Ask HN: Does 'Evidence Based Scheduling' work in practice?

I've just re-read an old article by Joel Spolsky about using a monte-carlo simulation to create accurate software release schedules [1]. Does anyone have any experience using this approach? If so, would you recommend it? [1] http://www.joelonsoftware.com/items/2007/10/26.html