Here we are in our second week of extended showtime data and I thought I’d share an analysis I’m not so sure I believe in but that might spark better conversation on the topic. I got inspired by Phil’s Thursday thread but felt like what I wanted to say about it didn’t lend itself to a Chatter reply. This would require tables and some math reasoning explained.
So while this might be full of crap, I might also be onto something.
Postgame Analysis on Spring Week 8 Numbers
If you’ve been following this development in the Chatter, you know that last week I produced a chart on Thursday that is a riff on the showtimes reporting that Phil has been doing since the Fall season, only now it includes 21 cities showtime data instead of just one. Here’s a table I made that extrapolates that data a bit with some other actual numbers from last week I’ll explain below the chart.
The first four columns come from the original and the Showtimes/Theater simply divides the Total Showings by the Theaters to give the average number of showtimes in theaters from our 21 city sample. Total Theaters is the number of theaters nationwide from last weekend as reported by BoxOfficeMojo’s final tallies for the week.
Then I made a big assumption that may not be valid. If you assume that there are a similar number of showings nationwide per theater as there are in our 21 city sample, you can get an approximate number of Nationwide Showings by multiplying the Showings/Theater by the Total Theaters. Among the reasons this might be completely invalid is that the number of theaters in our 21 city sample, despite representing more than 44% of the televisions in the country, appear to account for a far fewer percentage of the movie theaters. I don’t show it above but if you divide Total Theaters by Theaters above you get a wide spread among the different movies between 10% and 38% of the overall theaters represented by our 21 cities.
I attribute that to there being a very long tail of smaller theaters in small towns all over the country. I have relatives in a Louisiana town of less than 10,000 people and they have a movie theater but I bet it doesn’t have as many physical screens as the theater at LA Live next to Staples Center, so it has no hope at having nearly as many showings. I’m not sure how to reconcile that quite yet, but I thought it was worth at least admitting it is potentially an issue.
With that as a caveat, if you divide the actual weekend returns (from the same BoxOfficeMojo page) by that estimated number of Nationwide Showings, you get the average $/Showing. We know two things about this number:
- It’s high, for the aforementioned reason based on the assumption.
- It’s disproportionately high for low release films like “Compadres” since they likely occupy no theaters at all in small towns across America.
Despite those known weaknesses in this math, can all this help us this week?
Week 9 Showings Derivative Analysis
There are lots of analysis that are based on theater counts dropping combined with the per theater revenues, but what we have here is an admittedly flawed version of the same thing, only with the more specific data of showtimes. Showtime data is a proxy for what the local theater manager thinks about a returning film based on how many people attended it last week. He or she knows how full each auditorium was, knows the appetite of the local market, and can use those factors and more to influence how often a movie gets shown the next week, putting more fine a ceiling on its performance.
So along those lines, I produced the following:
This table uses the same logic as the first, using the 21 city showtime report from this week as its basis through the Nationwide Showings column. It then lists the average of ProBoxOffice.com and ShowBuzzDaily forecasts for this week as used in my model, extrapolating unforecasted films based on the average value of the forecasted ones.
How well would a movie have to do in order to make that average, given the estimated number of Nationwide Showings? That’s the Needed $/Show column. How does that compare to how that returning film did on a per show basis last week? That’s the Last Week Delta column.
As an example, if “Barbershop” lost 4% of its per showing audience it will meet its averaged pro forecast. Negative numbers for this metric are shown in green because it seems logical that a movie will naturally lose even per showing audience, so the larger that number is negative the better chance a film has at making its forecasted number.
By contrast, “Huntsman” would have to gain 50% of its per showing audience in order to make its averaged pro forecast, which seems highly unlikely and is why positive numbers in this column are shown in green.
Films for which there is no showtime data from last week to compare with are shown in grey.
I’m not ready to walk away from my picks this week based on this data, especially given the huge assumptions baked into them, but this is worth watching in the weeks to come. If “Criminal” makes Best Performer it will signal this line of reasoning may have merit. This more detailed showtime data gives us a window we didn’t have before and could give us more perceptive indicators, albeit ones that aren’t available until Thursday way past my picks article submission deadline.
Is this a decent idea? Am I missing something? Are there ways to improve this?
Comment here or on the original Thursday thread and I’ll be watching.
UPDATE 4/29 7a Pacific
In the Chatter thread that sparked this post, Austin Buyer’s Club pointed out that my interpretation of the final numbers is backwards. You want that $/Showtime number to be low to indicate a movie having a better chance of meeting its forecast and if you take the Week 8 actual $/Showtime, subtract the Week 9 target $/Showtime and then divide the whole thing by the Week 8 actual $/Showtime, the lower the number the better. So take the color coding and reverse it. Sorry for any confusion.