I love to look at predictions, especially after the fact. Harvard scientists have predicted the Salt Lake City Olympics medal count by an economic model explicitely not taking into account any sports knowledge but looking at economic factors and length of winter etc. Here's what they had predicted:
click and scroll a bit

Now that we know the result we can check how good their model is. One easy way is to compare the model's results with a primitive 'model' without any math in it. The first model that came to my mind is 'everything stays the same'. That means I 'predict' that each country will repeat last Olympics' result.

One obvious bit of sports knowledge should be taken into account: The host country will do much better than last or next time (a cursory glance at each medal count across Olympic games will show that to be true), due to a mixture of better motivation of athletes and pressure on judges by audience acclamation in many events.

So if we simply disregard the USA results instead of making an ad hoc allowance for hosting advantage we can compare the economists' model and the 'everything stays the same model' with the actual results (you'll find them at many places). We now see that last Olympics' results were at the very least not a worse predictor of the results in Salt Lake City than a complicated model.

As long as a primitive model is as good a predictor as a model with a lot of math I prefer the primitive model.

That reminds me when in the Netherlands a panel of experts predicted the football results for one season and their performance was compared to a 'model' that did nothing else but predict each single time a win of the home team the 'model' did better than any of the experts.

Wolfgang