What is backtesting?
Backtesting applies a rule to historical data to study how it would have behaved under stated assumptions. It is a research tool, not evidence of future performance.
Concrete example
Define a simple rule, run it only on information available at each historical point, include a stated cost assumption, and compare outcomes with a baseline.
Vocabulary
Look-ahead bias uses unavailable future information. Overfitting tunes a rule too closely to one sample. A baseline is a simple comparison method.
Prerequisites
Students need basic Python, statistics, and data-cleaning skills before building a backtest.
Common mistakes
Do not choose a rule after seeing the outcome, ignore costs, or report only the favorable time period.
Student exercise
Split a small historical dataset into a design period and a later evaluation period; record why results can still fail to generalize.
Competition relevance
Clear backtesting limits matter as much as a chart when explaining an IQIC quantitative method.
Prerequisite math and programming
Use algebra, probability, descriptive statistics, and basic Python before adding model complexity. Learn one concept at a time and retain every assumption.
Common mistakes
Do not confuse a historical result with a forecast, hide data transformations, or omit costs, limitations, and a baseline comparison.
Simple student exercise
Write a small reproducible analysis with a documented public input, one calculation, one labeled chart, and three limits on what the result can establish.
Competition relevance
IQIC reviewers need to understand a team’s method, evidence, assumptions, and limitations. Start with the competition overview.