StockRanker Workflow for Momentum-Based Equity Ranking
Summary
The document shows a stock-ranking workflow that derives 30-day and 90-day returns, ranks securities cross-sectionally, filters the universe, and creates a forward-return label for model training. A StockRanker module trains on historical data and predicts scores for a later period. A score-to-position step then selects a fixed number of names and assigns equal weights, while a trading handler sells securities outside the target set and rebalances the retained and selected positions.
The post itself is a troubleshooting request: execution fails when the StockRanker module attempts to start a process, with an operating-system error indicating that a required executable cannot be found. No resolution or model performance evidence is provided. The displayed date ranges and parameters describe this particular example; they do not demonstrate predictive quality. The forward-return label and time-separated prediction period are part of the workflow, but the post does not establish whether the setup avoids all leakage or produces robust out-of-sample results.
Key ideas
- The feature pipeline ranks recent returns and filters stocks before model training.
- A forward-return label is used to train a StockRanker model on historical observations.
- Predicted scores feed a selection step that creates equal-weight target positions.
- The trading handler submits exits for names outside the target set and orders for target holdings.
- The reported failure is a missing executable during process launch, and no fix or performance results are supplied.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.