What makes horror films interesting isn’t wondering if people die, but when. And it turns out that there are patterns of mortality timing that can tell us a lot about the genre as a whole. We churned through 40 Netflix gore-fests in an effort to tease out when characters actually die. The results were fascinating so let’s talk about. it.

This feels like a post with spoilers.

There will be a few popular horror-flick spoilers.

What’s the big takeaway?

If you are the first character to die in a movie, on average you don’t make it past the first fifth of the movie. On average, 19.5 percent of a movie’s running time elapses until the first death (give or take a standard deviation of 22.9 percent). The median percentage to first kill was 8.3 percent; given an average running time of 1 hour and 39 minutes, expect the first person to kick the bucket at about the 8 minute mark.

Here’s what the data looks like, tracking time to first kill as a percentage of total running time, plotted against when the movie was released:

It sort of looks like the time to first kill is increasing over time?

That’s a tempting conclusion, but it’s too hasty. Remember when we said this was pseudoscientific? A big problem here is sampling bias — Netflix’s algorithm (and library) emphasizes movies made in the past few decades rather than those before the ‘70s. So yes, it might look that people are living for longer stretches of screen time since the The Cabinet of Dr. Caligari kicked us off in 1920, but it’s just that there are more movies, and hence, more variability.

OK, so what’s the deal with the 100 percent in 2014?

You wanna talk outliers? Let’s talk outliers. In 2014, we had the masterful Babadook, which sucker-punched horror’s killer rule in the gut: Nobody dies except for a dog. (Non-humans were excluded for the purposes of this analysis.) And there’s also Honeymoon, which has little in the way of death till the very end.

I will posit that there’s been a bit of a surge in the slow-burn horror, perhaps as a response to Saw-type torture porn, which doesn’t rely on huge body counts to be effective. If you have fewer overall body bags, then that kill, when it comes, is much more powerful later in the film. On the other hand, films with a certain genre-awareness, like Tucker and Dale vs. Evil or Grabbers, have kills super early on — 1.1 and 3.2 percent of the way in, respectively, quickly establishing that yes, we are in a horror movie, and people will die.

Also, “they were dead the whole time!” movies sort of complicate things (here’s looking at you, The Others). So we just went with a 0 percent for those because corpses aren’t casualties.

You watched a lot of people die over the past 24 hours. Are you OK?

Thanks for asking. I’ve watched more of The Human Centipede than I ever wanted, which meant fast-forwarding to the 24:50 mark for a lethal injection. My favorite first kill — or at least the one that stands out amid the cavalcade of gore, dismemberment, and impaling — came via Scandinavia, thanks to Dead Snow: Red vs. Dead. A trucker attempts to give what he thinks is an unconscious pedestrian mouth-to-mouth, and turns out it’s a Nazi zombie. Good stuff!

Is there any way we can get more analytical about this?

I thought you’d never ask. The answer is: absolutely. I’ve dumped my raw data in a public Google doc. Consider it before deciding what fresh horror you want to watch this weekend.

Photos via Giphy