ML Monthly Retraining

Automated monthly workflow: train 4 XGBoost models, backtest OOS, compare, deploy best.

Prerequisites

Before starting, verify:

1. Data files exist: ../data/GLBX/nq_2m_1y.parquet and nq_5m_1y.parquet
2. Data covers at least the last 4 months (3 training + 1 test)
3. Working directory is Python/

If data is stale or missing, tell the user to download fresh data from Databento:

• Schema: ohlcv-1m (aggregate to 2m/5m) or ohlcv-1s for tick-level
• Dataset: GLBX.MDP3
• Symbol: NQ.FUT (front month continuous)

Workflow

Step 1: Pre-flight checks