How Fillipio Scores Every Stock

Every trading day, our pipeline runs before market open. Here's exactly what happens, step by step — no black boxes.

The Daily Pipeline

Step 0100:00 CET

Data Collection

Fresh OHLCV (Open, High, Low, Close, Volume) data is pulled for all 4,949 eligible US stocks and ETFs. We filter for: average 20-day volume > 100K, price > $1, and 200+ trading days of history. This excludes penny stocks and illiquid securities that produce unreliable signals.

Step 0201:00 CET

Technical Indicators (25+)

For each stock, we compute: RSI (14-day), MACD & histogram, Bollinger Bands (20,2), ADX (trend strength), SMA 50 & 200, EMA 12 & 26, ATR (volatility), On-Balance Volume, Stochastic oscillator, Williams %R, Rate of Change, volume ratio vs 20-day average, distance from 52-week high, and multi-timeframe returns (1d, 5d, 20d).

Step 0301:00 CET

ML Prediction (51 Features)

A LightGBM gradient boosted tree model processes 51 features per stock — including all technical indicators plus cross-sectional ranks (how a stock's RSI compares to all other stocks), day-of-week, and statistical features. It predicts 5-day forward returns and direction probability. The model is retrained monthly using walk-forward validation.

Step 0401:00 CET

Composite Scoring

Each stock gets a composite score 0-100. The formula: Technical × 0.35 + ML × 0.35 + Fundamental × 0.30 (when fundamental data exists). Without fundamentals: Technical × 0.50 + ML × 0.50. Scores map to signals: ≥80 STRONG_BUY, 65-79 BUY, 40-64 HOLD, 25-39 SELL, <25 STRONG_SELL.

Step 0501:00 CET

Risk Radar & Signals

The bottom 20 ML-scored stocks are flagged for Risk Radar — these average -3.2% loss over 5 days. Three strategies (Momentum, Value Hunter, Quality Growth) evaluate their entry/exit rules and execute paper trades with stop-losses and take-profit targets.

Step 0603:00 CET

Data Published

All scores, signals, predictions, and portfolio updates are synced to the frontend database. By the time European markets open, everything is fresh. US investors see the data before their morning coffee.

ML Model Performance

Our model is validated rigorously. Here are the key metrics:

81%

Ranking Accuracy

Top picks outperform bottom picks

-3.2%

Avg 5-Day Loss

Bottom 20 stocks avg decline

51

Features

Technical + statistical + temporal

Monthly

Retrained

Walk-forward, no look-ahead bias

Top ML Features (by importance)

  1. 1MACD momentum (rate of change of MACD)
  2. 2On-Balance Volume (OBV)
  3. 320-day price change
  4. 4Momentum rank (cross-sectional)
  5. 5RSI rank (cross-sectional)
  6. 6ATR as % of price (volatility)
  7. 7Bollinger Band width
  8. 8Distance from 52-week high
  9. 9Volume
  10. 10Day of week

See It In Action

The best way to understand Fillipio is to use it. 7-day free trial, full access.