.

WSJ.com/NumbersGuy March Madness contest

.

I asked college-hoops number crunchers (bolded) to submit their brackets, then compared them to predictions from sportswriters, politicians and basic algorithms such as choosing the best seed or the best record in each game. Below they're ranked by accuracy; the bracket with the fewest points lost so far is the most accurate. One point is deducted for a first-round misfire, two for a second-round miss, four for an Elite Eight team already eliminated, eight for a Final Four team eliminated, 16 for an eliminated finalist, and 32 for an eliminated champ. -- Carl Bialik (see related blog post at http://blogs.wsj.com/numbersguy/number-crunchers-vs-experts-vs-the-masses-306/)

.

pickersourcetype1st round wrong2nd round wrongElite 8 eliminatedFinal 4 elim.Final 2 elim.Champ elim.champtotal points lost

.

Ken Pomeroy (1, 9)http://www.kenpom.com/rate.phpalgorithm641000Kansas18

.

Coke Zero (2)private emailalgorithm1050000NC20

.

TeamRankings official (3)private emailalgorithm1031000Kansas20

.

Brad Evanshttp://tournament.fantasysports.yahoo.com/t1/78254expert1050000Kansas20

.

best seeds (9)http://nbcsports.msnbc.com/id/23487117/algorithm851000--22

.

RealTimeRPI.com (4)private emailalgorithm851000UCLA22

.

David Gellerprivate email (from Yahoo Sports' director of fantasy gamesexpert851000NC22

.

George Allenhttp://www.georgeallen.com/wp-content/uploads/2008/03/allen-brackets.JPGpolitician851000NC22

.

Sagarin rating (5)private emailalgorithm951000NC23

.

Sagarin Elo (5)private emailalgorithm951000NC23

.

AccuScore core (6)private emailalgorithm1051000UCLA24

.

ESPN users (9)http://games.espn.go.com/tcmen/whopickedwhomcrowd961000NC25

.

Jay Bilashttp://games.espn.go.com/tcmen/entry?entryID=2262515expert1151000Kansas25

.

SportsLine usershttp://splnexperts.mayhem.sportsline.com/print/brackets/printable-bracket/1/6/1/crowd662000NC26

.

Fran Fraschillahttp://games.espn.go.com/tcmen/entry?entryID=740600expert852000Kansas26

.

Andy Glocknerhttp://games.espn.go.com/tcmen/entry?entryID=1177expert971000UCLA27

.

TeamRankings momentum (3)private emailalgorithm1041100Kansas30

.

Reggie Rankinhttp://games.espn.go.com/tcmen/entry?entryID=623107expert1062000UCLA30

.

Jason Kinghttp://tournament.fantasysports.yahoo.com/t1/76736expert1072000Kansas32

.

Joe Lunardihttp://games.espn.go.com/tcmen/entry?entryID=1078940expert963000UCLA33

.

best recordshttp://nbcsports.msnbc.com/id/23487117/algorithm1282000Memphis36

.

AccuScore override (6)private emailalgorithm972100UCLA39

.

Russell Adamshttp://online.wsj.com/public/resources/documents/retro-NCAABrackets0803.htmlalgorithm993000UCLA39

.

Dan Shanoffhttp://online.wsj.com/public/resources/documents/retro-NCAABrackets0803.htmlexpert1352100UCLA39

.

Mike Hugueninhttp://tournament.fantasysports.yahoo.com/t1/76544expert1082100UCLA42

.

Dick Vitalehttp://games.espn.go.com/tcmen/entry?entryID=2370921expert1272100NC42

.

Barack Obamahttp://www.abcnews.go.com/images/Politics/Obama%20Bracket.JPGpolitician1063100NC42

.

Sagarin predictor (5)private emailalgorithm661110Kansas46

.

Bracket Science (7)private emailalgorithm872200UCLA46

.

Dan Wetzelhttp://tournament.fantasysports.yahoo.com/t1/76696expert1174100UCLA49

.

Greg Anthonyhttp://tournament.fantasysports.yahoo.com/t1/74068expert1194100UCLA53

.

Bill Simmonshttp://sports.espn.go.com/espn/page2/story?page=simmons/080319expert1264200UCLA56

.

John McCainhttp://www.johnmccain.com/brackets/View.aspx?view=johnmccainpolitician983110NC61

.

Inside Higher Ed (8)http://www.insidehighered.com/layout/set/print/news/2008/03/17/bracketalgorithm15105210Davidson87

.

.

Footnote 1: Mr. Pomeroy declined to participate, so I accessed his ratings the morning of the tournament's start and applied them to the bracket.

.

Footnote 2: Coke Zero offered a bracket-picking tool that chose games based on how much emphasis you wanted to place on various categories, including serious ones such as offensive power and a less-serious one -- the wildness of the school's mascot. Mike Gerstein used the tool to fill out a bracket that emphasized star power and offensive prowess, and didn't use the mascot category.

.

Footnote 3: TeamRankings based its official bracket on its core algorithm, while adding some competitive strategy for a pool of fewer than 100 players, such as this one. Its momentum bracket placed a high importance on teams' momentum coming into the tournament.

.

Footnote 4: RealTimeRPI used a prediction formula, taking into account home-court advantage for NCAA teams playing close to their home floor.

.

Footnote 5: Jeff Sagarin submitted picks based on his three ratings schemes, which are also published on USA Today's website: http://www.usatoday.com/sports/sagarin.htm

.

Footnote 6: AccuScore, which runs simulations of games, submitted one bracket with the winners of simulations of each game, and went upset-hunting for another bracket, where teams with the least probability of winning a game for its seed number -- even if that chance was greater than 50% -- were picked to lose.

.

Footnote 7: Bracket Science's Pete Tiernan mixed his gut instinct in with the numbers.

.

Footnote 8: It's no great surprise that Inside Higher Ed's bracket fell short; it was based on teams' academic performance. The Web site's article about its bracket was headlined, "A Bracket Not to Bet On."

.

Footnote 9: The numbers for ESPN and best seeds were initially incorrect. Both brackets had Duke beating Xavier, so they missed one Elite Eight team. Also, the numbers for the Pomeroy ratings initially showed that algorithm had five games wrong in the first round. In fact, it had six.