figsize=figsize)
|
||||
|
||||
for name, portfolio in results.items():
|
||||
if "portfolio_value" in portfolio.columns:
|
||||
values = portfolio["portfolio_value"]
|
||||
running_max = values.cummax()
|
||||
drawdown = (values - running_max) / running_max * 100
|
||||
ax.plot(portfolio.index, drawdown, label=name, linewidth=2, alpha=0.7)
|
||||
|
||||
ax.set_xlabel('Date', fontsize=12, fontweight='bold')
|
||||
ax.set_ylabel('Drawdown (%)', fontsize=12, fontweight='bold')
|
||||
ax.set_title(f'{ticker} - Drawdown Analysis', fontsize=14, fontweight='bold')
|
||||
ax.legend(loc='best', fontsize=10, framealpha=0.9)
|
||||
ax.grid(True, alpha=0.3)
|
||||
ax.axhline(y=0, color='black', linestyle='--', linewidth=1, alpha=0.5)
|
||||
|
||||
# Fill drawdown areas
|
||||
for name, portfolio in results.items():
|
||||
if "portfolio_value" in portfolio.columns:
|
||||
values = portfolio["portfolio_value"]
|
||||
running_max = values.cummax()
|
||||
drawdown = (values - running_max) / running_max * 100
|
||||
ax.fill_between(portfolio.index, drawdown, 0, alpha=0.1)
|
||||
|
||||
# Format y-axis as percentage
|
||||
ax.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'{y:.1f}%'))
|
||||
|
||||
plt.tight_layout()
|
||||
|
||||
if output_path:
|
||||
fig.savefig(output_path, dpi=300, bbox_inches='tight')
|
||||
print(f"✓ Saved drawdown plot to: {output_path}")
|
||||
|
||||
return fig
|
||||
|
||||
|
||||
def plot_returns_distribution(
|
||||
results: Dict[str, pd.DataFrame],
|
||||
ticker: str,
|
||||
output_path: str = None,
|
||||
figsize: tuple = (14, 8)
|
||||
) -> plt.Figure:
|
||||
"""
|
||||
Plot distribution of daily returns for all strategies.
|
||||
|
||||
Args:
|
||||
results: Dictionary mapping strategy name to portfolio DataFrame
|
||||
ticker: Stock ticker symbol
|
||||
output_path: Path to save the figure (optional)
|
||||
figsize: Figure size (width, height)
|
||||
|
||||
Returns:
|
||||
matplotlib Figure object
|
||||
"""
|
||||
fig, ax = plt.subplots(figsize=figsize)
|
||||
|
||||
for name, portfolio in results.items():
|
||||
if "strategy_return" in portfolio.columns:
|
||||
returns = portfolio["strategy_return"].dropna() * 100 # Convert to percentage
|
||||
ax.hist(returns, bins=50, alpha=0.5, label=name, density=True)
|
||||
|
||||
ax.set_xlabel('Daily Return (%)', fontsize=12, fontweight='bold')
|
||||
ax.set_ylabel('Density', fontsize=12, fontweight='bold')
|
||||
ax.set_title(f'{ticker} - Returns Distribution', fontsize=14, fontweight='bold')
|
||||
ax.legend(loc='best', fontsize=10)
|
||||
ax.grid(True, alpha=0.3)
|
||||
ax.axvline(x=0, color='black', linestyle='--', linewidth=1, alpha=0.5)
|
||||
|
||||
plt.tight_layout()
|
||||
|
||||
if output_path:
|
||||
fig.savefig(output_path, dpi=300, bbox_inches='tight')
|
||||
print(f"✓ Saved returns distribution plot to: {output_path}")
|
||||
|
||||
return fig
|
||||
|
||||
|
||||
def create_summary_report(
|
||||
ticker: str,
|
||||
results: Dict[str, pd.DataFrame],
|
||||
comparison_df: pd.DataFrame,
|
||||
output_dir: str
|
||||
) -> None:
|
||||
"""
|
||||
Generate comprehensive visual summary report.
|
||||
Creates all standard plots and saves them to output directory.
|
||||
|
||||
Args:
|
||||
ticker: Stock ticker symbol
|
||||
results: Dictionary mapping strategy name to portfolio DataFrame
|
||||
comparison_df: DataFrame with performance metrics comparison
|
||||
output_dir: Directory to save output files
|
||||
"""
|
||||
output_path = Path(output_dir)
|
||||
output_path.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
print("\nGenerating visualizations...")
|
||||
|
||||
# 1. Cumulative Returns
|
||||
try:
|
||||
plot_cumulative_returns(
|
||||
results,
|
||||
ticker,
|
||||
output_path=str(output_path / f"{ticker}_cumulative_returns.png")
|
||||
)
|
||||
except Exception as e:
|
||||
print(f"✗ Failed to generate cumulative returns plot: {e}")
|
||||
|
||||
# 2. Metrics Comparison
|
||||
try:
|
||||
plot_metrics_comparison(
|
||||
comparison_df,
|
||||
ticker,
|
||||
output_path=str(output_path / f"{ticker}_metrics_comparison.png")
|
||||
)
|
||||
except Exception as e:
|
||||
print(f"✗ Failed to generate metrics comparison plot: {e}")
|
||||
|
||||
# 3. Drawdown Analysis
|
||||
try:
|
||||
plot_drawdown(
|
||||
results,
|
||||
ticker,
|
||||
output_path=str(output_path / f"{ticker}_drawdown.png")
|
||||
)
|
||||
except Exception as e:
|
||||
print(f"✗ Failed to generate drawdown plot: {e}")
|
||||
|
||||
# 4. Transaction History (if TradingAgents results available)
|
||||
if "TradingAgents" in results:
|
||||
try:
|
||||
plot_transaction_history(
|
||||
results["TradingAgents"],
|
||||
ticker,
|
||||
strategy_name="TradingAgents",
|
||||
output_path=str(output_path / f"{ticker}_TradingAgents_transactions.png")
|
||||
)
|
||||
except Exception as e:
|
||||
print(f"✗ Failed to generate transaction history plot: {e}")
|
||||
|
||||
# 5. Returns Distribution
|
||||
try:
|
||||
plot_returns_distribution(
|
||||
results,
|
||||
ticker,
|
||||
output_path=str(output_path / f"{ticker}_returns_distribution.png")
|
||||
)
|
||||
except Exception as e:
|
||||
print(f"✗ Failed to generate returns distribution plot: {e}")
|
||||
|
||||
print(f"\n✓ All visualizations saved to: {output_dir}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
# Example usage / testing
|
||||
print("Visualization module loaded successfully!")
|
||||
print("\nAvailable functions:")
|
||||
print(" - plot_cumulative_returns")
|
||||
print(" - plot_transaction_history")
|
||||
print(" - plot_metrics_comparison")
|
||||
print(" - plot_drawdown")
|
||||
print(" - plot_returns_distribution")
|
||||
print(" - create_summary_report")
|
||||
|
||||
|
|
@ -24,3 +24,6 @@ rich
|
|||
questionary
|
||||
langchain_anthropic
|
||||
langchain-google-genai
|
||||
matplotlib
|
||||
seaborn
|
||||
numpy
|
||||
|
|
|
|||
Loading…
Reference in New Issue