I can make the end result whatever I want with programming the algorithms.
Or even the parameters to be used with the algorithms.
Many of the simulations for the Gulf War were predicting a conventional assault up through the Saudi Border, like what would have been seen in WWII (or an amphibious assault at Kuwait City). None of them involved a multi-prong assault not only there, but a dozen more places with the intent of sweeping behind the Iraqi forces and cutting them off from all support and supply. And in addition doing a heliborne assault into Kuwait City and taking the airport.
The problem was that they were basically trying to game a conflict in 1991, with by then 50 year old tactics and data. All of the "predictions" were based more or less on what the results would have been say if it was Patton or Ike in Europe. Or a slow slog through adverse terrain as in Korea or Vietnam. Nor the more mobile and spread out forces that the US had spent the last 30 years developing for areas like the deserts of the Middle East.
The algorithms could have been dead on, but in a case of GIGO the dataset they used was completely wrong.
Which the Battle of 73 Easting proved. Where when looking at the figures, the Iraqi Army should have won easily. In manpower both sides were equal, but the Iraqis had a third more armored vehicles, a huge amount of artillery already ranged in, and defensive positions. The Coalition were advancing on prepared positions, in bad light (a moderate sand storm), into unfamiliar territory. Yet, at the end of the battle around 1,000 Iraqis were dead, and over 1,300 were taken prisoner. Over half of their armored force (160 tanks out of 200) was destroyed (many of them their elite Republican Guards), and most of their armored fighting vehicles and artillery were destroyed.
The Americans lost 6 men, and a single M2 Bradley fighting vehicle.
And I can promise, if any of the computers had been given those figures they would have predicted a devastating American loss.
