## Nerdy March Madness contest

### 16-Mar-10

So those of you who know me well know that I’m a sucker for “nerdy” sports pools and contests. So pro-football-reference has once again come up with a unique type of contest – it’s a twist on the regular March Madness bracket filling out.

All “64″ teams are listed, and each team is assigned a “cost”. You can pick as many teams as you want, as long as the total price stays at 100 or less. The winner is the entry with the most total wins by all teams in the entry.

So if you want to join the contest, go do that first before reading some of my strategy.

Spoiler space….

space….

yup….

Okay.

So I came at this from a few different angles. I looked at the odds to win as published by the gambling community, to see people that had good odds at a relatively low cost. I also got folks Sagarin ratings, with the idea of looking at teams that were either under-seeded or under-costed. I definitely wanted to stay away from me picking who I thought would win, as I have a historical record as “world’s worst bracket filler-outer”

I then thought to see if I could do any sort of arbitraging. I added up the costs for each “pod” of 4 teams (1/8/9, 2/7/10, 3/6/11, 4/5/12/13), assuming that the 14-16 seeds would all lose. The cheapest ones cost 17, so if you got 6 of those, that would be 3 guaranteed wins each, or 18 total, plus any Sweet 16 and later wins for any of those folks (#3 and #4 seeds likely). But I wasn’t sure if 20 or so wins would be enough to win, though I thought probably not.

And plus, it seems a bit wasteful to spend money on teams that you KNOW have to lose. So, I moved on to figuring out expected wins by round by seed. I kind of made up probabilities for each seed to win each game. So for example, I have the #3 seed as a 90% chance to win Round 1, a 65% chance to win round 2 (assuming they make it there), etc. etc. That gives expected wins by seed up to the final 4. So of those 4 games, #1 seeds came up with 2.5 expected wins, which sounds about right. So then I used that number to try and come up with a max team. That is discounting the last 3 games, but hopefully will at least be close.

So the top pick there was Wake Forest, who as a 9 seed has 0.53 expected wins (.48 for beating the #8 seed in Round 1, and then .048 for beating the #1 seed in Round 2 (I guesstimated a 10% chance of the 8-9 winner beating the #31 seed) and then .006 chance of winning in the Elite 8). But they only have a cost of 2, so they have 0.265 wins per cost.

The top 17 teams in wins per cost (Wake Forest, Florida, Ohio State, New Mexico, Pittsburgh, Richmond, Vanderbilt, Gonzaga, UNLV, Cornell, Notre Dame, Kansas State, Villanova, Butler, Kentucky, Louisville, Northern Iowa) have an exact cost of 100, and come out at 19.72 expected wins, with 3 #2 seeds (all but WVU) and Kentucky in my picks, so a decent chance for 1-2 more wins in the Final 4.

But that is assuming that all #1 seeds are identical, as well as better than all #2 seeds, who are in turn better than all #3 seeds, which we all know is not true. So incorporating the team’s actual strengths will also tell me who might be under or over seeded. After searching for a few measures, I found that the folks over at basketballprospectus.com had already figured out odds for each team to win each game. So let’s use their numbers!

I did have to convert their numbers, which give the chance that a given team will reach a given round, into the percentages that a given team will win a particular number of games, but that wasn’t too bad with my trusty friend, Mr. Excel! The top 21 teams from that cost 100 and give me 24.93 expected wins, led by 11 seed Washington, who is given a 23.4% chance to win its first game at a cost of 1.

But that does at least peg a decent strategy as somewhere around 25 wins as a baseline. So I should be looking for 1 win per 4 points of cost at a minimum.

So I like a few low seed / low cost first round upset fodder, listed with seed, Pomeroy chances to win 1st round game, and cost

#14 Montana, 23.4%, 1

#13 Murray St. 40.9%, 2

#12 Utah St. 51.8%, 3

#13 Siena, 25.6%, 2 – they are probably under-represented here as they are playing a Purdue team that is now w/o one of their best players.

I think it’s good (but not really sure) that Utah St. and Siena would play each other in the 2nd round if they both won their first games.

So, if we are shooting for around 25, then all of a sudden the 17 cost 4-team “pods” start to look attractive.

Pitt/Xavier/Minnesota (3/6/11) cost 17, and have 3.47 expected wins (3 guaranteed wins). Wisconsin/Temple (4/5) have 3.59 expected wins for a 15 cost or 3.9 wins for 17 if you throw in #12 Cornell

In the end, here were my picks

BYU

Clemson

Duke

Florida St.

Georgia Tech

Maryland

Minnesota

Montana

Murray St.

Ohio St.

Pitt

San Diego St.

Siena

Temple

Utah St.

Wake Forest

Washington

Wisconsin

Xavier

That’s 100 cost, and 23.94 expected wins. You’ll see that I couldn’t help myself from subjectively screwing myself out of 1 win