Helmuth van Moltke (1880): "No plan of operations reaches with any certainty beyond the first encounter with the enemy's main force." It applies to software projects too, especially large ones. The enemy is the real world (or our lack of understanding of the real world.)

If we look at the original requirements every single project is a failure. If we look at the strategic goals and keep track of all trade offs maybe 30% is a fair guesstimate.