Why a Rolling Factor Can Leave Simulation Signals Empty
Summary
A platform support exchange explains one reason a strategy may produce trades in backtesting but no signals in simulated trading. The strategy’s prediction start and end dates were tied to live-trading parameters, while no lookback days were requested. Because its factor calculation used a rolling average, the simulation prediction set had no usable data and therefore generated no signals.
The practical lesson is to check whether the live prediction window includes enough historical observations for every rolling calculation. A windowed factor needs prior data before it can produce values, so matching prediction dates to live parameters without adding a lookback can leave the input set empty. The exchange identifies this as the cause in the reported case, but provides no code, platform-wide diagnosis, or evidence that the same issue explains every backtest-to-simulation mismatch.
Key ideas
- A rolling average needs historical observations before it can produce factor values.
- A prediction window tied to live parameters may omit the needed lookback period.
- An empty simulation prediction set can result in no trading signals.
- The reported explanation addresses one case and does not establish a universal cause.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.