A peak is not an average.
Record what the number represents.
Keep the measurement window, direction of movement, reset handling, policy and environment beside the result. The command sent to a policy is also different from the speed observed.
The evaluation version matters.
The evaluation script at the reviewed submodule commit reports displacement over a time window. It does not implement the upright-gated instantaneous peak described in the project log.
Inspect the reviewed evaluation code ↗Reproducing the reported peak requires its matching evaluation version and checkpoint. These values are not verified hardware performance.
What would you like to understand next?
Keep the source, your environment and your observations together.