Other Half of the BlogPoll Ballot

that happened right? Is all the alcohol finally affecting my brain?

no hate buddy, everyone puts together their ballots differently

any system which, at this point, ranks Va Tech ahead of Alabama needs to be re-evaluated. I say this not only as an Alabama fan, but as a rational human being. This just does not pass the sniff test, and all the reassurances in the world cannot convince any person who saw the Bama-VT game that VT is the better team.

This is a distressing trend I am seeing from stats based football people. When a system spits out a ranking such as this, you simply MUST reconsider your rankings, or at the very least qualify your doubts. Otherwise, you come across as someone who rather arbitrarily made up some formula, then ran the numbers, and stuck to it without any correction or reflection. In the business world, when your formulas spit out a result which makes no sense, you rework your formulas or at least qualify the results. When you submit a serious poll which has VT showing as a better team than Alabama, you undermine your own credibility.

A couple food-for-though questions:

One: You mentioned in your SSOS writeup on that other page:

There is some noise, like when someone racks up huge numbers while blowing out cupcakes, but those tend to get evened out by the teams that don’t do much thanks to tough opponents. There will be some noise in any SOS rank as long as there’s not a balanced schedule across the board, which is to say there will always be noise.

One possible way to reduce outlier noise is to median filter. The simplest way for your purposes would be to simply remove the highest and lowest values from each category and average the rest. Looking at your example:

For “Tot Def”, this would mean eliminating the Hawai’i and Miami numbers. The average would then be 41.25, which really isn’t too different than before. However, if Hawai’i’s mark had been, say, 110, the unfiltered average would have been 49.17. Basically it’s what the BCS system does with the computer rankings when averaging them up. Normally, the difference is small, but when you have individual oddballs, it helps keep the numbers closer to a median value rather than a full average.

Two: Alabama’s SSOS feels a little odd to me. Other SOS metrics (e.g. Sagarin and Colley) give Alabama pretty high marks right now – well above NCAA average. I’m not saying yours is wrong simply because it’s not in line (I have issues with Colley right now, actually), but it might be worth looking into the numbers to see if there’s something unexpected about them.

Ok, three: I am curious if you remove opponents’ performances against the team in question when calculating SSOS. For example, when calculating SSOS for Florida, do the “Tot Def”, etc. numbers for Florida’s opponents include their games against Florida? If so, you’re getting some negative feedback: the better Florida does, the more it hurts their opponents’ SSOS marks.