Knowledge MaxxingControl variable
A control variable is anything kept identical between two experiment runs, such as the data split or random seed, so any performance difference can be attributed to the one thing actually being tested.
Prerequisite: Experiment hypothesis
What matters
Without controlling other variables, a performance difference between two runs could be caused by an unrelated change rather than the thing you intended to test.
Example
When testing a new feature, keeping the exact same data split, random seed, and hyperparameters as the baseline run isolates the feature as the only real difference.
Common misconception
A control variable is not something that must be excluded from the experiment; it is something deliberately held constant so it cannot explain any observed difference.
Recall question
Why must other variables be held constant when testing one specific change?