Quantitative Research · KAEL Editorial Department
In addition to backtest results, what else needs to be checked?
Historical simulations can help test research ideas, but they rely on data and execution assumptions. When reading the results, you need to also understand under what conditions the model was run.
Check chronological order
Make sure each decision uses only the information already available at the time. The timing of data release, subsequent revisions and sample screening may introduce future information that is not easily detectable.
Restore execution constraints
Deal prices, fees, slippage, and capacity assumptions all affect simulation results. Write these parameters into the experimental configuration and check whether the results are sensitive to parameter changes.
Observe different intervals
Break the overall results into multiple time periods and environments to find the conditions under which performance is concentrated. Aggregate indicators may mask longer failure stages and cannot alone account for the stability of the method.
