I promised in my last post that I had an idea spurred by an astonishingly bad record amongst the "expert" picks when it comes to predicting game outcomes. About forty years ago one intrepid football fan took it upon himself to track expert picks and success rate in college and NFL football games. His conclusions were that the experts had a success rate of.476. Not bad for a baseball player, but not exactly Pro Bowl material for sports predicting. Think about it, a less than .500 record means they could have just flipped a coin to determine the outcomes and come out better. So, where does this lead me? I have a couple of experiments I want to try.

I came up with this plan last night, near the end of spending thirty five out of forty eight hours in a radio station during a snow snow with about three hours of sleep, and today I have already discovered some crucial flaws, but we're going to try it anyway and keep it as an on going project.

Here's the plan as I wrote it last night:

Since we are nearing NCAA tourny time, my plan is to try something a little different with my bracket picks this year. I'm not going to pick them. Instead, I'm going to let George Washington pick them. (Embarassing realization: I just checked to make sure it was indeed George Washington on the quarter.) (Before you laugh at me, how many of you thought I was referring to the dollar bill when I said George Washington?) Here's my plan. This year I'm going to to flip a coin to determine my choices. Heads the higher seed wins, Tales the lower seed. Now admittedly I haven't thought this completely through, so I will probably tweak my methodology slightly before the tourny starts. I'm going to spend the next month working out exactly the best way to do this. I think I will have two controls. One will be my own bracket with my own actual picks, the other will be the "expert pick. I will determine the expert at a later date, but rest assured, I'm not going to cheat. I'll pick somebody who "really knows" what he's doing. Next I will have a bracket determined with a simple coin toss. Heads the higher seed wins, tales the lower seed. My final bracket I will pick the first round and let a coin determine every ensuing round. In the time running up to the tourny, I'm going to be going back to past brackets at trying the quarter toss (or how 'bout the GW toss?) and see how it stacks up to the actual results. This will serve a couple of purposes: one, I think it will give me something to do while I'm not sleeping at 4:00 in the morning, two it will give us some preleminary data, and three, it will tell me ahead of time which bracket I should enter in the work pool this year. As soon as I figure out how to insert pictures into this blog I will post my first set of findings.

That was the plan as written at 4:00 in the morning after being up for nearly two consecutive days. Anyone spot the flaw?

If you said George Washington doesn't know anything about football, then you're right. If you said that stat I quoted up there (.476) pertained to football predictions, you are right again. I went back and did my first test run with the 2009 NCAA Tournament, and apparently George Washington is an Alabama State fan. GW picked them upsetting first seed Louisville, going on to defeat Ohio State, Arizona, and finally losing to West Virgina in the fourth round. Ok, flaw.

My second bracket experiment faired a little better. For the first round I gave the higher seeds a pass and then proceeded with the coin flip all the way to the end. This prevented any lunacy like Alabama State going to the elite eight, but still gave me some rather screwy results. Generally I'm a big fan of screwy results, but in this case, we're looking for the perfect bracket (or at least a better than .500 bracket), not to screw with the data. Clearly I needed a new tactic.

Once again, while at work, I revised my methodology and ran a few more brackets to see how it turned out. At this point I'm up to six brackets, six different methods of picking them. Each bracket seems to become a little more viable than the last, so I think I'm working towards something very cool. Coming up next I'll walk you through my method and show you the results of my insanity.

-C