Choosing and Building Continuous Futures Series for Reliable CTA Backtests

Summary

The article explains why futures research often uses constructed continuous series: individual contracts expire, so a longer history requires rules for selecting and joining successive contracts. A direct splice can create an artificial price gap at a roll, potentially triggering trend signals or distorting simulated profit and loss. It distinguishes three series: an open-interest-weighted index series for broad market observation, and forward- and backward-adjusted main-contract series that remove roll gaps while preserving different ends of the price history.

For CTA backtests, it generally favors forward adjustment because current prices remain aligned with the tradable lead contract, while historical prices shift. It warns that cumulative additive adjustments can distort distant price levels and that reconstructed histories may change when roll dates or spreads change. Main-contract selection rules, roll timing, and adjustment conventions also explain why vendors’ series differ. The article describes a platform workflow for loading individual contracts and rebuilding continuous data, but offers no empirical comparison of strategies across series. Researchers should validate the data construction rules and not treat a synthetic series as an executable contract.

Key ideas

• A continuous futures series joins expiring contracts according to explicit selection and roll rules.
• Unadjusted splicing can create artificial gaps that affect trend signals and backtest P&L.
• An index series summarizes the broader contract market but may not match any tradable contract price.
• Forward adjustment preserves the latest contract price while shifting history; backward adjustment preserves earlier prices while shifting later data.
• Roll selection, timing, and adjustment rules can cause vendor series to differ, and updated inputs may change reconstructed history.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.