TradingAgents/tests/unit/portfolio/test_performance.py

760 lines
27 KiB
Python

"""Tests for Portfolio Performance Metrics module.
Issue #31: [PORT-30] Performance metrics - Sharpe, drawdown, returns
"""
import pytest
from datetime import date, timedelta
from decimal import Decimal
from typing import List, Tuple
from tradingagents.portfolio import (
Period,
ReturnSeries,
DrawdownInfo,
TradeStats,
PerformanceMetrics,
PerformanceCalculator,
calculate_cagr,
calculate_rolling_returns,
calculate_monthly_returns,
calculate_yearly_returns,
)
# =============================================================================
# Test Fixtures
# =============================================================================
@pytest.fixture
def calculator():
"""Create a PerformanceCalculator with default settings."""
return PerformanceCalculator(risk_free_rate=Decimal("0.05"))
@pytest.fixture
def simple_returns():
"""Create a simple return series for testing."""
base_date = date(2024, 1, 1)
returns = [
(base_date + timedelta(days=i), Decimal(str(r)))
for i, r in enumerate([
0.01, -0.005, 0.02, 0.005, -0.01,
0.015, -0.003, 0.008, 0.012, -0.007,
])
]
return ReturnSeries(returns=returns, period=Period.DAILY)
@pytest.fixture
def positive_returns():
"""Create a positive return series."""
base_date = date(2024, 1, 1)
returns = [
(base_date + timedelta(days=i), Decimal("0.01"))
for i in range(20)
]
return ReturnSeries(returns=returns, period=Period.DAILY)
@pytest.fixture
def negative_returns():
"""Create a negative return series."""
base_date = date(2024, 1, 1)
returns = [
(base_date + timedelta(days=i), Decimal("-0.01"))
for i in range(20)
]
return ReturnSeries(returns=returns, period=Period.DAILY)
@pytest.fixture
def portfolio_values():
"""Create a portfolio value series."""
base_date = date(2024, 1, 1)
values = [
1000, 1010, 1005, 1025, 1030,
1020, 1035, 1032, 1040, 1052,
1045, 1060, 1055, 1070, 1080,
]
return [
(base_date + timedelta(days=i), Decimal(str(v)))
for i, v in enumerate(values)
]
@pytest.fixture
def drawdown_values():
"""Create a portfolio series with drawdowns."""
base_date = date(2024, 1, 1)
values = [
1000, 1050, 1100, 1080, 1000, # First drawdown
900, 950, 1000, 1100, 1150, # Second drawdown + recovery
1100, 1050, 1000, 1050, 1100, # Third drawdown + recovery
]
return [
(base_date + timedelta(days=i), Decimal(str(v)))
for i, v in enumerate(values)
]
@pytest.fixture
def trade_returns():
"""Create sample trade returns."""
return [
Decimal("0.10"), # 10% win
Decimal("-0.05"), # 5% loss
Decimal("0.08"), # 8% win
Decimal("0.15"), # 15% win
Decimal("-0.03"), # 3% loss
Decimal("0.00"), # breakeven
Decimal("-0.08"), # 8% loss
Decimal("0.12"), # 12% win
Decimal("0.05"), # 5% win
Decimal("-0.02"), # 2% loss
]
# =============================================================================
# ReturnSeries Tests
# =============================================================================
class TestReturnSeries:
"""Test ReturnSeries dataclass."""
def test_return_series_creation(self, simple_returns):
"""Test basic return series creation."""
assert simple_returns.period == Period.DAILY
assert simple_returns.annualization_factor == 252
assert simple_returns.num_periods == 10
def test_values_property(self, simple_returns):
"""Test getting just the return values."""
values = simple_returns.values
assert len(values) == 10
assert values[0] == Decimal("0.01")
assert values[1] == Decimal("-0.005")
def test_dates_property(self, simple_returns):
"""Test getting just the dates."""
dates = simple_returns.dates
assert len(dates) == 10
assert dates[0] == date(2024, 1, 1)
assert dates[1] == date(2024, 1, 2)
def test_annualization_factor_daily(self):
"""Test daily annualization factor."""
rs = ReturnSeries(returns=[], period=Period.DAILY)
assert rs.annualization_factor == 252
def test_annualization_factor_weekly(self):
"""Test weekly annualization factor."""
rs = ReturnSeries(returns=[], period=Period.WEEKLY)
assert rs.annualization_factor == 52
def test_annualization_factor_monthly(self):
"""Test monthly annualization factor."""
rs = ReturnSeries(returns=[], period=Period.MONTHLY)
assert rs.annualization_factor == 12
def test_annualization_factor_quarterly(self):
"""Test quarterly annualization factor."""
rs = ReturnSeries(returns=[], period=Period.QUARTERLY)
assert rs.annualization_factor == 4
def test_annualization_factor_yearly(self):
"""Test yearly annualization factor."""
rs = ReturnSeries(returns=[], period=Period.YEARLY)
assert rs.annualization_factor == 1
# =============================================================================
# Return Calculation Tests
# =============================================================================
class TestReturnCalculations:
"""Test return calculation methods."""
def test_calculate_returns_from_values(self, calculator, portfolio_values):
"""Test calculating returns from portfolio values."""
returns = calculator.calculate_returns(portfolio_values)
assert returns.num_periods == len(portfolio_values) - 1
# First return: (1010 - 1000) / 1000 = 0.01
assert returns.values[0] == Decimal("0.01")
def test_calculate_returns_empty(self, calculator):
"""Test calculating returns from empty values."""
returns = calculator.calculate_returns([])
assert returns.num_periods == 0
def test_calculate_returns_single_value(self, calculator):
"""Test calculating returns from single value."""
returns = calculator.calculate_returns([(date.today(), Decimal("100"))])
assert returns.num_periods == 0
def test_total_return(self, calculator, simple_returns):
"""Test total cumulative return calculation."""
total = calculator.total_return(simple_returns)
# Manual calculation: (1+0.01)*(1-0.005)*(1+0.02)*... - 1
expected = Decimal("1")
for r in simple_returns.values:
expected *= (Decimal("1") + r)
expected -= Decimal("1")
assert abs(total - expected) < Decimal("0.0001")
def test_total_return_empty(self, calculator):
"""Test total return with empty series."""
empty = ReturnSeries(returns=[], period=Period.DAILY)
assert calculator.total_return(empty) == Decimal("0")
def test_total_return_positive_only(self, calculator, positive_returns):
"""Test total return with all positive returns."""
total = calculator.total_return(positive_returns)
# 20 days of 1% each: (1.01)^20 - 1 ≈ 0.2202
assert total > Decimal("0.20")
assert total < Decimal("0.25")
def test_total_return_negative_only(self, calculator, negative_returns):
"""Test total return with all negative returns."""
total = calculator.total_return(negative_returns)
# 20 days of -1% each: (0.99)^20 - 1 ≈ -0.1821
assert total < Decimal("-0.15")
assert total > Decimal("-0.25")
def test_annualized_return(self, calculator, simple_returns):
"""Test annualized return calculation."""
ann_return = calculator.annualized_return(simple_returns)
# Should be positive for our simple returns
assert ann_return > Decimal("-1")
assert ann_return < Decimal("10") # Reasonable bound
def test_annualized_return_empty(self, calculator):
"""Test annualized return with empty series."""
empty = ReturnSeries(returns=[], period=Period.DAILY)
assert calculator.annualized_return(empty) == Decimal("0")
# =============================================================================
# Volatility Tests
# =============================================================================
class TestVolatility:
"""Test volatility calculation methods."""
def test_volatility_calculation(self, calculator, simple_returns):
"""Test basic volatility calculation."""
vol = calculator.volatility(simple_returns)
# Volatility should be positive
assert vol > Decimal("0")
# Annualized volatility typically 10-50% for equities
assert vol < Decimal("5") # Reasonable upper bound
def test_volatility_not_annualized(self, calculator, simple_returns):
"""Test non-annualized volatility."""
vol_ann = calculator.volatility(simple_returns, annualize=True)
vol_not_ann = calculator.volatility(simple_returns, annualize=False)
# Annualized should be higher by sqrt(252) factor
assert vol_ann > vol_not_ann
def test_volatility_zero_variance(self, calculator):
"""Test volatility with constant returns."""
constant = ReturnSeries(
returns=[(date.today() + timedelta(days=i), Decimal("0.01")) for i in range(10)],
period=Period.DAILY,
)
vol = calculator.volatility(constant)
assert vol == Decimal("0")
def test_volatility_insufficient_data(self, calculator):
"""Test volatility with insufficient data."""
single = ReturnSeries(
returns=[(date.today(), Decimal("0.01"))],
period=Period.DAILY,
)
assert calculator.volatility(single) == Decimal("0")
def test_downside_deviation(self, calculator, simple_returns):
"""Test downside deviation calculation."""
downside = calculator.downside_deviation(simple_returns)
# Downside should be positive and <= total volatility
assert downside >= Decimal("0")
def test_downside_deviation_positive_only(self, calculator, positive_returns):
"""Test downside deviation with only positive returns."""
downside = calculator.downside_deviation(positive_returns)
# No downside when all returns are positive
assert downside == Decimal("0")
# =============================================================================
# Risk-Adjusted Metrics Tests
# =============================================================================
class TestRiskAdjustedMetrics:
"""Test risk-adjusted performance metrics."""
def test_sharpe_ratio(self, calculator, simple_returns):
"""Test Sharpe ratio calculation."""
sharpe = calculator.sharpe_ratio(simple_returns)
# Sharpe can be positive, negative, or zero
assert isinstance(sharpe, Decimal)
def test_sharpe_ratio_zero_volatility(self, calculator):
"""Test Sharpe ratio with zero volatility."""
constant = ReturnSeries(
returns=[(date.today() + timedelta(days=i), Decimal("0.01")) for i in range(10)],
period=Period.DAILY,
)
sharpe = calculator.sharpe_ratio(constant)
assert sharpe == Decimal("0")
def test_sortino_ratio(self, calculator, simple_returns):
"""Test Sortino ratio calculation."""
sortino = calculator.sortino_ratio(simple_returns)
# Sortino should be defined
assert isinstance(sortino, Decimal)
def test_sortino_vs_sharpe(self, calculator, simple_returns):
"""Test that Sortino differs from Sharpe."""
sharpe = calculator.sharpe_ratio(simple_returns)
sortino = calculator.sortino_ratio(simple_returns)
# For asymmetric returns, Sortino should differ from Sharpe
# (unless all returns are below MAR or there's no downside)
# Just check both are calculated
assert sharpe != Decimal("0") or sortino != Decimal("0") or True
def test_calmar_ratio(self, calculator, simple_returns):
"""Test Calmar ratio calculation."""
calmar = calculator.calmar_ratio(simple_returns)
# Calmar should be defined
assert isinstance(calmar, Decimal)
# =============================================================================
# Drawdown Tests
# =============================================================================
class TestDrawdownAnalysis:
"""Test drawdown analysis methods."""
def test_max_drawdown_calculation(self, calculator):
"""Test maximum drawdown calculation."""
# Create a series with a known drawdown
# Start at 1, go to 1.1, drop to 0.9, recover to 1.05
# Max DD = (0.9 - 1.1) / 1.1 = -0.1818
cum_returns = [
Decimal("0"), # 1.0
Decimal("0.10"), # 1.1
Decimal("-0.10"), # 0.9
Decimal("0.05"), # 1.05
]
max_dd = calculator.max_drawdown(cum_returns)
# Max DD should be around -18%
assert max_dd < Decimal("-0.15")
assert max_dd > Decimal("-0.25")
def test_max_drawdown_no_drawdown(self, calculator):
"""Test max drawdown with no drawdown."""
# Monotonically increasing returns
cum_returns = [Decimal("0.01") * i for i in range(1, 11)]
max_dd = calculator.max_drawdown(cum_returns)
assert max_dd == Decimal("0")
def test_max_drawdown_empty(self, calculator):
"""Test max drawdown with empty series."""
assert calculator.max_drawdown([]) == Decimal("0")
def test_drawdown_series(self, calculator, drawdown_values):
"""Test drawdown series calculation."""
dd_series = calculator.drawdown_series(drawdown_values)
assert len(dd_series) == len(drawdown_values)
# First value should have 0 drawdown
assert dd_series[0][1] == Decimal("0")
def test_find_drawdowns(self, calculator, drawdown_values):
"""Test finding drawdown periods."""
drawdowns = calculator.find_drawdowns(drawdown_values, min_drawdown=Decimal("-0.03"))
assert len(drawdowns) > 0
for dd in drawdowns:
assert isinstance(dd, DrawdownInfo)
assert dd.max_drawdown < Decimal("0")
def test_drawdown_info_properties(self, calculator, drawdown_values):
"""Test DrawdownInfo properties."""
drawdowns = calculator.find_drawdowns(drawdown_values, min_drawdown=Decimal("-0.05"))
if drawdowns:
dd = drawdowns[0]
assert dd.start_date <= dd.trough_date
assert dd.peak_value > dd.trough_value
assert dd.duration_days >= 0
# =============================================================================
# Trade Statistics Tests
# =============================================================================
class TestTradeStatistics:
"""Test trade-level statistics."""
def test_trade_statistics(self, calculator, trade_returns):
"""Test basic trade statistics calculation."""
stats = calculator.trade_statistics(trade_returns)
assert stats.total_trades == 10
assert stats.winning_trades == 5
assert stats.losing_trades == 4
assert stats.breakeven_trades == 1
def test_win_rate(self, calculator, trade_returns):
"""Test win rate calculation."""
stats = calculator.trade_statistics(trade_returns)
# 5 wins out of 10 trades = 50%
assert stats.win_rate == Decimal("50.00")
def test_profit_factor(self, calculator, trade_returns):
"""Test profit factor calculation."""
stats = calculator.trade_statistics(trade_returns)
# Gross profit / Gross loss
# Profits: 0.10 + 0.08 + 0.15 + 0.12 + 0.05 = 0.50
# Losses: 0.05 + 0.03 + 0.08 + 0.02 = 0.18
# PF = 0.50 / 0.18 ≈ 2.78
assert stats.profit_factor > Decimal("2")
assert stats.profit_factor < Decimal("3")
def test_average_win_loss(self, calculator, trade_returns):
"""Test average win and loss calculation."""
stats = calculator.trade_statistics(trade_returns)
# Average win: 0.50 / 5 = 0.10
assert stats.avg_win == Decimal("0.1000")
# Average loss: -0.18 / 4 = -0.045
assert stats.avg_loss < Decimal("0")
def test_largest_win_loss(self, calculator, trade_returns):
"""Test largest win and loss identification."""
stats = calculator.trade_statistics(trade_returns)
assert stats.largest_win == Decimal("0.15")
assert stats.largest_loss == Decimal("-0.08")
def test_expectancy(self, calculator, trade_returns):
"""Test expectancy calculation."""
stats = calculator.trade_statistics(trade_returns)
# Expectancy = (win_rate * avg_win) + (loss_rate * avg_loss)
assert isinstance(stats.expectancy, Decimal)
# Should be positive for our sample (more wins than losses by amount)
assert stats.expectancy > Decimal("0")
def test_empty_trades(self, calculator):
"""Test statistics with no trades."""
stats = calculator.trade_statistics([])
assert stats.total_trades == 0
assert stats.win_rate == Decimal("0")
assert stats.profit_factor == Decimal("0")
# =============================================================================
# Benchmark Comparison Tests
# =============================================================================
class TestBenchmarkComparison:
"""Test benchmark comparison methods."""
def test_benchmark_comparison(self, calculator):
"""Test basic benchmark comparison."""
base_date = date(2024, 1, 1)
portfolio = ReturnSeries(
returns=[(base_date + timedelta(days=i), Decimal(str(r)))
for i, r in enumerate([0.02, -0.01, 0.03, 0.01, -0.02])],
period=Period.DAILY,
)
benchmark = ReturnSeries(
returns=[(base_date + timedelta(days=i), Decimal(str(r)))
for i, r in enumerate([0.01, -0.005, 0.02, 0.005, -0.01])],
period=Period.DAILY,
)
comparison = calculator.benchmark_comparison(portfolio, benchmark)
assert "alpha" in comparison
assert "beta" in comparison
assert "information_ratio" in comparison
assert "tracking_error" in comparison
def test_beta_calculation(self, calculator):
"""Test beta calculation."""
base_date = date(2024, 1, 1)
# Portfolio moves 2x the benchmark
benchmark_rets = [0.01, -0.01, 0.02, -0.02, 0.015]
portfolio_rets = [0.02, -0.02, 0.04, -0.04, 0.03]
portfolio = ReturnSeries(
returns=[(base_date + timedelta(days=i), Decimal(str(r)))
for i, r in enumerate(portfolio_rets)],
period=Period.DAILY,
)
benchmark = ReturnSeries(
returns=[(base_date + timedelta(days=i), Decimal(str(r)))
for i, r in enumerate(benchmark_rets)],
period=Period.DAILY,
)
comparison = calculator.benchmark_comparison(portfolio, benchmark)
# Beta should be approximately 2
assert comparison["beta"] > Decimal("1.5")
assert comparison["beta"] < Decimal("2.5")
def test_mismatched_periods(self, calculator):
"""Test benchmark comparison with mismatched periods."""
base_date = date(2024, 1, 1)
portfolio = ReturnSeries(
returns=[(base_date + timedelta(days=i), Decimal("0.01")) for i in range(10)],
period=Period.DAILY,
)
benchmark = ReturnSeries(
returns=[(base_date + timedelta(days=i), Decimal("0.01")) for i in range(5)],
period=Period.DAILY,
)
with pytest.raises(ValueError, match="same number of periods"):
calculator.benchmark_comparison(portfolio, benchmark)
# =============================================================================
# Complete Metrics Tests
# =============================================================================
class TestCalculateMetrics:
"""Test complete performance metrics calculation."""
def test_calculate_metrics(self, calculator, simple_returns, trade_returns):
"""Test full metrics calculation."""
metrics = calculator.calculate_metrics(
simple_returns,
trade_returns=trade_returns,
)
assert isinstance(metrics, PerformanceMetrics)
assert metrics.start_date == simple_returns.dates[0]
assert metrics.end_date == simple_returns.dates[-1]
assert isinstance(metrics.total_return, Decimal)
assert isinstance(metrics.sharpe_ratio, Decimal)
assert metrics.trade_stats is not None
def test_calculate_metrics_empty(self, calculator):
"""Test metrics with empty series."""
empty = ReturnSeries(returns=[], period=Period.DAILY)
metrics = calculator.calculate_metrics(empty)
assert metrics.total_return == Decimal("0")
assert metrics.sharpe_ratio == Decimal("0")
assert metrics.num_drawdowns == 0
def test_calculate_metrics_with_benchmark(self, calculator, simple_returns):
"""Test metrics with benchmark."""
base_date = date(2024, 1, 1)
benchmark = ReturnSeries(
returns=[(base_date + timedelta(days=i), Decimal("0.005"))
for i in range(10)],
period=Period.DAILY,
)
metrics = calculator.calculate_metrics(simple_returns, benchmark_returns=benchmark)
assert metrics.benchmark_alpha is not None
assert metrics.benchmark_beta is not None
assert metrics.information_ratio is not None
def test_best_worst_day(self, calculator, simple_returns):
"""Test best and worst day identification."""
metrics = calculator.calculate_metrics(simple_returns)
assert metrics.best_day == Decimal("0.02")
assert metrics.worst_day == Decimal("-0.01")
def test_positive_negative_periods(self, calculator, simple_returns):
"""Test counting positive and negative periods."""
metrics = calculator.calculate_metrics(simple_returns)
# Count manually: 0.01, -0.005, 0.02, 0.005, -0.01, 0.015, -0.003, 0.008, 0.012, -0.007
# Positive: 6, Negative: 4
assert metrics.positive_periods == 6
assert metrics.negative_periods == 4
# =============================================================================
# Utility Function Tests
# =============================================================================
class TestUtilityFunctions:
"""Test utility functions."""
def test_calculate_cagr(self):
"""Test CAGR calculation."""
# Start: 1000, End: 1610.51, Years: 5
# CAGR = (1610.51/1000)^(1/5) - 1 = 0.10 (10%)
cagr = calculate_cagr(
Decimal("1000"),
Decimal("1610.51"),
Decimal("5"),
)
assert abs(cagr - Decimal("0.10")) < Decimal("0.01")
def test_calculate_cagr_zero_start(self):
"""Test CAGR with zero start value."""
cagr = calculate_cagr(Decimal("0"), Decimal("100"), Decimal("5"))
assert cagr == Decimal("0")
def test_calculate_cagr_zero_years(self):
"""Test CAGR with zero years."""
cagr = calculate_cagr(Decimal("100"), Decimal("200"), Decimal("0"))
assert cagr == Decimal("0")
def test_calculate_rolling_returns(self, simple_returns):
"""Test rolling returns calculation."""
rolling = calculate_rolling_returns(simple_returns, window=3)
# Should have num_periods - window + 1 values
assert len(rolling) == simple_returns.num_periods - 3 + 1
def test_calculate_rolling_returns_window_too_large(self, simple_returns):
"""Test rolling returns with window larger than series."""
rolling = calculate_rolling_returns(simple_returns, window=100)
assert len(rolling) == 0
def test_calculate_monthly_returns(self):
"""Test monthly return aggregation."""
# Create daily returns spanning multiple months
returns_data = []
for month in [1, 2, 3]:
for day in range(1, 11):
dt = date(2024, month, day)
returns_data.append((dt, Decimal("0.001")))
daily = ReturnSeries(returns=returns_data, period=Period.DAILY)
monthly = calculate_monthly_returns(daily)
assert len(monthly) == 3
assert (2024, 1) in monthly
assert (2024, 2) in monthly
assert (2024, 3) in monthly
def test_calculate_monthly_returns_wrong_period(self):
"""Test monthly aggregation with wrong input period."""
weekly = ReturnSeries(returns=[], period=Period.WEEKLY)
with pytest.raises(ValueError, match="daily returns"):
calculate_monthly_returns(weekly)
def test_calculate_yearly_returns(self):
"""Test yearly return aggregation."""
# Create daily returns spanning multiple years
returns_data = []
for year in [2023, 2024]:
for i in range(10):
dt = date(year, 1, i + 1)
returns_data.append((dt, Decimal("0.001")))
daily = ReturnSeries(returns=returns_data, period=Period.DAILY)
yearly = calculate_yearly_returns(daily)
assert len(yearly) == 2
assert 2023 in yearly
assert 2024 in yearly
def test_calculate_yearly_returns_wrong_period(self):
"""Test yearly aggregation with wrong input period."""
monthly = ReturnSeries(returns=[], period=Period.MONTHLY)
with pytest.raises(ValueError, match="daily returns"):
calculate_yearly_returns(monthly)
# =============================================================================
# Edge Cases and Error Handling
# =============================================================================
class TestEdgeCases:
"""Test edge cases and error handling."""
def test_calculator_with_zero_risk_free_rate(self):
"""Test calculator with zero risk-free rate."""
calc = PerformanceCalculator(risk_free_rate=Decimal("0"))
assert calc.risk_free_rate == Decimal("0")
def test_calculator_with_custom_mar(self):
"""Test calculator with custom MAR."""
calc = PerformanceCalculator(min_acceptable_return=Decimal("0.05"))
assert calc.min_acceptable_return == Decimal("0.05")
def test_very_small_returns(self, calculator):
"""Test with very small returns."""
tiny = ReturnSeries(
returns=[
(date(2024, 1, i), Decimal("0.0001"))
for i in range(1, 11)
],
period=Period.DAILY,
)
metrics = calculator.calculate_metrics(tiny)
assert isinstance(metrics.total_return, Decimal)
def test_very_large_returns(self, calculator):
"""Test with very large returns."""
large = ReturnSeries(
returns=[
(date(2024, 1, i), Decimal("0.50")) # 50% daily
for i in range(1, 6)
],
period=Period.DAILY,
)
metrics = calculator.calculate_metrics(large)
# Total return should be huge: (1.5)^5 - 1 ≈ 6.59
assert metrics.total_return > Decimal("5")
def test_mixed_positive_negative(self, calculator):
"""Test with alternating returns."""
alternating = ReturnSeries(
returns=[
(date(2024, 1, i), Decimal("0.02") if i % 2 == 0 else Decimal("-0.02"))
for i in range(1, 21)
],
period=Period.DAILY,
)
metrics = calculator.calculate_metrics(alternating)
# Should have roughly zero total return
assert abs(metrics.total_return) < Decimal("0.10")