Remember that guy Nate Silver from Chicago, who was all over the news during election season with the incredibly accurate predictions on his website FiveThirtyEight.com? He had computers crunching vast arrays of numbers. He also does well in baseball season. Now it's Oscar season, and he's baaaack.
This is depressing to a seat-of-the-pants guy like me. If you entered the Sun-Times Outguess Ebert contest, you have a good chance of outguessing me. That's based on my record over the last 10 years (a lukewarm 68%) and the near-certainty that several entrants will have perfect scores. Still, I've had my good years. In 2005, I got 82% in the contest categories. In 2006, 91%. Don't ask about 2001, with 45%. (Russell Crowe in "Gladiator" beat Ed Harris in "Pollock"? No way!)
My problem is I allow myself to be overcome with emotion. I tend to predict winners I love, and that may be why I picked Ellen Page for 2008 best actress in "Juno," and Marion Cotillard won for "La Vie en Rose."
What Nate Silver does has nothing do to with matters of the heart. His system doesn't even involve having to actually see the movies. It's all statistical manipulation. Here are his website's latest predictions:
Best Picture: "Slumdog Millionaire" (99.0% chance of victory) Best Actor: Mickey Rourke, "The Wrestler" (71%) Best Actress: Kate Winslet, "The Reader" (68%) Best Supporting Actress: Taraji P. Henson, "The Curious Case of Benjamin Button" (51%) Best Supporting Actor: Heath Ledger, "The Dark Knight" (86%) Best Director: Danny Boyle, "Slumdog Millionaire" (99.7%)
Silver says he's almost certain he'll get one of those wrong, and the obvious candidate is Taraji P. Henson. All the conventional wisdom lines up in favor of Penelope Cruz for "Vicky Cristina Barcelona," so I may get that category wrong, too, because I persist in believing Viola Davis has a chance for "Doubt."
What Silver is doing, in doubting his computers, is revealing he may have seen some of the movies after all, and has heretical doubts about his own system. Yet he was way out ahead last year, and nearly alone at the time, in correctly predicting the victory of Barack Obama.
What is his secret? On the Oscars and politics, he isn't saying. Thanks to Wikipedia, however, I can reveal his QERA formula, which refers to his "Quik Earned Run Average" for judging the performance of a baseball pitcher. Here it is:
QERA =(2.69 + K%*(-3.4) + BB%*3.88 + GB%*(-0.66))
You're very welcome. Although his movie formulas remain secret, I think I know why he is even more sure (99.7%) that Danny Boyle will win as Best Director than he is that Heath Ledger (86%) will win as Best Supporting Actor, even though everyone thinks Ledger will win. That's because Boyle won the Directors Guild of America award. Only six times since 1948 has it failed to predict the winner. Add that to "Slumdog's" best picture lead, and Boyle does better than Ivory Soap (99.44%).
Still, humans sometimes beat computers at chess. Who is best among human predictors? David Poland’s "Gurus o' Gold" at his MovieCityNews site have a splendid record of 20 of 24 in the last three years. Here are their latest leaders, reflecting how strongly they feel on a 1-10 scale:
Slumdog 9.06 Actor: Penn 5.78, Rourke 5.25 Actress: Winslet, 6.53, Streep 6.0 Supporting Actor: Ledger, 9.31 Supporting Actress: Cruz 6.54; Tomei, 6, Davis 4.75. Director: Boyle 8.81.
What does the human say about the computer program? "538 is silly and has nothing close to the statistical data that he does in any election," Poland told me Friday. "As far as I can tell, there is almost no love for that film ["Button"] or what is seen as a bit of a Stepin Fetchit role performed as well as she or anyone could do it. I would say that Melissa Leo [Actress] or Andrew Stanton [screenplay] are many times more likely as an upset."
Poland adds helpfully: "I think he's brilliant, but after years of talking Oscar stats as a measure, one becomes utterly aware that it's a fool's errand."
What it amounts to is that Actor and Supporting Actress are the tricky categories, and after the Sunday night ABC Oscarcast all of these calculations will be meaningless. You're welcome!