Signal model, pick scoring, and risk notes

1. How signals are produced

The model computes momentum and macro regime features on daily bars, then classifies each day as BUY, SELL, or HOLD using threshold rules. Confidence reflects signal strength, not certainty.

2. Public pick scoring

Public stock picks are ranked deterministically from optimized signal history, signal recency, price freshness, market-cap liquidity, fundamentals, and risk penalties. Crypto picks use top-liquid assets only and combine trend, liquidity, volatility, market-cap rank, and source freshness.

3. Main inputs

• Price momentum from split-adjusted daily candles.
• Macro context from FRED series (rates, curve, inflation, labor, recession).
• Fundamentals and events from cached company, earnings, and recommendation data.
• Crypto price, volume, market cap, and multi-period return data from cached provider responses.
• User-selected controls: sensitivity, horizon, smoothing, and risk exits.

4. Risk controls in backtests

• Take-profit and stop-loss percentages.
• Maximum holding window before forced exit.
• Direction-adjusted win rate and average return on closed trades.

5. Important limitations

• Backtests can overstate future performance.
• No transaction costs, taxes, or slippage are modeled.
• Liquidity, spreads, and portfolio constraints are not included.
• Provider outages or stale data can affect outputs.
• Model parameters can drift as market regimes change.
• Crypto prices can diverge across venues and can move sharply outside normal market hours.
• Prediction-market prices are context only and are not the primary pick driver.