Predicting the future ... in real time
As the college basketball progresses and March draws ever closer, "RPI" no doubt will be a commonly used acronym among fans discussing their team's chances* of earning an NCAA tournament invitation.
* Jerry Palm, who operates CollegeRPI.com, recently released his first bracket projection of the season, and Maryland did not make the cut. The Terrapins' RPI ranking currently sits at No. 74, and their strength of schedule slots at No. 45.
About a year ago, Chip Yu and a few others decided to enter into the crystal ball business and started RealTimeRPI.com. Using a statistical prediction model based on a multi-variable regression formula, RealTimeRPI.com projects the score of each game, the final record of each Division I squad and which teams will make the Big Dance.
"Who will win a game, that's the future," Yu, an analyst for RealTimeRPI.com, said in a telephone interview. "We want to sell and present a rating."
A team's predicted finish can change after each game, as well as after each game any of its opponents plays. For instance, on Saturday -- prior to the start of Maryland's ACC slate -- RealTimeRPI.com predicted that the Terrapins would finish 12-17 and 3-13 in conference play. On Sunday, before Maryland tipped off against Florida State, its predicted finish changed to 14-15 (5-11). On Monday, after the Terrapins had defeated the Seminoles, their predicted finish stood at 15-14 (6-10), which is where it remained as of two minutes ago.
Their projected record did not change after Tuesday's loss at Wake Forest because RealTimeRPI.com predicted that the Terrapins would fall in Winston-Salem.
Here are the factors RealTimeRPI.com considers in compiling its predictions: winning percentage, strength of schedule, opponents' strength of schedule, a team's scores, opponents' scores, winning margins and home-court advantage.
Yu said home-court advantage usually contributes four to eight points to the predicted tally of the home team's final score in each game. How many points the home team is afforded is based on its record at home in the "last couple of seasons," Yu said. Non-mathematical factors, such as atmosphere, crowd noise, etc., are not considered. RealTimeRPI.com affords Maryland 6.5 home-court advantage points.
On the surface, it would appear that RealTimeRPI.com bases its game predictions on which team holds a higher RPI ranking, but Yu insisted that is not the case. According to the Web site, RealTimeRPI.com's game predictions are accurate 76.8 percent of the time.
If you're thinking, 'Gee, 15-14 -- not to mention 12-17 -- sounds like a ridiculously low projection for Maryland,' you're not alone. Yu said he thinks Maryland will end up outperforming its current projection, but based on its body of work at the current moment, he said that's the final record the Terrapins deserve.
"The good thing about this computerized model is it's not biased," Yu said. "It's all with data, with what [the teams] have done."
As we all know, the unpredictable human element present in every game makes such mathematically based predictions faulty to a certain extent. RealTimeRPI.com did not expect College of Charleston to defeat North Carolina*. Neither, I'm sure, did the Tar Heels. That's part of the beauty of sports.
* The site did, however, predict that Clemson would beat North Carolina last night.
Some of RealTimeRPI.com's predictions seem sketchy at best (Apparently, Duke is only going to lose once the rest of the regular season.), but if you're interested in seeing another viewpoint on how the season will shake out, it's at least worth perusing.
Posted by: fushezzi | January 14, 2010 2:47 PM | Report abuse
Posted by: Barno1 | January 14, 2010 2:56 PM | Report abuse
Posted by: wbbradb | January 14, 2010 3:23 PM | Report abuse
Posted by: jpfterps | January 14, 2010 3:42 PM | Report abuse
Posted by: Barno1 | January 14, 2010 3:58 PM | Report abuse
Posted by: Rational_Man | January 14, 2010 4:06 PM | Report abuse
Posted by: cbmuzik | January 14, 2010 4:17 PM | Report abuse
Posted by: lavar609 | January 14, 2010 5:10 PM | Report abuse
Posted by: deadskin | January 14, 2010 6:08 PM | Report abuse
Posted by: petecard | January 14, 2010 6:53 PM | Report abuse
Posted by: Realness1 | January 14, 2010 7:47 PM | Report abuse
Posted by: mstov | January 14, 2010 9:01 PM | Report abuse
Posted by: carolina1 | January 15, 2010 6:30 AM | Report abuse
Posted by: petecard | January 15, 2010 8:07 AM | Report abuse
Posted by: Barno1 | January 15, 2010 11:57 AM | Report abuse
The comments to this entry are closed.