They really would. I feel like kaggle is mostly teams slowly grinding for very small incremental improvements over each other. Adding a time component would really sort the wheat from the chaff.

It would certainly change the dynamic. Not sure about wheat from the chaff - faster != better.

Metrics for scoring: error/cost/fitness, time to solution (t), dollars to solution ($), resources in terms of {carbon, kWh, USD,}, costs in terms of costed opcodes,

Whether there's a logically plausible causal explanation (for quickly-discovered probably just associative relations)

The monthly ones are fun.