I would argue that point-based estimation is just as inaccurate as time-based estimation in that it only serves as a proxy for effort, which is done over time.

I would argue that point-based estimation is just as inaccurate as time-based estimation in that it only serves as a proxy for effort, which is done over time. Point systems (and time systems) have two big issues * You're throwing away all your data points when you condense a story down to a single point. Systems exist to try and avoid things like anchoring bias, but you just throw it away. * You imply uncertainty exists about the task duration, but don't try and measure how uncertain you are. Using Monte Carlo simulation here can really help address these two issues.