The conditions are part of the result
In precision engineering, a measurement has meaning in the context of its test conditions. Computational research has an equivalent requirement: a result should travel with the information needed to understand and reconstruct the experiment.
A chart alone cannot provide that context. The input data may have changed. A dependency may behave differently. A default parameter may have been altered between two runs.
Give the experiment an identity
The SC reproducibility initiative asks authors to describe software environments and computational experiments so that others can attempt to repeat the work. It treats these details as part of scientific communication. [1]
Make comparison possible
For a market experiment, we favor a record that identifies the question, data version, code revision, configuration and execution command. It should also preserve the output and explain which assumptions were used. This makes differences between runs easier to investigate.
Reproducing a result is an engineering achievement. Establishing that it says something useful about a market requires further evidence. Keeping those questions separate helps prevent a clean execution from being mistaken for a validated strategy.