Bracket tip from Tech prof: Go with Kansas

The genesis was a halfcourt shot. Dr. Joel Sokol, an engineering and math graduate of Rutgers who took his Ph.D. at M.I.T., had come to work at Georgia Tech and begun to follow the Yellow Jackets. Even with the gifted freshman Chris Bosh, Tech finished the regular season 14-14 and missed the NCAA Tournament.

“Some people said we’d have gotten in if we’d won one more game,” Sokol said Monday. “We played Tennessee in one of those holiday tournaments and lost on a halfcourt shot.”

On Dec. 15, 2002, Tech led Tennessee 69-67 at Philips Arena with 0.5 seconds remaining. The Volunteers’ Ron Slay inbounded to Jon Higgins, whose cast from midcourt sailed true. (Watch the YouTube clip and you’ll hear Bob Rathbun yell, incorrectly: “We’re going to overtime!”)

Said Sokol: “It seemed sort of ridiculous that whether a halfcourt shot goes in could make a difference.”

He and a few fellow professors took to fiddling with numbers, as math professors will. They developed the LRMC rankings, which are now furnished to the NCAA tournament committee and are available online at Tech’s School of Industrial and Systems Engineering website. And LRMC stands for … what?

“Logistic Regression/Markov Chain,” Sokol said. Then he laughed. “If we had thought just a little bit more when we were starting — we didn’t do such a great job of (naming).”

Long story short: Going by data, the profs found that one more Tech victory might indeed have made a difference. “We would have been right around the bubble,” he said. “People were pretty much right.”

When the 2004 NCAA Tournament began, Georgia Tech was the No. 3 seed in the Midwest. The nascent LRMC rankings — as compiled at, ahem, Georgia Tech — picked the Jackets to make the Final Four. Which they, ahem, did.

“It was completely unbiased,” Sokol said. “They were a really good team and had a lot of close wins, and we got lucky as well. They made us look good.”

Sokol’s computations mightn’t be as widely viewed as Ken Pomeroy’s ratings, but LRMC is a useful tool come Bracket Season, where we find ourselves today. Unlike RPI, on which the NCAA committee relies greatly, the LRMC makes allowances for margin of victory.

Sokol stopped filling out a bracket when he started supplying data to the committee, fearing a conflict of interest. (He did lapse once, entering the $1 billion Warren Buffett contest two years ago. He didn’t win.) He lets his numbers do his picking. LRMC projects a Final Four of Kansas, Oklahoma, North Carolina and Michigan State; it has the Jayhawks winning, which would be fine by Sokol. He has a soft spot for KU.

The only year LRMC tabbed the correct Final Four was 2008, when four No. 1 seeds made it. But most folks went with North Carolina, which entered its semifinal at 36-2, or Memphis, which reached the final at 38-1. LRMC picked Kansas over both. The Jayhawks took the title after Memphis blew a nine-point lead with two minutes remaining.

“The Kansas-Memphis game would have been great fun to watch before I did this,” Sokol said. “It was a stomach-churner whole way through. I kept thinking, ‘I really need this team to win for our rankings.’ It was worse than if I had 20 bucks on the game.”

As for this year: “The math is always going to be wrong somewhere. If you look at teams — you watch them play and know someone just came back from injury, which the numbers don’t take into account — you might be an expert in an individual game. And if you’re in a big bracket contest, the more weird (your bracket is), the better.”

The LRMC bracket has Oregon, the No. 1 seed in the West, losing to Duke in the Sweet Sixteen and No. 11 Gonzaga upsetting Seton Hall and Utah to reach the third round. Otherwise it’s mostly chalk. LRMC isn’t picking any No. 12 seed over a No. 5. That’s the thing about working from a season’s worth of numbers: There aren’t many wild hairs.

Example: Wichita State. LRMC has the Shockers ranked 20th in the nation; the NCAA seeded them 11th and relegated them to a play-in game against Vanderbilt. “As much as our method likes them and as much as I personally like them,” Sokol said, “they would have to pull off a bunch of upsets to go far.” LRMC ranks Arizona No. 10 and Miami No. 18; those figure to be the next two teams in Wichita’s path.

Sometimes numbers are no fun. Syracuse has lost five of its past six games and was lucky to make the field. LRMC sees the 10th-seeded Orange as a first-round winner over Dayton. “Our method treats early- and late-season games the same, which tends to give better predictions,” Sokol said. “But I do cringe sometimes.”

Rule No. 1, in bracketology as in Lee Child’s Jack Reacher books: Hope for the best, expect the worst. Citing “the Bayesian framework,” Sokol said: “A third of the time, the (Vegas) line turns out to be wrong by 11 points or more. There’s just so much randomness.”

Thomas Bayes was a Presbyterian minister who wrote about probability in the 1600s. In 1837, Davidson was founded by Presbyterians. In 2008, Davidson and Stephen Curry nearly upset Kansas, LRMC’s darling, in the regional final. All the world’s a bracket.

To enter Bradley’s Bracket Fiasco and win fabulous cash prizes, click here.

Support real journalism. Support local journalism. Subscribe to The Atlanta Journal-Constitution today. See offers.

Your subscription to the Atlanta Journal-Constitution funds in-depth reporting and investigations that keep you informed. Thank you for supporting real journalism.