FML 301: Analysis Guidelines

Pete Johnson

This is the third and last in a series of three posts intended to educate new players on different aspects of Fantasy Movie League. Special thanks to M37, who contributed ideas (and is quoted) in this entry and to Ari for edits theater count and weekly drop content, and to Phil's Phul Phlicks for his Wednesday showings posts.

First, you learned about the basics, the nuances of the bonuses, and the weekly calendar. Then you got exposed to a wide variety of resources you have at your disposal. In this final educational post for new players I'll discuss some guidelines that have proved useful when deciding on your weekly screen lineup.

Beginning with the Fall 2015 season, I began to run a weekly series of data analysis articles. Some of them turned out to be very insightful while others, not so much. In this post, I'll cut through the weeds and highlight the conclusions that have proved most useful in shaping decisions.

The most important thing to understand, though, is that even if you can use data to refine your choices each week, there is still some guesswork and often you have to "go with your gut". If predicting weekend box office revenue was a perfect science, Hollywood would make money on every movie it produces and every FML player would have the same screen combination every week. It's this variability that keeps it all interesting.

Accuracy of Professional Forecasts and New Films

The professional forecasting sites are extremely useful to FML players but they are far from perfect. In fact, it's important to note that their main purpose is to help actual theater managers nationwide with staffing and concession inventory decisions prior to each weekend so they can cater to their movie-going customers effectively.

The most comprehensive article I've written on the subject so far was in Week 9 of the Fall season. There, I reasoned that for returning films, if you average the and ShowBuzzDaily forecasts you get an accurate prediction around 75% of the time within +/- 10% of the forecast.

New films, though, are much harder, even for the professionals. ShowBuzzDaily has shown to be a little more accurate than others, but even they are only right no better than 50% of the time. There are more unknowns with a new film where with a returning one there are previous box office totals and comparable films (both in terms of genre and time of year) to use as guidelines for how much revenue drop there will be from week to week.

Since the Week 9 article, running regression tests on a larger sample than what was presented in that article, I found that +/- 15% works best across all films and that's what I use in my model for my weekly picks post.

Best Performer Candidates

Not every film matters every week when it comes to choosing FML line ups. During the Fall Season I found that most of the time only the top 4 BP candidates matter, but since then using regression tests and an average of professional forecasts I found that the top 6 BP candidates works best for my weekly model. While that reduces the number of viable combinations to choose from, it hardly makes a selection for you. Most weeks at least one of the top BP candidates are new films and given the inaccuracy of even the professional forecasts for them, it becomes clear where the challenges lie in playing FML.

Even in weeks that heavily involve returning films, the FB$ pricing is often good enough that even a swing within that +/- 15% can decide which film gets Best Performer and which one doesn't. In Week 5 of Fall 2015, had "The Intern", a returning film, sold $1,272 less in tickets it would have changed what film earned BP and which combination eared the then perfect $10M bonus. Sometimes, results can be that close and be well within margin of error on the professional forecasts.

Theater Counts

Sources like BoxOfficeMojo will publish theater counts, but as FML player M37 likes to point out, it is important to understand what that number really means. As he explained in an email to me during the Fall 2015 season:

For example, your local 20-plex may have Hotel Transylvania on four SCREENS, but it counts as only one THEATER. The data listed at BoxOfficeMojo (and most of what is reported) is theater counts, the numbers of cinemas which are showing a film.