[!todo] Seed note. A starting point, not a finished note yet.
Monte Carlo forecasting predicts when a backlog will finish without estimating any single item. It takes a history of weekly throughput (items completed per week), then simulates the future by repeated resampling: draw a random past week’s throughput, subtract it from the remaining item count, draw again, and count the weeks until the backlog empties. One pass is a trial; running a few thousand trials yields a full distribution of completion dates. You read commitments off the percentiles, typically the 85th, the date you beat 85% of the time. The method works because the natural variation in a team’s own delivery history substitutes for the size of the work, which is exactly the quantity story-point estimation tries and fails to guess. It assumes the future resembles the recent past, so the forecast must be re-run as team size or work mix changes. Seeded from Software Estimation and Forecasting.