Hokepoints: Bracketology Because It's Bracketology Week 2013
Site Notice: This Thursday we're planning a basketballgasm liveblog, culminating in the Michigan-South Dakota State game. Probably getting started with the afternoon games, so you've got from now until then to get your brackets filled out and get your work done before productivity goes to Bolivia. Viva March!
My new tradition. I'm not really the basketball guy around here, however I do seem to perform really well when it comes to March Madness brackets, getting back more than my pay-in every year since 2000 (won twice). The first six years of that was luck—since then I've just been working really hard at it so I don't lose the streak.
For those filling out their brackets today here's some helpful stuff. My favorite tool for clearing the biases is the Wall Street Journal's blind comparison. Also never miss the annual GARGANTUBRACKET by Czabe.com, the blog Bracket Science and the gloriously cheap calculator at Poologic, which lets you program how many upsets you want and find inefficiencies to exploit. Use SCIENCE! to take money from your friends and co-workers!
The last tool is my own (<<<<<grab it here>>>>>). It turns KenPom's ratings into a confidence %, and then automatically pulls up which venue the game will be at and whether there's any injuries you need to know about for either team. Who likes drop-down menus?
What I do is normalize the closest 16-1 matchup (Kansas vs. WKU at 22.6% difference in KenPom's "Pyth") as 100% for the 1 seed to win, and use the KenPom ratings to percentile everyone else's games into a confidence number. Then I roll through anything under 70% and decide if my knowledge of those teams might justify taking the under.
Here's the first round, where "Confidence" is a measure of how likely the top seed might be to win. The venue is listed so you can identify things like don't take Boise over K-State in KC, or how 12-seed Cal (a team worse than Virginia, Iowa, Denver, Baylor, Kentucky, Stanford, UConn, Maryland, and Sothern Miss according to Kenpom) is basically playing at home in San Jose.
[UPDATE: I had some errors in the below chart. Now fixed. The tool was fine but I've added an option to set your own chaos factor.]
| High Seed | Low Seed | Difference | Confidence | Venue |
|---|---|---|---|---|
| Indianapolis Regional | ||||
| 1 Louisville | 16 North Carolina A&T | +62.6% | 100.0% | Lexington, Ky. |
| --or-- | 16 Liberty | +68.8% | 100.0% | Lexington, Ky. |
| 8 Colorado St. | 9 Missouri | -1.2% | 48.8% | Lexington, Ky. |
| 5 Oklahoma St. | 12 Oregon | +5.4% | 55.6% | San Jose, Calif. |
| 4 St. Louis | 13 New Mexico St. | +17.9% | 68.6% | San Jose, Calif. |
| 6 Memphis | 11 St. Mary's | -3.2% | 46.7% | Auburn Hills, Mich. |
| --or-- | 11 MTSU | -1.2% | 48.8% | Auburn Hills, Mich. |
| 3 Michigan St. | 14 Valparaiso | +15.6% | 66.3% | Auburn Hills, Mich. |
| 7 Creighton | 10 Cincinnati | -6.0% | 56.2% | Philadelphia, Pa. |
| 2 Duke | 15 Albany | +40.2% | 91.8% | Philadelphia, Pa. |
| Los Angeles Regional | ||||
| 1 Gonzaga | 16 Southern | +48.1% | 100.0% | Salt Lake City, Utah |
| 8 Pittsburgh | 9 Wichita St. | +8.1% | 58.4% | Salt Lake City, Utah |
| 5 Wisconsin | 12 Ole Miss | +7.7% | 58.0% | Kansas City, Mo. |
| 4 Kansas St. | 13 La Salle | +3.9% | 54.0% | Kansas City, Mo. |
| --or-- | 13 Boise St. | +6.1% | 56.3% | Kansas City, Mo. |
| 6 Arizona | 11 Belmont | +6.2% | 56.4% | Salt Lake City, Utah |
| 3 New Mexico | 14 Harvard | +25.3% | 76.3% | Salt Lake City, Utah |
| 7 Notre Dame | 10 Iowa St. | +1.1% | 51.1% | Dayton, Ohio |
| 2 Ohio St. | 15 Iona | +27.5% | 78.6% | Dayton, Ohio |
| Arlington Regional | ||||
| 1 Kansas | 16 Western Kentucky | +48.6% | 100.00% | Kansas City, Mo. |
| 8 North Carolina | 9 Villanova | +4.0% | 54.1% | Kansas City, Mo. |
| 5 VCU | 12 Akron | +8.1% | 58.4% | Auburn Hills, Mich. |
| 4 Michigan | 13 South Dakota St. | +26.1% | 77.1% | Auburn Hills, Mich. |
| 6 UCLA | 11 Minnesota | -5.0% | 44.8% | Austin, Texas |
| 3 Florida | 14 Northwestern St. | +40.0% | 91.6% | Austin, Texas |
| 7 San Diego St. | 10 Oklahoma | +5.7% | 55.9% | Philadelphia, Pa. |
| 2 Georgetown | 15 Florida Gulf Coast | +33.3% | 84.6% | Philadelphia, Pa. |
| Washington D.C. Regional | ||||
| 1 Indiana | 16 Long Island | +50.5% | 100.00% | Dayton, Ohio |
| --or-- | 16 James Madison | +51.3% | 100.00% | Dayton, Ohio |
| 8 NC State | 9 Temple | +9.0% | 59.3% | Dayton, Ohio |
| 5 UNLV | 12 California | 6.6% | 56.9% | San Jose, Calif. |
| 4 Syracuse | 13 Montana | +37.4% | 88.9% | San Jose, Calif. |
| 6 Butler | 11 Bucknell | +2.7% | 52.8% | Lexington, Ky. |
| 3 Marquette | 14 Davidson | +8.8% | 59.1% | Lexington, Ky. |
| 7 Illinois | 10 Colorado | +1.3% | 51.4% | Austin, Texas |
| 2 Miami FL | 15 Pacific | +30.6% | 81.8% | Austin, Texas |
General tips:
If you're in a big pool, run multiple brackets, each with carefully selected upsets. There's no such thing as an NCAA tournament without lots of big upsets and at least one surprising run. The 1 seeds all made it to the Final Four just once. If you submit one milksop bracket you're up against every other milksop bracket and will get beat by the one crazy guy who had LSU going to the Elite 8 or something. Hitting on a carefully selected upset that rearranges a bracket and lets you ride a different high seed to the Final Four is the most typical route to a win.
If you're in a small pool, play conservative. One or two points won't usually make a difference in a small pool, but the likelihood of something crazy like that one guy's wife who picks based on the cuteness factor of mascots winning is cut down so you don't need to take risks to get ahead.
Pick the upsets the most carefully. I love picking 6-11 upsets because if you get it wrong they're bound to get wiped out by the 3 anyway. If you roll the dice on a 3-seed or lower losing early though, you'll feel like an idiot as the rest of your pool collects the easy points. A tournament without upsets never happens, but neither does a tournament with all the upsets. You can totally undo a great pick with a terrible one elsewhere.
Get value for your upsets. Know who's in your pool and the inefficiencies. Fans will generally take their favorite team to go two rounds later than they really belong and conference teams to go a round further. This is an inefficiency.
Be really really lucky. This is really the only rule.
I meant Pitt.
I don't dare test my linking skills, but Nate Silver may be of some help. He gives Michigan a 2.4% chance of winning the champoinship, better than Wisconsin or MSU.
I don't know...I picked a lot of Big Ten teams because they've been playing tournament-level basketball all season. But in general pick against the big ten teams because the people in your pool are more likely to overrate them, so if they do lose you could wind up with a pretty clear road to the top of the pool.
I say keep the Live Blog going all weekend, provided MICH advances. Didnt we do that the first year MICH finally made the tournament again?
only if you moderate it the whole time.
That could get pretty ugly. And the Ban Hammer trigger finger might go nuts. lol
The professors and researchers at Georgia Tech have used LRMC (Logistic Regression/Markov Chain) to yield the most accurate results of any rating system in the past 5+ years. Take a look at the second link for mroe info on the statistical analysis, pretty fascinating stuff.
http://www2.isye.gatech.edu/~jsokol/lrmc/
http://www2.isye.gatech.edu/~jsokol/lrmc/about/
Another favorite of mine comes from the guys at MiniTab. In their method, they apply regression models to a particular rating system. Last year they used LRMC but this year, after analyzing how different systems performed throughout the regular season, they decided to use Sagarin's ratings for the models.
http://blog.minitab.com/blog/the-statistics-game/predicting-the-2013-nc… (tournament predictions)
http://blog.minitab.com/blog/the-statistics-game/using-minitab-to-choos… (determining the best college basketball rating system)
EDIT: Here's the Nate Silver analysis for anyone interested (full disclosure: i have not read it yet, only skimmed)....
http://www.nytimes.com/interactive/2013/03/18/sports/ncaabasketball/nat… (predictions/team probabilities by round)
http://fivethirtyeight.blogs.nytimes.com/2013/03/18/parity-in-n-c-a-a-m… (analysis)
March 19th, 2013 at 10:07 AM ^
I don't really get the "KP Diff" stat, what does that tell? Can someone with higher intelligence in statistics explain it to me?
March 19th, 2013 at 10:38 AM ^
I correctly predicted a national champion based on KenPom rankings (Find team who are top 10 in both offensive and defensive efficiency). Every year, there are at least 3-4 teams who qualified under both conditions. I then use my eye test to determine if they look like a champion. Last year, it was easy for me to pick Kentucky because they passed the eye test and the KenPom test.
This year? I'm not so sure. There's only one and that is Florida. Even with them qualifying under KenPom standard, this year has a lot of parity and no elite top team. Florida blew out weak competition, but when it comes to good team or close games, they wilt under pressure. They don't pass the eye test to me. I am not sure on who is the likeliest national champ out of the top 10-15 team from KenPom. There are that many team who can make a strong argument for winning it all. As I go through the bracket, I ended up with 3 B1G final four and with Indiana winning it all. They have the pieces to win it all, but I don't trust Crean.
March 19th, 2013 at 11:25 AM ^
Thanks for the guidelines.
However, this would be a whole lot easier if you just emailed me your bracket off list.
March 19th, 2013 at 11:28 AM ^
March 20th, 2013 at 12:19 AM ^
All four #2 seeds have made the Sweet Sixteen exactly once over the past 15 years. You might want to rethink this.
March 19th, 2013 at 11:49 AM ^
Seth, that KenPom tool is fantastic.
March 19th, 2013 at 12:43 PM ^
Are for wussies. You win or you lose. (Unless you actually tie after all those games). Winner take all. And mutliple brackets isn't prognosticating, it's gambling. It's craps more than bracket science. One bracket per pool, so you know who to root for in each and every game.
It really helps to know who is in your pool if you are playing to win. Some people pick to make the games more fun. There is a way (pointed out very well here) to play the pool in order to win (or at least give you slightly better odds of winning). This is awesome, Seth.
Comments