ATLANTA -- Computer ratings won't be a part of college football's new four-team playoff. But the sport isn't shutting down the servers entirely.
The 13-person selection committee that will choose the four teams for the playoff will be armed with an arsenal of statistics, sabermetrics and other data.
The source isn't anyone associated with the old-school BCS combination of six different computer ratings. The new provider is the brainchild of three former college baseball players, and it found its way into the selection committee's toolbox by way of a Twitter connection.
A new era, indeed.
SportSource Analytics, a fledgling company, has more than 50 million statistics at its disposal, according to its founders. Former Vanderbilt pitchers Stephen Prather and Drew Borland, ex-Georgia Tech pitcher Scott Prather and recently hired stats guru Marty Couvillon are charged with supplying the selection committee with everything it needs to make educated decisions about the quartet of teams that will have a chance to play for a national championship.
"Obviously, the committee is going to watch the games and film, but at the end of the day they're going to have to compare teams side by side," Scott Prather said. "They want us to constantly keep our data up to date. We're not trying to influence the decision. We don't have a vote. We're only supplying them with a resource and information."
SportSource Analytics hopes to explain just about anything about college football since 2001, when its database of seemingly infinite statistics begins. A few examples:
• From 2004 to 2013, among active head coaches with at least three years' experience, the three coaches with the highest percentage of points-producing offensive possessions are Texas A&M's Kevin Sumlin (48.2 percent), FSU's Jimbo Fisher (46.6) and Washington's Chris Petersen (46.4). The three coaches with the highest percentage of touchdown-producing possessions are Sumlin (39.8 percent), Petersen (37.5) and Ohio State's Urban Meyer (36.4).
• Nick Saban has coached only two games at Alabama since 2007 when the Crimson Tide had "inferior talent" compared to their opponent, based on a four-year trailing average of recruiting class rankings.
• In 2013, Georgia's offense scored on 68.2 percent of its possessions against FBS foes when it made at least one first down. It scored 77.7 percent of the time when it made at least two first downs. The Bulldogs scored touchdowns only 13.6 percent of the time when they allowed at least one sack during a drive.
• SEC referee Ken Williamson's crew averaged 2.09 holding calls per game last season, nearly twice as many as every other SEC crew. Referee David Smith's crew had the fewest at 0.50 per game. (Referee Tom Ritter's crew averaged 14 total penalties per game, most of any crew.)
• Georgia Tech coach Paul Johnson's program might be built on the triple-option offense, but his defense might ultimately be what determines whether the Yellow Jackets win or lose. During Johnson's six-year tenure, the Jackets are 37-8 when allowing 150 rushing yards or fewer; they're 11-24 when allowing more than 150.
• Since 2001, no FBS team has a winning record in games in which it trailed at the half. LSU has the best record at 23-23. Texas is next at 24-27.
• Over the past 13 seasons, Kansas State had the best record when leading at the half at 81-3. Boise State is next at 133-7.
The Prather brothers and Borland came up with the idea for SportSource Analytics while Stephen Prather was in graduate school at Vanderbilt in 2007. The company's three founders were college football fans fascinated by the "Moneyball" concept employed by Oakland A's general manager Billy Beane, who used a sabermetric approach to build a baseball team. So together, after Stephen Prather drew up a mock business plan while completing his MBA, they decided to launch a coaching-search firm based on statistical data.
"After we built the database, a school would hire somebody and we'd ask: 'Why would they hire that guy?'" Scott Prather said. "It didn't make any sense."
The founders shelved their idea for a few years, while the Prather brothers started their careers in commercial real estate and Borland worked as a computer software architect. (The three men still have other full-time jobs, in addition to their duties at SportSource Analytics.) But when the real estate market slowed down after the economic recession in 2009, the Prather brothers decided to get serious about the idea. They called then-Georgia Tech athletic director Dan Radakovich for advice.
"He told us not to go the search firm route and instead sell data to athletics directors," Scott Prather said. "He told us to compile data about coaches and not teams."
Borland, who earned dual bachelor's degrees in computer science and mathematics at Vanderbilt, went to work building the database. Initially, Borland decided to focus on head coaches. Who was the best coach by winning percentage? Which coach had the best offense, defense and special teams? Which coach was the best when his team was winning at halftime or trailing at the half?
"I wanted to build a tool where people could answer questions about coaches," Borland said.
Within a few months, Borland expanded the database to include offensive and defensive coordinators. Borland's database grew from approximately 1 million statistics to more than 50 million by January 2011. The company launched the Coaches by the Numbers website and introduced it as a subscription service, in which fans could pay a monthly fee for access.
"It failed miserably," Borland said. "There aren't enough people out there who are interested."
So SportSource Analytics decided to audible in 2012. Instead of writing content and supplying statistics that fans might find interesting, they decided to focus on more detailed data that might be valuable to college coaches and administrators.
"It became so complex," Borland said. "We were grading offensive efficiency and defensive efficiency. We thought it was information that coaches would consider gold. We wanted to be considered the truth about 128 [FBS] teams."
The company's first clients were the superagents Jimmy Sexton and Trace Armstrong of Creative Artists Agency, who used SportSource Analytics' coaching dossiers to sell their clients to prospective schools. Agents Russ Campbell and Patrick Strong of Balch Sports, who represent coaches such as North Carolina's Larry Fedora, Boise State's Bryan Harsin and Cal's Sonny Dykes, also use the service.
"We'd love to be a consultant when teams start looking for a head coach," Scott Prather said. "Right now, we're already supplying data. But we feel like we study coaches more than anyone. ADs are always talking about their short lists. This is where you could go to look up anybody on a short list. If you're going to buy stock in a company, you look at historical data. You can now do the same thing with a coach."
It didn't take long for FBS coaches to figure out that SportSource Analytics knew perhaps as much about their teams as they did. Teams like Colorado, Georgia Tech, Ohio State, Ole Miss and Virginia were among the first to use the company's deep well of statistics to study their opponents and their own teams. Another half-dozen teams, including Arizona State, Notre Dame, Oregon and Wisconsin, signed up for the service this year.
Coaches, typically graduate assistants or quality-control assistants, can log on to the SportSource Analytics website and find whatever stats they're looking for. Borland and Couvillon, the founder and operator of cfbstats.com and now the company's chief data architect, load information into the database through play-by-play sheets, drive charts and box scores from games. Borland said the software parses out words and converts them into statistics and queries. He estimates the software can generate at least 70 pieces of data from one play in a game.
"It turns human language into computer code," Borland said.
Some schools have even started sending the company daily practice data in Excel sheets that include formations, personnel and down-and-distance. Coaches might use the data for self-study to eliminate tendencies in play calling and to ensure their reserves are getting on the field enough.
"Our secret sauce has been taking a concept and boiling it down to something anyone can use," Scott Prather said. "There are only a handful of people who know how to manipulate the information. We want our data to be used by anyone who can turn on a computer."
In January 2013, Stephen Prather sent a message to College Football Playoff chief operating officer Michael Kelly on Twitter. Prather told him he had a product that might help the selection committee members evaluate teams. Much to his surprise, Kelly responded. He sent Kelly a sample of the "Coaches by the Numbers" dossier and gave him an online demo of the service. The founders met with Kelly in the summer of 2013, then went to Dallas to show CFP executive director Bill Hancock, chief financial officer Reid Sigmon and Big 12 commissioner Bob Bowlsby how it worked. They were impressed.
Hancock refers to SportSource Analytics as the playoff's "outsourced, full-service sports information office."
"There's no analytics," Hancock said. "Obviously, the word analytics is in the company name, and they might be doing analytics for other clients, but not for us. There's some hangover from the BCS days of people wanting the data to be manipulated or compiled. But we wanted just raw data. That's what we asked for, and that's what they're giving us."
By the end of 2013, SportSource Analytics had built a custom platform for the playoff, and it unveiled it to the committee in January 2014. A representative from SportSource Analytics will be in Dallas every time the selection committee meets this season -- in an adjoining room, in case a committee member needs a question about a specific team answered.
"They can get the committee members what they need, almost instantly with a couple of clicks," Kelly said.
Unlike the six computer ratings used in the BCS formula for 16 years, SportSource Analytics won't offer its opinions on teams. But it's going to provide the foundation for the committee members to make theirs.
"I think the main difference in our platform and what they've used in the past is that we're trying to look at what's happening on the field," Borland said. "We're not getting into predictive analytics. They want to have something to compare teams. They don't want a number that gets spit out of a computer. We try to keep our data below calculus and above arithmetic."