a few years ago, the new york giants defeated the new england patriots in the super bowl in what was a tremendous upset. it was the only game that the patriots lost all year.

i don't want to say that i predicted it, although some people gave me credit for doing so. i didn't and don't claim any special football knowledge. i just happened to have a math degree and was able to realize that the intuition people have about streaks is usually false. it can be true under certain factors that are not truly probabilistic, but when you do the math in a situation where uncertainty is involved it generally works out the exact opposite way.

the simple way to state it is that the longer the streak is, the more likely it is to break. so, a team that walks into a championship game after winning the last eighteen games (or whatever it was) is actually at a hefty disadvantage. they are bound to lose eventually - by skill or by luck.

i think people get confused because they want to apply this idea of inertia. the intuition is that if something has happened repeatedly, then it is more likely to continue happening than stop happening. but, that is just wrong.

i'm looking at the sunspot cycle. the best scientists don't really have much of a grasp regarding how this really works. there's a few guesses. the overriding mentality seems to be that, because the cycles have been decreasing over the last several cycles, it follows that they should continue to decrease. this is the same probabilistic fallacy that declared the patriots' superbowl win to be an inevitability, and the game a formality.

i'm going to take a dissenting view, and cite the exact same logic, but do it properly. if cycle 25 is a decrease, it will be the 4th decrease in a row. i do not believe we have ever seen four decreases in a row - meaning that that would be very unlikely. i consequently suggest that we're bound to see an increase....

it might not be a big increase, mind you. i'm just pointing out that a streak that long would be unlikely, given it's never happened before.