I've been working in the probabilistic estimating business for a decade or two.
One of the seminal books that started it all is Short-Term Forecasting. This is the basis of Box-Jenkins and Mr. Jenkins quote that is universally misquoted by most people in the Agile community.
All Models are Wrong, Some Models are Useful
As well - of course - is the nonsense that Forecasts are not estimates, popularized by #NoEstimates advocates.
The Box-Jenkins modeling process expands to the ARIMA (Auto-Regression Integrated Moving Average) models we use for cost and schedule models. This approach makes use of past performance to forecast future outcomes. This empirical method is used nearly universally for forecast time series in the past for outcomes in the future. From stock markets to Estimates to Complete and Estimates at Completion. In our domain, we archive all the performance numbers and then compare them against the planned performance models. This is then used to make adjustments to the Plan or the Estimate from the Root Cause of the variance.
This is called Close Loop Control, and that works on all domains where stochastic process underly the work process, from software development, to process control, to flight control systems, the natural systems.
