Backtest Custom Ranking Models

Step by step instructions to generate your own Custom Ranking Model:

1. Prepare your Stock Universe (Screener)

This is the most important first step — define which stocks the ranking model will be applied to.

• Go to the Stock Screener

• Create a new screener specifically for this task

• Add minimal requirement filters like a price floor, volume floor, market cap floor, etc.

• See this example here for top US stocks (click Actions > Duplicate to make it your own and use it in the ranking model)

• Filtering is key to making this work - minimal requirements should be broad enough to allow the ranking model to work on a large selection of stocks, but also make sense for the purposes of the ranking model itself

2. Create Your Custom Ranking Model (Google Sheets)

Ranking models are just a list of metrics + importance scores. Use Legacy Tools > Ranking Model for inspiration.

• Open Google Sheets and create a new blank spreadsheet.

• In Sheet1 (default sheet), create exactly two columns:

  • Column A: metric

  • Column B: score (importance weight; can be positive or negative)

• Fill in your metrics and scores.

Example for a simple Momentum model (as shown):

Metric: sharpe_ratio → Score: 10

Metric: z-score_relative_2y → Score: 10

(You can use more metrics. Download the reference table linked below the video for all available metric names — use the exact “table metric type” names.)

• Set sharing: Anyone with the link can view.

• Copy the shareable link (Google Sheets URL).

Pro tip: Try using an LLM (like Grok or ChatGPT) to help design good combinations of metrics and weights.

Here is a Google Sheet example: https://docs.google.com/spreadsheets/d/1nteDBLRh-6VZnQboCIoPtZvXCtmNn6_NdRo0vPi8STI/edit?usp=sharing

3. Set Up the Backtest

Time to combine the screener + ranking model in the backtester.

• Go to any Ranking Model Strategy like Millennium Alpha, Momentum, Vision or Low Beta

• Click “Replicate Strategy”

• Choose “Custom Filter” and select your screener and screener tab from the drop-down menus

• Choose “Custom Ranking Model”

• Paste the Google Sheets URL in the field

• Set up the rest of the backtest as usual (rebalancing frequency, number of positions, risk controls, benchmark)

4. Run the Backtest

• Launch the backtest.

• The first run with a new ranking model takes longer (it has to rank the universe). Subsequent runs using the same ranking model are much faster (cached as long as you don’t delete the original backtest).

5. Review and Iterate

Iterate by:

• Changing the Google Sheet (metrics/weights) → re-run.

• Adjusting screener filters.

• Testing different stop losses, rebalancing frequency, position count, etc.

• Reusing the same ranking model across many backtests without re-uploading.

Additional Important Notes

• For US stocks: Highly recommended to add strong pre-filters (price > $5–10, volume floor, market cap floor, or limit to top 1,500 by dollar volume) before applying your ranking model.

• The ranking model ranks the entire available universe first, then the screener filters are applied, then the top N stocks are selected.

• You can explore the legacy ranking models (Millennium Alpha, Momentum, Vision, Low Beta) to see examples of good metric combinations.