BCS computers and the 'Pirates' solution

ShareThis

Try to pin down Jeff Sagarin on which of his three college football rating formulas is the most tried and true, and he'll likely turn the table with an algorithm involving the 1960 World Series.Sagarin, 60, is an MIT grad and one of the early pioneers in the development of a methodology for rating sports teams. He's been providing ratings to USA Today since 1985.His computer ratings are among six utilized by the Bowl Championship Series -- in conjunction with the Harris and USA Today polls -- in determining its weekly standings.The reason The Unbalanced Line questioned Sagarin about his ratings this week is the problematic disparity between his overall ratings and the ratings he provides to the BCS.In Sagarin's overall ratings, Penn State is first, Texas second, Florida third, Alabama fourth, and Texas Tech fifth. If those were the ratings provided to the BCS, Penn State would be No. 2 in this week's standings, and Texas Tech would be third.In his so-called ELO-CHESS ratings, the Nittany Lions are fourth, and those are the ratings Sagarin provides to the BCS.Even though the two polls and the computers are each weighted a third of the BCS standings formula, it's clear the computers still wield clout.Despite its No. 2 ranking in the polls, Penn State wound up third in the standings because of its computer average of four. Texas Tech finished in a first-place tie with Alabama in the computer ratings, which enabled the Red Raiders to jump Penn State in the standings, despite their No. 3 ranking in the polls.Sagarin also has Texas Tech ranked No. 1 in his ELO-CHESS ratings. He has Texas ranked third; Florida, seventh; and Alabama, second. The reason for the difference between his overall ratings and his ELO-CHESS ratings? In the ELO-CHESS ratings, margin of victory is not factored into the equation. And that's the way the BCS wants it.In 2002, the BCS commissioners decided it wasn't appropriate to include margin of victory in its standings, evidently because of the perception it was motivating teams to run up scores to earn BCS "style points."Sagarin's third rating formula, the PREDICTOR, takes into account score margin. He says it's more accurate in predicting upcoming games, but it's not as "politically correct" as ELO-CHESS.In his PREDICTOR ratings, the top five teams, in order, are: USC, Florida, Penn State, Texas and Oklahoma.His overall ratings are a synthesis of the two.But which is the trustworthiest? Here's where the conversation with Sagarin turns to the 1960 World Series. The New York Yankees outscored the Pittsburgh Pirates, 55-27. It was the most lopsided combined score in Series history. Yet, amazingly, the Pirates won in seven games.With regard to the BCS computers, Sagarin says, "The argument always comes back to the Pirates and Yankees. The computers are all using the Pittsburgh Pirates Solution." Which is to say, they're removing score margin from the equation.Sagarin agrees score margins are relative -- to a certain degree. But he also says if he were a BCS commissioner, he'd operate the standings the way they're doing it, without factoring in margin of victory. Why? Because it has the "most collective support."The combined score in the 1960 Series suggests the Yankees were the better team, but wins and losses say otherwise. Who's right?Sagarin ran those World Series numbers, using the synthesis of his ELO-CHESS and PREDICTOR formulas and... the Pirates and Yanks came out dead even, both with an overall rating of 5.00.Just for kicks, let's test Sagarin's three formulas on the Penn State-Iowa and Texas Tech-Oklahoma State games Saturday and see which, if any, proves to be more accurate.By his hypothetical ELO-CHESS calculations, Penn State beats Iowa by 17, and Tech beats OK-State by 12. His PREDICTOR has Penn State winning by 8, and Tech by 4. And, in his overall ratings, Penn State wins by 14, and Tech by 7.The Unbalanced Line's unformulated call: The Nittany Lions win by 11; Tech gets upset by a field goal; and the computers drop the Red Raiders several spots.As quirky as the BCS computers can be sometimes, they're every bit as credible, if not more so, than the Harris and USA Today polls.Consider this: Texas is ranked sixth and seventh respectively in those polls, a spot below No. 5 Oklahoma in the Harris and three spots below No. 4 Oklahoma in the coaches' poll. The Longhorns' computer average is 3, the Sooners' is 6.Who's right?Texas and Oklahoma have 8-1 records, and the Longhorns beat the Sooners, 45-35, on a neutral field. The Pittsburgh Pirates solution makes it a no-brainer.The Unbalanced Line prefers the Obama solution: an eight-team playoff.(Contact John Tucker at jtucker@unionleader.com)(Distributed by Scripps Howard News Service, www.scrippsnews.com.)