Here are my brackets for this year:
These are meant purely for fun. Also, they are meant for the type of fun which involves upset points.
Specifically, these sheets are designed for a pool where participants earn the difference in seeds as upset points. In other words, if you’ve accurately picked Eastern Kentucky (15 seed) to beat Kansas (2), you are rewarded with an additional 13 points, while correctly picking Kansas is worth no additional points.
Here are my last minute lessons from 2014:
1- Michigan St. is overvalued.
I really like the Spartans, but so does everyone else. So, while MSU might win, even if they do and you’ve picked them on your sheet, you’d still have a lot of work left to do to win your pool. If you are in a pool with 100 people and you accurately picked MSU as your champion, you still have to hope you’ve beat about 15 other people on all of the other games.
2- I wanted to pick Virginia as a title pick, but couldn’t pull the trigger.
Two things stood out.
First, in a post on hoopvision, Jordan goes through teams that, like Virginia, experienced a large increase in overall efficiency before and after January 1.
Here’s his table.
In nearly all cases, the team which “got hot” at the end of the season, relative to their early-season performance, struggled come March. Great stuff, Jordan.
On another note, Virginia lost by 35 (!!!) earlier this year against Tennessee. In one of my posts last year, results suggested that such one-game futility is not a trait shared by the recent national champions. As a result, I couldn’t have Virginia go all the way.
3) If you have a moment, this expose by nropp.com is quite thorough, and gives great insight into upset values for round 1. Manhattan (currently 14:1) and Delaware (unlisted odds) stand out as bold calls.
4) While it is great that the viewpoint of the analytics community is more prominent now than ever (Deadspin, Sports Illustrated, and ESPN all consulted statisticians in their picks, for example), I wonder if this over-saturation of analytics advice warrants a change in strategy.
For years – even through the Duke title in 2010 – much of the analytics folks were hidden on smaller websites, forums, and not given much of a voice with respect to predictions. As an example, Ken Pomeroy’s 2010 log-5 gave Duke a 25% chance of winning the national title, while fewer than 10% of the public gave Duke such a chance.
In 2014, this appears much different. My guess is that practically half of anyone’s pool will have at least consulted with one of the websites I listed above. This includes people who probably just copied the 538 bracket straight up.
And even if people managed to avoid the websites above, they are probably following the advice of one of another sites “bracket experts,” who, ironically, are now likely consulting Ken Pom, Massey, or Nate Silver’s predictions themselves before making their choices. Simply through transitivity, the analytics community has probably touched the vast majority of brackets in 2014.
As a result, we see two four-seeds among the most commonly picked champions. These two four-seeds are among Ken Pom’s top-10 teams and among Nate Silver’s top-4 teams. I wonder if the public would have similarly backed these four-seeds even just a few years back.
Moreover, I wonder if there’s a hidden value in other teams like Kansas, Syracuse, or Villanova, teams liked by the analytics folks (but not loved), but who are currently picked by paltry few people to do well, even as top seeds.