CMP.11:8 - Common Anti-Patterns and How to Avoid Them

Tempting move	Failure	Useful repair
Generalize from a slow implementation	Another algorithm may avoid its work.	Prove a requirement for every admitted transcript or procedure.
Count distinct inputs without considering acceptable answers	Several inputs may legitimately share one output.	Count output-distinct classes.
Supply an inconsistent adversary	No actual input could produce its replies.	Retain a nonempty compatible input family.
Fix randomness, then choose a different hard input for each seed	This can miss a successful randomized algorithm.	Use one prior input distribution or a valid coupled argument.
Ignore input-dependent preprocessing or stronger access	The algorithm already acquired the supposedly missing information.	Charge or expose that resource and restate the model.