Sunday, February 24, 2008

Oscar predictions (math-nerd style)

The Oscars air tonight, and I’ve updated my regression analysis predicting the Best Picture winner. Despite a few skeptics, the model correctly predicted “The Departed” winning last year. (Lucky or smart? Who knows.)

(See last year’s post for a more in-depth explanation of the regression model. http://invisiblelawstudent.blogspot.com/2007/02/predicting-best-picture-winner.html )

I’ve gone back and updated the data to include last year’s nominees. I’ve continued to only use data from the last five years. I still think this is an excellent way to account for the changing tastes of voters. (What does the behavior of voters from the 1950s really tell us about voters today?)

The updated data again confirms that that the guild awards are still the best predictors of Oscar success. The PGA is still negatively-correlated with winning an Oscar. For movies such as “Little Miss Sunshine,” “Brokeback Mountain,” and “The Aviator,” winning the PGA was essentially the kiss of death. A PGA award can be overcome, but if a film is on the bubble, don’t expect a PGA award to help its Oscar chances. The Golden Globes and box office earnings remain insignificant predictors.

Here’s the updated prediction model for this year:

Oscar Win = SAG x .423 + DGA x .757 + PGA x -.454 + ACE x .285 + -.025

Here’s how the model would predict the winners from the last five years:















Movie

SAG

DGA

PGA

ACE

Points


The Departed

0

1

0

1

1.017


Babel

0

0

0

0

-0.025


Letters from Iwo Jima

0

0

0

0

-0.025


Little Miss Sunshine

1

0

1

0

-0.057


The Queen

0

0

0

0

-0.025


Crash

1

0

0

1

0.683


Brokeback Mountain

0

1

1

0

0.277


Good Night and Good Luck

0

0

0

0

-0.025


Capote

0

0

0

0

-0.025


Munich

0

0

0

0

-0.025


Million Dollar Baby

0

1

0

0

0.732


The Aviator

0

0

1

1

-0.195


Finding Neverland

0

0

0

0

-0.025


Ray

0

0

0

1

0.26


Sideways

1

0

0

0

0.398


Return of the King

1

1

1

1

0.985


Lost in Translation

0

0

0

0

-0.025


Master and Commander

0

0

0

0

-0.025


Mystic River

0

0

0

0

-0.025


Seabiscuit

0

0

0

0

-0.025


Chicago

1

1

1

1

0.985


Gangs of New York

0

0

0

1

0.26


The Hours

0

0

0

0

-0.025


The Two Towers

0

0

0

0

-0.025


The Pianist

0

0

0

0

-0.025



















Notice the correct prediction for Crash.

Here’s how the model predicts this year’s nominees:


Movie

SAG

DGA

PGA

ACE

Points

There Will Be Blood

0

0

0

0


-0.025

No Country for Old Men

1

1

1

0


0.7

Atonment

0

0

0

0


-0.025

Juno

0

0

0

0


-0.025

Michael Clayton

0

0

0

0


-0.025

No surprises here – “No Country” is the clear leader. However, it's PGA win shows that it may not perfectly fit Oscar voters' tastes, and its lack of an ACE win highlights its vulnerability. It's still the most likely to win, but don't discount the chances for a "Michael Clayton" or "Juno" to sneak in.

Here’s hoping this helps you take the prize for that Oscar pool. Enjoy the show!

2 Comments:

Blogger Unknown said...

Hi,

Your regression model is flawed. As someone previously commented, you need use a logit/probit model (more commonly known as logistical regression) in order to get the desired result. A linear regression using a binary 0/1 for an oscar winner is faulty because 0 and 1 don't have numerical values. In other words, the model is trying to predict 0 or 1, but you should be trying to predict true/false.

See the following two for reference:

http://online.wsj.com/article/SB110928974020363759.html

http://www.tuck.dartmouth.edu/pdf/pr20040225_oscars.pdf

5:06 PM  
Anonymous Anonymous said...

Hey there! This is my first visit to your
blog! We are a collection of volunteers and starting a new project in a community
in the same niche. Your blog provided us useful information to work on.

You have done a outstanding job!

Here is my web site Www.Dailymotion.Com

10:44 PM  

Post a Comment

<< Home