Beard Merit Badge
Join Date: Apr 2001
Casino cash: $7338
Is this guy cdcox?
On Monday morning, Paul Bessire woke up to a world without Michael Vick. It was not a real world since Vick still has a tentative hold on the Philadelphia Eagles' starting quarterback job, but much of Bessire's life is spent in the haze of possibility. And with Vick injured on Sunday he wondered: How would the Eagles do with Nick Foles as their quarterback for the rest of the season? He plugged Foles into the algorithm that is the basis for his PredictionMachine.com site and ran 50,000 simulations of the Eagles' next 11 games with Foles as the starter. Then he did the same for the team with Vick, careful to account for something his computer program does not – Vick being healthy for a whole year – then read the results.
The computer gave Vick's Eagles a probability of winning 8.5 games this year, Foles 7.9. Because the simulation runs the entire NFL season, it showed the Eagles with Vick winning the NFC East at 8-8 and falling just short to the Dallas Cowboys with Foles.
It's a tight probability race in what QB gives the Eagles the better edge. (AP)
"It's a small difference but it's big," Bessire says. "It goes from the Eagles winning the division to the Cowboys winning the division."
This is the sports story Bessire wants to tell. It's one that is different from the sports journalists who stand in cluttered locker rooms, but it is a sports narrative nonetheless, built with numbers instead of quotes as corroboration and produced by a man who is 31 with a master's degree in quantitative analysis. And it is a story professional teams seem to want him to tell them.
He says several professional sports teams have used his information, looking for ways to predict player performance. Among his biggest sports clients, he says, are NFL teams. He has come up with a way to predict how college players will perform in the NFL and he says six teams have been interested enough in his findings to pay for them.
In an NFL where every bit of information is considered a secret, no matter how trivial, Bessire can't reveal the names of the teams to whom he provides information. And those teams would never admit using it, certainly not as a primary source of evaluation. The closest confirmation comes from one club that acknowledges the existence of Bessire's system, but there is no sense as to how heavily it is used.
[Weekly draft report: Jadeveon Clowney commitment an NFL concern?]
The pro teams buy his information, even though Bessire's primary customer is a gambler looking for an edge. His website has an unmistakable appeal to a sports bettor with references to records against the spreads. They buy his analysis because they apparently see value in his algorithms. He knows this because one NFL team chose his best available player four straight times in last year's draft. Another took his best available on three consecutive picks.
"It blows my mind," Bessire says of the idea that NFL teams are making picks based on his draft lists.
His connection to the NFL teams comes from his draft expert, Matt Richner, who was an intern with the Seattle Seahawks in 2008 and early 2009 while studying quantitative and qualitative analysis in graduate school at Seattle University. Using league contacts made during his internship, Richner was able to develop a client list, Bessire says.
Alabama is great, but not NFL great in a hypothetical matchup against the worst of the pros. (AP)
He obviously is not alone in doing this. Teams in many sports are using statistical analysis to more accurately predict performance. But what is not clear to the public is how much that information is being used.
"There are teams that have a good ol' boy perception that are much farther along in this regard than people know," Bessire says.
Much of his draft analysis comes from assessing how a college player will do against professional competition. To do this he looks at a variety of factors – like a player's scouting combine results compared to other NFL players' combines. He also studies what a player has done on the field, adjusting for things like the level of conference in which they play, coaching staffs and teammates. Then he puts it all in the computer and waits a few seconds for the results.
As a way of illustrating this, he and his staff recently ran a simulation to see what would happen if Alabama, college football's top team, played the four winless teams in the NFL (Jaguars, Giants, Steelers and Buccaneers). The program said the NFL teams would win more than 90 percent of the time.
Bessire promoted the Alabama vs. pros simulation in a press release and it was picked up by media outlets across the country, intrigued by the possibility of such a matchup. In part because of these random simulations and also his work with teams, he has become a bit of a media sensation, appearing on dozens of sports radio shows and making occasional contributions to several sports websites. There is a bit of hype to his work. For instance, he doesn't need to simulate a season 50,000 times. Three hundred is the magic number. But he figures 50,000 sounds better.
And while he seems to enjoy the attention, he is at heart, a man in love with sports and the numbers that explain it. He grew up in Wisconsin, tall and athletic and brilliant at math. "He's a nerd but he doesn't fit the nerd set either," says Mike Fry, an associate professor in Cincinnati's business school who first met Bessire as a student at the school.
To perhaps best understand the way Bessire thinks, you have to know how he picked his college. He applied to every school ranked in college basketball's top 25 at the time as well as the Ivy League. His final choices were Arizona, Utah, Yale, Cincinnati and Duke. He narrowed those to Cincinnati and Duke and ultimately chose Cincinnati because of its highly rated business school. It didn't hurt that the Bearcats were ranked No. 1 at the time.
There was no sense he would turn predictions into a career. He had no interest in gambling when he was younger and is quick to say he does not consider himself a bookmaker or is seeking a following of bettors. Rather, he was intrigued by a website called Whatifsports.com that tried to determine the result of historical fantasies. What if the 1976 Reds played the 1927 Yankees? When he started a six-month internship at Western & Southern Financial Group in Cincinnati he was stunned to learn Whatifsports was in the building next door. Soon he was doing regular work for the site.
He worked with Fry on his master's thesis: an analysis of how different NBA players interact with each other while on the court. The resulting document ran roughly 80 pages "with a lot of math," Fry says, and was a change from the flood of theses on business analytics that came pouring in from other students.
Not long after, he went into business for himself, soon learning there was a market for people who discovered he had an uncanny ability to successfully predict sports results. He got a fulltime job with WhatifSports. He worked there a few years until deciding late in 2009 to leave and start his own business, which he did in 2010.
"He was willing to take a chance," says Mike Magazine, a professor of Bessire's at Cincinnati who now teaches an NCAA tournament predicting class with him. "He went out on his own and built his company from nothing. We were very concerned. He is very bright and he could have gone off in business and done well. But he did this."
"Everything I need to do a simulation comes from an injury report and a box score," Bessire says. "That confuses people because there are so many advanced analytics out there."
It's an amazing claim given the swarm of numbers that fly across computer screens. And yet it makes sense. Every statistic, no matter how complicated, comes from something. A box score can seem like a rudimentary accounting of the most basic of measurements in a game – passing yards, rushing yards, fumbles kept and lost – but those figures tell truths that become the cornerstone of larger truths.
For Bessire, the most important of statistics are per-play averages. Nothing is more significant to him than a team's cumulative offensive yards per rush, offensive yards per pass, defensive yards per rush and defensive yards per pass. Everything spins from those results.
Bill Belichick's weekly Patriots injury report presents a challenge for Paul Bessire's analysis. (AP)
"There is nothing more sophisticated out there than a box score," he says.
[Related: Andrew Luck doing fine for Colts as Peyton Manning destroys the league]
He uses these base statistics whether predicting a player's performance or a team's potential. There are other mitigating factors, of course. Strength of schedule is important, especially when trying to determine how well a team will do, as are injury reports.
Bessire has learned to parse the documents that teams hand out daily, putting a percentage of probability that a player's pulled groin or sprained ankle will allow him to play in the next game. Using the Green Bay Packers' yards per pass statistics are little good if Aaron Rogers can't play. Over the years, Bessire thinks he has come to understand the silent language of the vague injury report – assessing trends for many of the coaches who release their information in predictable ways. The biggest exception to this is New England Patriots coach Bill Belichick, who releases injury information in a fashion that makes it impossible to properly assess.
"I think he enjoys doing that," Bessire says.
The beauty of the numbers is finding simplicity in the complex. One of Bessire's great successes is correctly predicting all 11 NFL playoff games against the spread in 2011. This came in the middle of a run in which he went 20-3 in postseason predictions. He assessed a wide variety of numbers, accounting for among other things, each team's strength of schedule. But in the end, only one statistic mattered most: his algorithm kept showing that the teams with the best pass rush would win every time.
Then they did.
"I don't want to say that a great pass rush and a good quarterback are the most important things for winning football games, that would be bad for business," he says with a chuckle. "But certainly getting to the quarterback is one of the biggest factors."
On Wednesday, Bessire broke his prediction machine – a device he calls the Predictalator – which is not a gigantic HAL-like entity but rather a server in Texas. He was trying to run simulations on this Sunday's Broncos-Jaguars game that is gaining fame as having the largest line in the history of the post-merger NFL, and ran into his own checks and balances. You see, Bessire built into his algorithm a block to keep people from entering ridiculous scenarios. Apparently, the Predictalator thought the 28-point spread by which the Broncos are favored is too absurd to consider. It refused to take the data he tried to input.
He was stuck. He spent the next hour with his brother Joel, who is in charge of the site's technical aspects, trying to find a way to override the problem.
Eventually they did. The Predictalator hummed again with possibility.
And the business of simulating the world of football could begin again.
I found this really interesting as well. Goes a long way in explaining why the Chiefs are doing so well this year.