Warning, analytics lovers: The next three weeks will not be easy.

Like most sports, college basketball has been steadily infiltrated by math types who have built careers on analyzing every win, every loss, every strength of schedule — all in the name of gaining that .1 percent edge for those who take this bracket-filling thing seriously.

Then the games start. March Madness. It's a one-game, and, sometimes, one-shot proposition that may as well be called the Anti-Analytics championships.

Jim Valvano. Danny and the Miracles. George Mason. Bucknell over Kansas. VCU and Butler (The first time). From a pure statistics point of view, none of that should have happened. Because it did, we spend the next three weeks obsessing over upsets, unknowns and the unimaginable.

But the math geeks don't give up.

Among them is Ed Feng, who runs the website thepowerrank.com, which promises better predictions through analysis. A PhD in chemical engineering from Stanford, Feng decided to leave the world of polymers and plasma and devote himself to the important stuff — like figuring out who will win Thursday's Iona-Iowa State game. (By the way, picking that game, or any game in the first week, doesn't really matter in the big picture, Feng says. More later.)

He's done well. Feng's 2002 through 2015 pre-tournament numbers predicted the winners of more than 71 percent of March Madness games, a stat even he admits gets skewed a little because some things really never do happen — like a No. 16 seed beating a 1 seed. At least not yet.

About that other 29 percent?

"You really can't see everything coming, in some ways," Feng said. "The 70-foot bank shot, the fluke injury to key players. There are some things you just don't know about."

Ah, that 70-foot bank shot. That happened last week in the quarterfinals of the American Athletic Conference tournament. Connecticut was losing by three with 0.8 seconds left when Jalen Adams took an inbounds pass and launched the ball bucket-ward. It went in. Mayhem ensued. The shot forced a fourth overtime and UConn came out on top. In addition to helping UConn punch its once-shaky ticket to the dance, it served up a nice refresher in the randomness of sports — and just in time for tournament week.

Feng's job is to find that delicate balance between the unthinkable and the hard numbers, and give subscribers the best chance of winning their contest, whether it be a 10-person office pool, or some huge contest offered by ESPN. Some of this involves straight analytics — basically, analysis of who beat who, and by how much, and the strength of the opponents they played. He also offers a "contrarian" analysis, in which he studies brackets that have already been submitted to allow people to best position themselves for victory.

Last year, with undefeated Kentucky dominating the sport, Feng noticed a relative dearth of brackets being filled out in Duke's favor. He recommended taking the Blue Devils to go all the way, which, in fact, happened. Picking the eventual champion, especially if that champion isn't everyone's favorite, makes a much bigger impact on most pools than nailing five upsets in the opening week.

Of course, Duke's win didn't come without its own bit of anti-analytics — something nobody saw coming.

In last year's title game against Wisconsin, Duke was struggling — losing by nine to a veteran team that had knocked out the Wildcats two days earlier.

Coach Mike Krzyzewski went with a hunch and called on Grayson Allen, who, at the time, was an overshadowed freshman averaging four points a game.

Allen made a 3-pointer to start Duke's comeback and screamed, "Let's Go," in an attempt to bring his team back to life. He finished with 16 points and became the focal point of a comeback for a team filled with NBA talent.

"Grayson put us on his back," Krzyzewski said.

While Allen won't sneak up on anyone this year — he's Duke's leading scorer at 21 points per game — there's almost certainly another Grayson Allen lurking in some part of the country, ready to break a bracket near you.

It brings back memories of a game from 15 years ago — not a "One Shining Moment" game, mind you, but one that offered a valuable lesson nonetheless.

The great Temple coach, John Chaney, was in his fifth and last regional final, still looking for his first trip to the Final Four — a destination he would never reach.

In a game against Michigan State, Chaney's vaunted matchup zone did just about everything it was designed for. Except one. It did not account for unheralded David Thomas, a 5-point-per-game scorer who came off Tom Izzo's bench (and now serves as Izzo's director of basketball operations).

Thomas went 8 of 10 for a career-high 19 points, including a key 3-pointer with a minute left that sealed the game.

With the tears still fresh on his face, Chaney gave an explanation that summed up both the pros and cons of analytics — and did so years before they became an industry unto themselves.

"You have to look at statistics with everyone, and sometimes it works and sometimes it doesn't," Chaney said. "There's always one guy. You always deal with the known and leave the unknown alone. I've done that all my life. That's why we've won 70 percent of our games. I live and die by that."