Backtest Results
Historical performance analysis of our trading strategies. All results are based on out-of-sample backtesting with no lookahead bias.
Performance Metrics Overview
Comprehensive performance comparison across all strategies
| Metric | Position-Sized | Consensus | Buy & Hold |
|---|---|---|---|
| Annual Return | 10.07% | 11.29% | 8.46% |
| Sharpe Ratio | 0.78 | 0.87 | 0.47 |
| Sortino Ratio | 0.92 | 0.96 | 0.60 |
| Calmar Ratio | 0.29 | 0.41 | 0.15 |
| Max Drawdown | -34.67% | -27.62% | -56.78% |
| Annual Volatility | 12.84% | 13.00% | 18.08% |
| Win Rate | 65.22% | 81.58% | 0.00% |
| Total Trades | 23 | 38 | N/A |
| Backtest Period | 35.8 years | 35.8 years | 35.8 years |
Performance Analysis
Comprehensive comparison of strategy performance
Position-Sized
+2991.45%
Consensus
+4480.19%
Buy & Hold
+1750.13%
Peak-to-trough declines over time
252-day rolling returns comparison
Model Analysis
Historical model prediction probabilities by horizon
Strategy Performance
Historical backtest performance vs Buy & Hold
11.29%
0.87
-27.62%
Win Rate
81.58%
Total Trades
38
Time in Market
0.00%
Historical backtest performance vs Buy & Hold
10.07%
0.78
-34.67%
Win Rate
65.22%
Total Trades
23
Time in Market
0.00%
Strategy Descriptions
Binary approach that invests 100% when the consensus threshold is met.
Rules:
- • ≥2 models signal LONG → 100% invested
- • <2 models signal LONG → 100% cash
- • Simple, easy to implement
Best for: Traders who prefer binary decisions and can tolerate higher volatility.
Gradual approach that varies position size based on model agreement.
Rules:
- • 3 models LONG → 100% position
- • 2 models LONG → 90% position
- • 1 model LONG → 30% position
- • 0 models LONG → 0% (cash)
Best for: Traders who prefer gradual adjustments and smoother equity curves.
Model Training Insights
Classification performance on test set and cross-validation results
| Horizon | AUC | Accuracy | Precision | Recall | F1 Score | CV AUC | Features |
|---|---|---|---|---|---|---|---|
3M | 0.753 | 0.659 | 0.814 | 0.635 | 0.714 | 0.680 ± 0.234 | 10 |
6M | 0.884 | 0.774 | 0.916 | 0.753 | 0.827 | 0.834 ± 0.125 | 10 |
12M | 0.975 | 0.912 | 0.975 | 0.909 | 0.941 | 0.978 ± 0.034 | 10 |
AUC: Target ≥ 0.60 (higher = better predictive power)
Precision: Accuracy of LONG predictions (higher = fewer false signals)
Recall: Ability to identify opportunities (higher = catches more good periods)
CV AUC: Cross-validation performance (lower std = more stable)
Confusion Matrices
How well each model classifies LONG vs CASH signals
3M Horizon
Accuracy: 65.9%
Actual: CASH ✓
Actual: CASH ✗
Actual: LONG ✗
Actual: LONG ✓
6M Horizon
Accuracy: 77.4%
Actual: CASH ✓
Actual: CASH ✗
Actual: LONG ✗
Actual: LONG ✓
12M Horizon
Accuracy: 91.2%
Actual: CASH ✓
Actual: CASH ✗
Actual: LONG ✗
Actual: LONG ✓
Cross-Validation Results
Train vs Validation AUC per fold - detecting overfitting
3M Horizon CV Results
5 folds completed
6M Horizon CV Results
5 folds completed
12M Horizon CV Results
3 folds completed
Feature Importance
Top 10 most important features for each prediction horizon
Top 10 Features - 3M Horizon
Most important features for prediction
Top 10 Features - 6M Horizon
Most important features for prediction
Top 10 Features - 12M Horizon
Most important features for prediction
Backtest Methodology
Data & Approach
- ✓Out-of-sample testing (no lookahead bias)
- ✓Walk-forward validation
- ✓Models trained on historical data only
- ✓Signals lagged by 1 day (realistic execution)
Key Metrics
- Annual Return: Annualized total return
- Sharpe Ratio: Risk-adjusted returns (higher is better)
- Max Drawdown: Largest peak-to-trough decline
- Win Rate: Percentage of profitable trades
Want to learn more about our models and methodology? View detailed methodology →