Top Chef is back! My partner and I are huge fans and so excited to watch incredibly talented chefs make amazing food. It’s also a chance for us to guess who the winners and losers are before the judges’ kick someone out.
During one episode of season seventeen, right as the knives sliced together announcing Judges’ Table, my partner turned to me and said, “oh no! Lee Anne is a goner!”
She was one of our favorites. While the judges hadn’t cared for her mixed berry clafoutis and cinnamon donuts, she had won the Quickfire Challenge. Surely she was safe!
But no. Padme said those dreaded words. “Lee Anne,…”
How did my partner know? What was this dark magic? She explained to me that whoever appears on screen the most is sure to be eliminated.
Her theory was that in a last moment of pity, the producers go into the editing room and throw in as many interviews of the loser as possible. This gives that chef one more, longer period of time on screen before they are never seen again (or until they come back as a guest judge or all-star).
I, being a natural skeptic (and such a joy at parties), decided I needed to investigate. If this was true, I would impress all of my friends with my crystal ball abilities! I had the machine learning to figure it out.
Process: count the faces
I leveraged an OpenCV tutorial to recognize faces. Then it was a matter of modifying the code to go frame by frame and count the number of scenes in which each contestant appeared prior to Judges’ Table.
Oddly, I tried to parallelize the process by splitting each episode into quarters and halves and running the recognition on my GPU, but my GPU couldn’t handle more than one frame at a time with memory constraints.
Also, running 1 GPU-optimized OpenCV run was slower than splitting the video into eights and running it on my 8-core CPU. That’s a topic I may dig into later.
Check out my hacky python code here.
Does it work? Not quite.
I know what you’re thinking – Betteridge’s law of headlines. However, I’ll break down the numbers for those willing to read on.
The eliminated contestant appeared the most on-screen three out of thirteen episodes (not counting the finale) or 23%. However, if you guessed one of the three most frequently shown contestants, you’d be right seven out of thirteen episodes or 54%. And if you guessed someone in the top five most shown would be eliminated, you’d be right 69% of the time.
Of course, as contestants get eliminated, guessing someone in the top five isn’t very impressive. Here’s the breakdown.
The rank by the number of scenes of the eliminated and winning contestants, where 1 is the most scenes in an episode, are below.
| Episode | Rank of appearances of eliminated contestants | Rank of appearances of winning contestants |
| 1 | 2 | 14 |
| 2 | 7 | 4 |
| 3 | 10 | 6 |
| 4 | 6 | 5 |
| 5 | 2 | 11 |
| 6 | 3,7 | 4,9 |
| 7 | 6 | 9,8 |
| 8 | 1 | 1,5,7,8 |
| 9 | 1 | 4 |
| 10 | 3 | 1 |
| 11 | 4 | 2 |
| 12 | 5 | 2 |
| 13 | 1 | 2 |
Who appeared the most?
You may be dying to know who did make the most appearances. Was it fan favorite and winner Melissa? Nope. While Melissa won the hearts of the audience, it was Kevin who was the twinkle in the producers’ eyes.
Despite missing 2 episodes and not making it to the finale (except as a sous chef), he managed to appear in the most scenes total across all episodes and appeared the most in the most individual episodes.
He was the fan favorite in season six and made it back from Last Chance Kitchen. Maybe the producers liked him in that season or maybe they liked a come back kid story.
Who was popular in the best and worst episodes?
The last thing I investigated was IMDB rankings. At the time of this writing, the highest ranked episode was episode thirteen with 7.7 stars and the lowest ranked episode, despite its amazing Olympian guest judges, was episode ten with 6.6 stars.
For episode thirteen, producer favorite Kevin was eliminated and fan favorite Melissa won. Kevin also appeared the most on camera and Melissa the second most in that episode. Were fans excited to see more of Melissa? Were they happy to see Kevin eliminated because he hogged the spotlight? Your guess is as good as mine.
The results of episode ten were Karen was eliminated and Stephanie won. Stephanie also appeared the most on camera, Melissa the second most, and Karen was the third most. My guess is it had nothing to do with who won or was on camera.
It doesn’t work, but it was worth it to try
So, you can’t guess who will be eliminated from Top Chef based on the number of times they appear on camera. However, it was fun to give it a shot and to use machine learning and facial recognition to try to find out.
Thanks for reading!