In my opinion, the more upsets there are in March Madness, the better. A good way to measure how many upsets there are is by the average seed of each round. I’ve calculated the average seed per round for each tournament going back to 1985 to give historical context to each year’s tournament. These are the past four years:
If you average the average seed from the round of 64 down to the round of 1, you get the All Rounds Average Seed which represents how many upsets there were in the tournament. It’s a good way to represent the tournament in one number. Now, all the years can be compared with this metric. Here, you can see 2014 had a lot more “Madness” than other years.
This is how 2017 compares to the extremes in each round:
This is how 2017 ranks with the other 33 years. A lower rank means a lower average seed for that round.
This year’s tournament was the 9th least underdog tournament of the 33 years. The round of 32 was particularly favorite heavy, while the Elite Eight and Final Four weren’t bad with 7 seed South Carolina and 11 seed Xavier. 2014 had the highest overall average seed of 6.1 due to #8 Kentucky and #7 Connecticut in the championship game.
This year’s overall average seed was put on a scale from 0 to 10 with the theoretical max of 14.1 on one end and the theoretical min of 2.9 on the other end. It was a 0.8/10. 2014 was a 2.9/10 which puts things in perspective. Even the most upset year was only 29% of the way to a completely upset tournament.
If the overall average seed was scaled with the historical max and min as the endpoints instead of the theoretical max and min, this year’s tournament would have a 2.4/10 “Upset Factor”. 2014 is a 10/10 and 2007 is a 0/10.
One notable thing this year was that the average seed dropped from 4.1 in the Sweet Sixteen to 3.8 in the Elite Eight. That was the second smallest drop ever. The average drop in seed in that round is 1.2: from 4.4 to 3.2. Again, that was due to #7 South Carolina and #11 Xavier winning.
The average of all the years follows the lower seed winning result closer than the higher seed winning result.
Now we’ll look at historical trends in the average seed.
There’s been more variation in the past 5 years.
Notice something? Each round, the average seed varies more. In fact, the standard deviation increases by the same amount, 0.27, each round.
This is a 2D graph of the average seed per round for each year. Each line is one year, and more recent years have a higher opacity. I made this in MATLAB because it provides greater control over the opacity. Here’s a nice look at the detail from round 2 to round 3.
This is Excel’s 3D plot of the average seed per round for each year. It doesn’t offer any other angle and it’s ugly, so I imported the data into MATLAB and plotted it.
The 3D line plot looks better, but it’s still messy. I tried a variant on the line graph called the ribbon.
I also made a bar plot which shows the same thing.
There’s too much data to be displayed clearly, but these graphs give us an interesting visual summary of the history of March Madness by using average seed.