Home /permanent

Metric Hacking

Metric Hacking is when a metric is optimised in a way that improves the number without improving the thing it's meant to measure.

It's a common failure mode of Proxy Metrics: metrics can and will be gamed. It's closely related to Goodhart's Law: "When a measure becomes a target, it ceases to be a good measure".

In Kaggle competitions, it refers to tricks that optimise the competition metric in a way that doesn't align with the organiser's goals.

See Reliance on Metrics is a Fundamental Challenge for AI.