Thursday, March 19, 2009

March Madness Picks: UNC Takes it if Gonzaga Doesn't Upset Them

Update 6 April: I correctly called the final winner (UNC), along with most of the rest of the country. But my overall success rate of predicting winners was 60.3%. Maybe this is my last year doing this.

I know nothing about basketball. But like an unemotional, disciplined investor, I developed a quantitative approach and I stick with it. Last year I got into my office pool at the Sweet Sixteen stage and picked 14 out of 15, including the winner, despite the disbelieving howls at some of my picks when I made them. The method is simple - best scoring margin wins, period.

This year I was inspired by to try to figure out the relative skill of teams; that is, if you have a good scoring margin but you're in a crappy conference, when you line up against somebody from the Big 12 or Big East, you might be in trouble. Masseyratings does some giant networking thing (A beat B, who beat C but was beaten by D, who beat A...etc.) Whereas I have an Excel sheet and a day job. But to take skill level into account, I looked at the tournament last year and tried to assess skill level by looking at how many teams each conference sent to the tournament. If you're in the Pac-10 or the Big East (which had 6 and 8 teams in the 2008 tournament respectively) chances are you're a seasoned team, because you have to play all the other good teams from your conference. On the other hand, if you're the only representative of East Bumblef**k Valley Conference at the tournament, maybe not.

Well it didn't work. To scoring margin, I added the difference between each pair of team's representation at the tournament, multiplied by an adjustable fudge factor. But I could never make it come out more accurately in terms of number of correct win/loss predictions than plain old scoring margin, which in any event is the best correlation I could find between tournament progress and any commonly tracked team statistic (R=0.694). Number of entrants in tournament doesn't give as much information as stats for each of the conferences, but if you're interested in that, you can play with spreadsheets because I'm officially done.

Here are the picks. I only pick win-lose, not spreads. For each game I only care about the winner - so say I predicted that in Round 1, A beats B and C beats D; in Round 2 I predict A beats C. Say that what really happens is that in Round 1 A beats B as I predicted but D beats C. As long as A beats D and wins the Round 2 game, that counts as a correct prediction. Consequently if I predicted a team would win a game in Round 2, but they got knocked out already in Round 1, then I know I'm wrong before Round 2 even begins. They can't win a game they're not in.

Bottom line, I agree with Obama that UNC will take it, although Gonzaga has a real chance to squeak by them in the third round.

FIRST ROUND PICKS - 22 W/L out of 32. 68.75% correct. Terrible. Lower than 75% I got last year at this stage.

I was right:
Connecticut over Chattanooga
Purdue over Northern Iowa
Washington over Mississippi St
Memphis over Cal State Northridge
Villanova over American
Texas over Minnesota
North Carolina over Radford
LSU over Butler
Gonzaga over Akron
Oklahoma over Morgan State
Duke over Binghamton
Oklahoma St over Tennessee
Louisville over Morehead State
Pittsburgh over East Tennessee St
Michigan State over Robert Morris
Missouri over Cornell
Arizona State over Temple
Xavier over Portland State
USC over Boston College
Wisconsin over Florida State
Siena over Ohio State

I was wrong:
BYU over Texas A&M
Cal over Maryland
Clemson over Michigan
Illinois over Western Kentucky
Stephen F. Austin over Syracuse
North Dakota State over Kansas
Utah St over Marquette
Wake Forest over Cleveland State
West Virginia over Dayton
Utah over Arizona

SECOND ROUND PICKS - 75% correct.

I was right:
Louisville over Siena
Mich State over USC
Connecticut over BYU (really Texas A&M)
Purdue over Washington
Missouri over Utah St (Really Marquette)
Memphis over Cal (really Maryland)
Pittsburgh over Oklahoma St
Xavier over Wisconsin
Duke over Texas
North Carolina over Butler (really LSU)
Gonzaga over Illinois (really Western Kentucky)
Oklahoma over Clemson (really Michigan)

I was wrong:
Stephen F. Austin over Arizona State
North Dakota State over West Virginia
Wake Forest over Utah
UCLA over Villanova (a little paranoid about this one) (turns out I should have been more than a little paranoid)

THIRD ROUND PICKS - 62.5% correct.

I was right:
Pittsburgh over Xavier
Connecticut over Purdue
Louisville over Wake Forest (really Arizona)
North Carolina over Gonzaga (thought it would be very close)
Oklahoma over Stephen F. Austin (really Syracuse)

I was wrong:
Memphis over Missouri
UCLA over Duke (very close) (was really Villanova over Duke)
North Dakota State over Michigan State (iffy about this one but I'll stay with it; turned out to be Kansas vs. Michigan State, Michigan State won)

FOURTH ROUND PICKS - 25% correct.

I was right:
North Carolina over Oklahoma

I was wrong:
Louisville over North Dakota State (turned out to be Michigan State over Louisville)
Pittsburgh over UCLA (will be Villanova, and they won)
Memphis over Connecticut (actual game was Connecticut vs. Missouri, Connecticut won)

FIFTH ROUND - 50% correct.

I was right:
North Carolina over Pittsburgh (will be Villanova)

I was wrong:
Memphis over Louisville

CHAMPIONSHIP: - 100% correct.
North Carolina over Memphis (will be Michigan State)

No comments: