Guest Planker Will Blythe suggests creating a nationwide March Madness pool. Sounds like a good idea, but then comes the tricky question: what scoring system to use? This is a longstanding beef of mine, and the first night of Sweet 16 action seems like as good a time as any to air it. Did you have North Carolina in the Sweet 16? Yes? Well, you and 96 percent of the country can go ahead and pat yourselves on the back. How about Western Kentucky--did you have them, too? Really? And you live outside of the greater Bowling Green area? Now that's something to write home about.
But many NCAA tournament pools give you the same number of points for picking the Tar Heels and Hilltoppers. That's just a stupid way of conducting any sort of gambling operation, especially in sports. You should be rewarded for knowledge and intuition that goes beyond what a computer formula could tell you--that's why, for instance, people gambling on NFL games have to bet against the spread, so that you factor out of the equation the boring reality that some teams are a lot better than other teams. You should be far more handsomely rewarded for predicting Western Kentucky's success than for predicting North Carolina's--it takes a lot more basketball knowledge/good guessing/dumb luck.
Now, most pools don't have any form of this risk adjustment (often called "upset scoring"), and so the people who do well, by and large, are those who don't pick very many upsets. But, equally troublesome, the most common form of upset scoring makes things even worse. This pool, for instance, gives you bonus points for each upset--that is, for each time you pick a lower seed to beat a higher seed. Facebook used a similar system for a while, before turning its pool over to CBS this year. But this creates its own set of problems. For one, it values all upsets equally--you get the same bonus whether you pick a nine seed over an eight seed or a 15 seed over a two seed. And on the flip side of the coin, it values all non-upsets equally. So there's a huge and distorting incentive, in any game with teams seeded closely together, to pick the higher seed and get the upset bonus. If I were in a pool with this scoring system, I would pick all four nine seeds and all four ten seeds to win in the first round, regardless of what I thought about any individual game. I would never pick a four seed over a five seed to reach the Sweet 16. This, needless to say, is not a good scoring system.
The right way to do a pool is to award, for each game predicted correctly, a number of points that corresponds to the seed you predict correctly. You pick a one seed, you get one point; you pick a ten seed, you get ten points. This provides the right incentive to pick upsets you think might happen, but it also punishes you appropriately for picking a ton of upsets that don't pan out. Then, of course, as most pools do, you add a multiplier in each subsequent round, so that later rounds are worth more. My personal preference is 1/2/4/8/16/32 (so the same base number of points is available in each round), but some pools go with something more like 1/2/3/6/9/12 or 1/2/4/6/8/10. So, in my preferred system, for correctly picking Western Kentucky into the Sweet 16, you'd get 12 * 2 = 24 points, whereas for correctly picking UNC, you'd get 1 * 2 = 2 points. I've yet to find a pool that uses precisely this rubric, but if I ever end up running one, it's what I'll use.