MTV’s reality dating show Are You The One combines 10 guys and 10 gals whose “scientifically determined” hetero love-match sits across from them — without either party of the matches knowing the identity of their match. Each week the contestants pair up at the end of the episode and the total correct number of matches is dramatically displayed in number of bright beams of light — yet the particular matches, the who-is-matched-with-who-correctly — is left for the contestants to figure out for themselves. When my girlfriend showed me the trailer for this series, she quickly added that I would in fact be expected to not watch with her, likely for concerns that some people (me) would bundle this show into the general category of Trash Television. I, however, saw a great opportunity to use some completely over-kill data science tools to tackle the results of this fine television series. The season finale is TONIGHT, and these results are hot off the compiler, so here you go:
The (spoiler alert) entirety of the code and results are posted as open source code on my github page. The analysis uses the very popular (for data scientists) Jupyter Notebook to hardcode in the results from each week (credit to Stephanie N. for diligent data collection), then takes a rather hacky and cumbersome — but effective — brute force approach by simulating 100,000 possible pairings (with replacement), and tests the error for each weeks worth of data. In cases where the simulated pairings are consistent with each episode’s beams of light, the error counter is zero. Cases where the total error is zero are printed as predicted pairings. You’ll notice that three predictions are made, all redundant.
I worked on this hack while at NeuroSummit2016 and presented it to a captive audience, to whom I am grateful for their attention to my talk amidst other talks about saving lives and high art. I was especially and actually thankful that my talk was preceded by a thorough unpacking of a Britney Spears track, so at least we covered a big range. Note that a simple googling will show you that many other people cracked the above pairings days ago (before me), but I’m pretty sure this is the first open source Python code framework that solves the show.
Shouldnt It Be 3 million pairings? There are 10! Ways of pairings 10 guys and 10 Girls.
Good point! I didn’t point out that I eliminated one of the couples since they paired early on and since there were no cross-pairings, they didn’t provide any useful data. I also provided some exclusionary criteria based on a confirmed match in episode 4. So virtually it is 8! and the total pair numbers is 40k, though I’ll need to check that the exclusions are made outside of my 100k loop, in which case the number of possibilities would be 360k and your point would be valid. I’ll of course check that after the show (on now). Thanks for the input — this is why open sourcing is great 🙂