Alpha Builder Documentation
Harness the power of AI to build, analyze, and optimize your investment strategies
1. Portfolio Creation
Alpha Builder (powered by Boosted.ai) allows users to build AI-driven portfolios through three main stages: Name and Details, Portfolio Preferences, and Optimization Goals. Each portfolio is created by setting parameters that guide how the AI selects, weights, and manages securities.
Core Settings
- Long Positions: Define the number and percentage of stocks the AI will take long positions in.
- Short Positions: Define parameters for stocks the AI will short.
- Portfolio Setup: Determine the investment horizon, starting value, trading cost, and additional constraints (e.g., stop-loss, take-profit triggers).
- Optimization Goals: Choose how the AI optimizes—most users select Relative Performance (Alpha) to outperform benchmarks.
The Boosted.ai engine applies machine learning models trained on financial data to generate signals and rankings, then optimizes them using algorithms to balance return and risk.
2. Ranking Analysis
Once the model is trained, it produces Rankings — an ordered list of securities from most to least attractive. This ranking is the foundation for portfolio construction and performance measurement.
How Rankings Work
- Each stock is analyzed across fundamental, technical, and macro drivers.
- Every driver receives a positive or negative explain score.
- The AI calculates a total explain score and assigns a rating (e.g., 5-star, A–F).
- High-scoring stocks become buy candidates; low-scoring ones become sell candidates.
Tear Sheet Metrics Overview
The Tear Sheet summarizes the model's predictive accuracy, risk-adjusted returns, and total performance. These results show that the AI-generated portfolio significantly outperformed the benchmark while maintaining a strong Sharpe ratio, indicating efficient risk management.
Risk Adjusted Returns
| Metric | Meaning | Your Portfolio |
|---|---|---|
| Sharpe Ratio | Measures return per unit of volatility. | 0.9854 |
| Beta | Correlation to the market. <1 means less volatile. | 0.9622 |
| Information Ratio | Excess return per unit of tracking error vs benchmark. | 0.7516 |
| Treynor Ratio | Return earned per unit of systematic risk. | 0.1904 |
Accuracy
| Metric | Meaning | Value |
|---|---|---|
| Number of Predictions | Total trades/signals generated. | 3,240 |
| Correct Predictions | Trades that outperformed the benchmark. | 1,700 |
| Information Coefficient | Predictive strength of predictions. | 0.0494 |
| Up / Down Periods | Positive vs negative trading days. | 2,491 / 1,984 |
| Up/Down Ratio | 0.56 % |
Return
| Metric | Description | Value |
|---|---|---|
| Total Return | Overall return of the portfolio. | 2049.53 % |
| Annualized Return | Average yearly gain. | 18.82 % |
| Benchmark Return | Performance of the benchmark. | 536.29 % |
| Benchmark Annualized Return | Annualized return of the benchmark. | 10.96 % |
| Avg. Actual Dividend Yield | Dividend income from holdings. | 1.68 % |
| Avg. Benchmark Dividend Yield | 1.89 % |
In Summary
| Component | Function | Output |
|---|---|---|
| Portfolio Creation | Defines investment constraints, goals, and stock universe. | Optimized AI portfolio. |
| Ranking Analysis | Evaluates and measures AI model's predictive accuracy. | Tear Sheet metrics (Sharpe, Returns, Accuracy). |