Tuesday, January 26, 2010

The VFX Predictinator, Part 2

Just joining us? Please read "The VFX Predictinator Part 1."

So, how could anyone accurately predict the winner of the visual effects Academy Award using known, quantifiable data? My wife’s little idea suddenly became an obsession. We analyzed the last 20 years of Oscar nominees and winners and started playing with various quantifiable pieces of data and tried to find (or create) some sort of logical narrative behind Academy voters’ choices. What causes the over 6,000 Academy voters to choose these films? What influences these voters?

Critical Acclaim and Box Office
A consensus of critical acclaim from movie critics and strong domestic box office gross both play a significant role in the film's victory. These were the two bits of criteria that drove my first wave of research, and although these criteria did not prove to be slam dunks, the data illustrated that stronger films were more likely to win the Oscar. We recorded acclaim with the Rotten Tomatoes Tomatometer, which gauges acclaim on a percentage scale. Our earliest year of analysis is 1989, since beyond that year, we felt like the aggregate Tomatometer score begins to weaken due to fewer reviews in their pool. For box office tally, we looked at the film's final domestic box office gross (from boxofficemojo.com), and assigned each film a box office value relative sum of each year's nominees, to reduce inflationary discrepancies.

The strong critical acclaim for 2000's "Gladiator" gave it a strong advantage relative to its fellow visual effects Oscar nominees "The Perfect Storm" and "The Hollow Man," which both earned much lower Tomatometer scores.

The Tomatometer point value was weighted more heavily than box office as acclaim is a much better indicator of who will win the Oscar (see my previous work at predicting the Oscar for more discussion on this topic).

Additional Academy Nominations
We noticed that films that ride a wave of additional Academy Award nominations grab the gold for visual effects more frequently. Generally speaking, if a film with technically well-done effects (yet comparatively unremarkable or un-groundbreaking) ends up with a boatload of additional Oscar nominations for acting, directing, or art direction, more likely than not, Academy voters will jump on the bandwagon muttering to themselves "well, that movie got 12 nominations, so the effects must be pretty good! I'll vote for it!" The most extreme example occurred in 2000, where the effects Oscar went to "Gladiator" (which earned 12 total nominations that year), rather than "The Perfect Storm" or "The Hollow Man" (which earned 2 and 1 nominations, respectively). The most lopsided year in this category of criterion was 1997, where "Titanic," the winner of the Oscar, earned 14 total Academy nominations that year, while its competition, "Starship Troopers" and "The Lost World" earned only one nomination each. If a film earned four or more Oscar nominations, we assigned it a point value.

"Gladiator" was nominated for twelve Academy Awards in 2000, while its two fellow nominees earned a total of three nominations.

Month of Release
Producers of 'prestige' pictures--films that are generally regarded as 'important' and Oscar-worthy--know exactly the best time to distribute their films. A film debuting in the fall, particularly November and December, will be in the forefront of Academy voters' minds. With their film fresh in their memory, producers have a much better chance of nabbing Oscar nominations and wins, versus the same film (with the same acclaim) released in February or March. With this evidence, we decided to include the calendar month of release to the formula. For example, we noticed that in each of the "Lord Of The Rings" film victories, the movies defeated films that were exclusively released in the summer; the "Lord of the Rings" movies were each released in December. We assigned point values to nominees' month of release, with films later in the calendar year earning a higher value.

Sequels and Oscar-Winning Predecessors
What about battling the 'been there, done that' attitude of Academy voters? In looking at past visual effects Oscar winners, we noticed that sequels to films had a slight disadvantage in voting. The Academy may feel a sequel's effects were good enough to get a nomination, but decided against voting for it since, generally, sequels tread on familiar, well-tested ground. For example, "The Lost World," may have suffered from this issue, since voters may have thought, "well, didn't they do convincing dinosaurs in the first film? I don't want to reward a retread." In addition, we noticed that sequels whose predecessor actually won the Oscar for visual effects also had a slight statistical disadvantage. 2007's "Pirates 3" had the double disadvantage of not only being a sequel, but a sequel to a film that previously won the Oscar. The year this criterion made the greatest impact was 1992, where "Batman Returns" and "Alien 3" lost the gold to non-sequel "Death Becomes Her." Both of the losing films were sequels, and "Alien 3" took an extra hit since its predecessor, "Aliens," won the effects Oscar in 1986. For these two categories, we assigned negative point values.

Up until this point, I've been listing 100% quantifiable, undeniable criteria to help drive our formula. But to push the formula further into the realm of invincibility, we had to come up with two more criteria that aren't 100% quantifiable, and may be controversial. We labored over these final two pieces of criteria, and pared them down to the most barest of definitions. Let's dive in!

Trevor Wood, Ben Morris, Bill Westenhofer and Mike Fink with their Academy Award for "The Golden Compass."

Organic Creatures and Facial Animation
Looking at the Oscar winners over the past 20 years, a specific trend is undeniable. The visual effects Oscar goes to a film that features synthetic, organic characters. This phenomenon became much more significant in the post-"Jurassic Park" era, where advances in computer graphics allowed filmmakers to tell extraordinary stories with fully animated characters that play a significant role in the narrative. We asked of each film: does the movie's primary visual effects consist of organic creatures? And, as a second piece of criterion, if so, does the film contain CG facial animation, i.e. organic acting?

Can you see where this is going? "Babe" beat out "Apollo 13;" talking barnyard animals trump space travel. "Fellowship of the Ring" won over "Pearl Harbor;" Balrogs and cave trolls trump exploding airplanes. "Benjamin Button" tops "The Dark Knight;" CG human trumps miniature car chases. And, most significantly, "The Golden Compass" topples "Transformers;" talking polar bears trump robots made of metal and chrome. Academy voters have a strong preference in voting for films with organic creature effects, particularly with facial acting performances, rather than films whose visual effects surround action set pieces, extraordinary environments, or digital stunt work. We awarded films a point value for these two criteria.

Even with all of these fairly accurate criteria under our microscope, we still had a problem accurately predicting certain years' competitions. "Death Becomes Her's" 1992 win and "What Dreams May Come's" 1998 win were naggingly baffling. Even though "Death" was not a sequel like its competition (giving it a slight edge), it suffered at the box office compared to fellow nominee "Batman Returns," especially with the 1-2 punch of "Batman's" huge box office and strong critical acclaim. Similarly, in 1998, "What Dreams" had strong statistical competition with "Mighty Joe Young." Both films had nearly identical box office and critical acclaim, but "MJY" had an edge with its organic creature work while "What Dreams" primarily had environmental visual effects.

We had a difficult time crafting a formula that correctly predicted "What Dreams May Come" getting the Oscar over "Armageddon" and "Mighty Joe Young." And then, Robin Williams showed us the way.

Lead Actor with an Academy Award
What made "Death" and "What Dreams" special? We came to the realization that these films had something in common most visual effects films don't have: lead actors with some serious performing prestige. Both Meryl Streep and Robin Williams, leads in "Death" and "What Dreams," previously won acting Oscar statuettes. This is a fairly rare occurrence for visual effects films. When looking through visual effects nominated films through the last 20 years, you frequently see lead actors of serious stature (Johnny Depp, Robert Downey Jr., Will Smith, Jeff Goldblum) but very rarely will you see a lead actor with an Academy Award under its belt. In fact, the phenomenon only occurred seven times among sixty nominated films. We created the criterion, 'Has the lead actor previously won an acting Academy Award?', and gave it a point value.

Meryl Streep's presence in "Death Becomes Her, might have given Academy voters one more reason to vote for the film.

Adding this last bit of criterion helped "Death" and "What Dreams" defeat its competition; the thought is Academy voters may have been swayed by the presence of Streep and Williams in those films. Or it could have been a desperate attempt to skew the data so our formula became fool-proof. Or, maybe it's a little of both.

For each year of analysis (1989-2008), we added the point values of each nominee, and the film with the largest point value became 'our pick' for the winner of the Academy Award. And in each and every year from 1989 to 2008, the 'our pick' was, indeed, the winner of the Oscar.

In Part 3, we'll summarize our findings, and give you a look at how the formula works. The formula that correctly predicted the winner of the visual effects Oscar for twenty out of the last twenty years. Read Part 3.


Unknown said...


I see a second career here!


mrfusion said...

Great article and fantastic work on the Predictinator, but I don't think you'll need it for this year. :)

Akira said...

Everything makes sense so far...

ccamp_fx said...

You should adjust your scoring technique to take into account the changing of the Academy voting procedures for visual effects. Pre-1995, the award was voted by the effects members only, while 1995 and beyond are voted by the Academy as a whole. Starting with 'Babe', you will have the award voted on by a much larger percentage of people who do not understand the technical aspects of what was involved of bringing the effects to the screen, and possibly will allow their votes to be skewed by the other factors you mentioned: box office, number of other noms, how well they liked the film as a whole, seeing something new or different compared to the other nominees, et. al. With the larger, less-knowledgeable-about-effects voting block (i.e., actors) determining the outcome of the category, actually winning award should be seen as less important as just being one of the three nominations, which are still decided by effects peers.

Todd Vaziri said...

ccamp_fx said...
>You should adjust your scoring technique to take into account the changing of the Academy voting procedures for visual effects.

Meh, sounds hard. :)

At one point I considered doing exactly that, but we noticed the 1989-1995 and post-1995 voting behaviors were actually quite similar (which allowed us to achieve 100% accuracy with our formula). If we couldn't get 100% accuracy, I'd be more interested in splitting up the years into two separate formulas... but our universal seems to work.