One thing that we have worked hard to understand will sound a lot like a question you ask your investment advisor…

How much risk should I have in my portfolio (of bracket picks)?

We have examined this question thoroughly.

If you took all the better seeds all the way through your bracket, you would likely finish between the 65^th and 95^th percentile, depending on the year and who you picked among the 1 seeds to win once reaching the Final 4.  Great you are better than most entries, but you still aren’t winning.

Let’s take a closer look at the last three years to get an idea on how much risk you may need.  We used a scoring system where the 1^st round is worth 1 point, the 2^nd round is worth 2 points, etc.

In the chaos that was the 2010 NCAA Tournament, you would have had to take 19-30 points of risk, and be right.  In 2009, you would have need 10-21 points of risk.  2008 had very few upsets and you would have needed 6-17 points of risk.

We repeated this drill for every year and we looked at the ESPN scoring model which makes each round worth the same number of points, but allocated among fewer games.

Here is the rule of thumb that we believe transcends point allocation dependencies.

We think it is reasonable to target 4 lower seeds winning in round 1 (not including 7-10, 8-9 matchups),  2-4 lower seeds in round 2, and 1-2 games in rounds 3 and 4.

If you can get more than 50% of these correct, you should be able to get 13-17 points ahead of the competition.  Then you have your Final 4 games to close any remaining gap.

In a soon to be released article, we will look at how you should dissect this analysis in two categories:  When do Brackets mis-behave because of poor seeding and when do teams mis-behave.

Here is the Final Four Shortlist:

A New Look at our Final 4 List and a little about our methodology

We look at our Final 4 possibilities through five filters and after we have those teams selected, we examine possible combinations based on historical distributions.  The second part can’t be done until we have an actual bracket, but we can look at teams through our five filters.

We have seven teams that are getting through all the filters right now:  Ohio State, Kansas, Texas, Villanova, Maryland, Georgetown, and West Virginia.  It is likely that 3 of the Final 4 teams will come from this list.

We have eight teams that get through our statistical filters, but get caught in one of our historical or situational condition filters:  Pittsburgh, BYU, North Carolina, UConn, Duke, Kentucky, Wisconsin, and Syracuse.  It is likely that one of these teams will slip through and make the Final 4.

Our filters generate a lot of questions.  Without giving away our “secret sauce”, we will tell you that we look at conference affiliation, point distribution, and historical performance.  Our statistical filters look at historical distributions of on the court performance and computer rankings and we balance accuracy (how many teams can we get right) and precision (how many teams can we get right without including 200 teams).

We will tell you a little about things we look at after the bracket is released.  One of the things we look at is do we have a team that was in last year’s Final 4?  This happens 73% of the time.  We also look to see if we have two teams from one conference.  This happens 69% of the time.  However, both of these things happening only happens 38% of the time.  We also look at seed distribution and computer ranking distribution to see if it falls within historical precedent.

For these variables that we have found, there are probably five times as many that we have tested and have failed our statistical tests.  We are currently looking at some new ones and will continue to do so right up to Selection Sunday.  If we can’t get them coded in time, we will comment about it in the blog.

Many of the things that don’t pass statistical testing we will discuss in our blog, because while it may not pass statistical tests, they are still statistical trends that we want you to be aware of.  For example, playing in your home state or playing in a dome for the first time are yet to pass statistical tests, but they could pass once we add 2011 results.

One thing we aren’t afraid to do is “test and learn”.  In other words, provide a mea culpa and from what we learn from one year, incorporate it into next year’s model.