Ben Zauzmer - Oscarmetrics: The Math Behind the Biggest Night in Hollywood

Oscarmetrics

Ben Zauzmer

Oscarmetrics

2019. Ben Zauzmer. All rights reserved.

No part of this book may be reproduced in any form or by any means, electronic, mechanical, digital, photocopying or recording, except for the inclusion in a review, without permission in writing from the publisher.

Published in the USA by:

BearManor Media

P O Box 71426

Albany, Georgia 31708

www.bearmanormedia.com

Printed in the United States of America

ISBN978-1-62933-440-0 (paperback)
78-1-62933-441-7 (hardback)

Book and cover design by Darlene Swanson www.van-garde.com

DISCLAIMER: OSCAR, OSCARS, ACADEMY AWARD, ACADEMY AWARDS, OSCAR NIGHT, A.M.P.A.S. and the Oscar design mark are trademarks and service marks of the Academy of Motion Picture Arts and Sciences. The appearance of the registered trademark symbol is implied wherever the terms Oscar, Academy Award, or Oscar Night, or the plural versions of these terms, appear in this book. This book is neither authorized nor endorsed by the Academy of Motion Pictures Arts and Sciences.

Contents

Id like to thank the Academy.

Opening Monologue

MARIA:

Lets start at the very beginning. A very good place to start.

The Sound of Music (1965): Nominated for 10 Oscars; won 5 (Best Picture, Best Director, Best Musical Score, Best Film Editing, Best Sound)

L auren Bacall, the legend of stage and screen, was finally on the verge of receiving her first Oscar. After a career of starring in movies from To Have and Have Not (1944) to John Waynes final film, The Shootist (1976), this was her moment. She had won a Screen Actors Guild award and a Golden Globe for The Mirror Has Two Faces (1996), and now many viewed the subsequent Oscar as a mere formality.

Juliette Binoche was the relative newcomer. Though she had been popular in France for several years, her name recognition in America was a fraction of Bacalls heading into 1996. She played a supporting role in The English Patient , which won Best Picture later that evening, and she gained some pre-Oscars momentum with a British Academy of Film and Television Arts (BAFTA) victory that some Oscar predictors ignored.

The envelope opened, the winner was Binoche. A look of shock, a walk to the stage. First words into the microphone: Im so surprised. Its true I didnt prepare anything. I thought Lauren was going to get it.

Why? Did the sentimentality of Bacalls topping off her lustrous career with an Oscar after more than half a century seem insurmountable? Does a Screen Actors Guild win plus a Golden Globe guarantee an Oscar? In truth, there were plenty of signs to indicate a close race, but neither Binoche nor the media noticed.

If youve ever read articles by expert Oscar columnists in the run-up to awards season, theyre likely relying on traditional reporting techniques like interviews and film analysis, not data and calculators.

Thats where I come in.

Rewind a few years to my freshman year at Harvard. With the 2012 presidential election just ten months away, and pitchers and catchers due to report to spring training in a few weeks, two of my favorite mathematical prediction seasons were just around the corner. I studied applied math in college, and what could be a more engaging application of math than trying to predict the future, especially the future of something as entertaining as politics or baseball?

Thats when it hit me. What about my other passion: movies? Surely, in the vast realm of the internet, there was a person who had revolutionized the art of Oscar prediction with data, the way Bill James did for baseball and Nate Silver did for politics. Heck, even one of the Oscar nominees that year Moneyball (2011) focused on mathematically predicting a field traditionally dominated by nonmathematical thinkers. I went to Google and found nothing.

My immediate reaction was that it must be impossible, or someone else would have already stepped in. But I had to know. So, like any good applied math major (or concentrator at Harvard, for some unknown reason), I went hunting for a dataset filled with everything I would need: Oscar results from every year in every category, with all of the pre-Oscars data points for each movie, and all of the same data for the upcoming years nominees. Again, Google failed me.

Therefore, being the cool college freshman I was, I proceeded to spend a solid month on the third floor of Lamont Library, not only enjoying one of the warmest rooms and comfiest chairs in Cambridge during a typically bitter Massachusetts winter, but also taking advantage of the quiet and Wi-Fi to build my own Oscar dataset. As far as I know, it was the only one quite like it in the world. I used the Academy website, Wikipedia, Rotten Tomatoes, Metacritic, IMDb, and a host of other sources, often even hunting through old press releases to collect individual data points, one at a time.

One month later, dataset finally in hand, I built some formulas. The gist of these formulas lies in weighting data points such as which categories a film is nominated in and which pre-Oscars awards a film has won. More weight goes to the inputs that have done the best job of predicting each category in previous years.

If all of the data points happen to point in the same direction, math is unnecessary, at least as far as determining a favorite goes. But what do we do when the indicators dont agree? Say the BAFTAs pick one nominee for Best Actress, but the Screen Actors Guild picks another. If were really unlucky, the Golden Globes choose a third and fourth, for their comedy/musical and drama actress categories.

As a matter of fact, thats exactly what happened in 2001. Halle Berry won the Screen Actors Guild award for Monsters Ball , and Judi Dench took the BAFTA for Iris , while the Golden Globes went to Nicole Kidman for Moulin Rouge! and Sissy Spacek for In the Bedroom .

Traditionally, we trust human intuition to weight all of these factors appropriately. But to be frank, this is where math can outshine people. There are absolutely some components of Oscar prediction that human intuition handles better than math, but weighting data points isnt one of them.

Once my formulas were ready, I made predictions in 20 Oscar categories and put them up on a decidedly second-rate website I built in a few minutes. The word Harvard often garners press coverage for stories that otherwise would never make the news. Same thing goes for the word Oscars. Turns out, if you put them together, theyre powerful enough to make even my basic website the subject of articles from around the world. The pressure was on.

The night of the ceremony arrived, and I co-hosted an Oscar party in one of Harvards Hogwartsian common rooms while nervously anticipating the results. At first, the evening only went so-so. I started off 9-for-14, much better than flipping a five-sided coin each time (one side for each nominee in the category), but not quite groundbreaking either. Things began to pick up steam as the major categories came in. My picks for both screenplay awards were correct, and Best Director and Best Actor also went according to Oscarmetrics. And then, it was time for Best Actress.

Best Actress that year was, by all accounts, a two-horse race: Viola Davis for The Help (2011) versus perennial nominee Meryl Streep for The Iron Lady (2011). A consensus was forming around Davis, including by many of journalisms most prominent Oscar predictors, for a variety of reasons some mathematically justifiable, some not so much. But the math slightly favored Streep by just a fraction of a percentage point.