January 6 Weekend Forecast: 'Underworld' Opens the New Year
To begin the new year, our data science team is proud to announce a weekly box office forecast series.
At Pilot, we’ve developed an array of sophisticated models that can provide a variety of different analyses depending on our clients’ needs. For instance, short-term forecasts calculated the week before a film’s release might rely more heavily on models using social media data.
On the other hand, long-term forecasts calculated a year or more before a film’s release — sometimes even before teaser trailers are released — might rely more heavily on models using historical data.
These weekly forecasts will tend to fall into the former category, although we may occasionally draw a comparison between short-term and long-term forecasts to illustrate how last-minute social media data can prove useful in identifying sleeper hits or surprise box-office bombs.
Below, you’ll find our team’s confidence interval prediction for the first major release of 2017, Underworld: Blood Wars. All dollar amounts are rounded to the nearest thousand, and represent domestic opening only. A larger confidence interval may represent greater variance. (Although both Hidden Figures and A Monster Calls expand wide this weekend, we’ve chosen to limit our forecast to the lone new wide-release film.)
|Title||Lower Bound||Mean||Upper Bound|
|Underworld: Blood Wars||$9,346,000||$13,936,000||$18,525,000|
Why confidence intervals? Each of our models is trained on data spanning thousands of films, allowing us to perform more sophisticated and granular analysis compared to a traditional comps spreadsheet. This is because all of a film’s possible comps (and more) can be considered when training a model. In order to ensure that our prediction is as close to the true box office value as possible, we employ a Markov chain Monte Carlo method to generate a confidence interval based on thousands of iterations of our models.
However, it is important to note that these intervals may not be normally distributed. What does this mean? Although we provide a lower bound, mean, and upper bound, it might not be the case that this distribution is shaped like a traditional bell curve (below).
Instead, it might be the case that the distribution is skewed left toward the lower bound or right toward the upper bound. The distribution might even have multiple peaks (below).
In this forecast for a different unnamed film, the distribution of possible box office values is clustered at $16 million and $20 million, signifying that the model is equally confident in these two forecasts. A simple look at the mean forecast doesn’t end up telling the full story, and it turns out that the upper bound is a better forecast than the mean in this case.
Each individual forecast is different, and it’s our mission to approach each analysis with statistical rigor. Our weekly forecasts won’t typically dive into this level of technical detail, but by subscribing to our reporting services, you can read additional in-depth analysis like this for future films. Our reports also analyze social media data to provide granular film marketing recommendations based on audience demographics and brand synergy.
Learn more at our official website!