Hacker News new | past | comments | ask | show | jobs | submit login
Tinder Auto-Matcher Bot (github.com/verdie-g)
34 points by verdie-g on Nov 25, 2019 | hide | past | favorite | 16 comments



>Q: I didn't get any match even tough Tinder tells me that I >got X+ likes!?

>A: Several reasons could explain that: >[...] >They are not in your elo anymore [...]

Kinda creepy that you can like a person, that person likes you too, and yet the algorithm might override your choice and decide to not create the match.


> yet the algorithm might override your choice and decide to not create the match

I believe part of the motivation for this is to keep the experience good for straight women. If straight women leave the app, it's going to die.

Because straight women often have more potential matches than they can go through, Tinder is more aggressive about prioritizing those.

If they present a woman with a man whose score was high, he might be physically attractive enough for her, but he might have "behaved badly" in recent conversations and therefore be low-value match for her in the end. If she keeps getting matches with men who behave badly when chatting starts, she'll probably quit the app.

Among my straight female friends who use Tinder, harassment fatigue is the #1 reason they quit or take breaks.


What kind of elo do you have on tinder anyway?


It's a combination of how many people are swiping left/right on you and the ELO of said people.


It’s no or more less creepy than Facebook or Twitter deciding which of your friends’ posts you will read today (as well as which you will not), and pretty much every internet-connected person on Earth has settled for and accepted this, so I guess this is what we get now.


oz exists in our algorithms now


Hey, this is pretty cool! I made something similar but I was doing image comparison via cross-correlation template matching using OpenCV: https://github.com/jidicula/tinderizer.

A small correction to your README though:

> These pictures are in very low definition though.

There is a way to access the original (highest) resolution of the submitted images through the API. My code picks the lowest-res one though since I'm doing image comparisons.

However, looking at what the output is now, it looks like Tinder has tightened up their API since I wrote this. It used to be the case that each teaser profile that likes you would have all of its images accessible through the API. This is no longer the case, and only 1 image is accessible for each profile that likes you.


Dating sites definitely need some ML and AI thrown at them for the user, so we can make basic decision trees and save ourselves time and effort.

Is there an actual person in the photos? Is it obvious which person it is? Are there any that aren't close up selfies? Are there any close up selfies? Are they smiling? Any red flags like pictures of Machu Picchu or marathons or massive glasses of alcohol or teardrop tattoos under the left eye?

We can visually identify thousands of species of birds at this point, surely computer science can tackle this important problem next, no?


At the very least, dating sites should have all the data they need to figure out what a user's "type" might be and prioritize profiles matching that type.


This was the entire idea behind OkCupid. Strong analytics and lots of Q&A to match you with people based on "compatibility".

Turns out people don't give a shit about compatibility, they just want to date hot people. Figures.


"hot people" is what I was getting at by "type". Machine learning nowadays should be able to figure out "hey, this user likes people who look like this, so let's prioritize potential matches who look like this".

If you wanna get extra dystopian, link the profile to a PornHub account and figure out the user's "type" that way.


What people consider "hot" can be extremely different. For me any attraction to a woman completely disappears if doesn't have natural brunette hair, isn't interested in STEM, likes cats, has different political views from me, etc. I'm sure I'm not the only one with strict preferences. I've always felt like dating apps didn't put enough effort into capitalizing on that. Luckily I'm married and don't have to worry about that anymore.


People did. OKC didn’t.

OKC pivoted to volume and frequency of swipes over quality of matches, the superficial engagement metric that generally ruins content.


Well yes but that all happened after the Match Group acquisition a few years ago (right about the time they introduced "swiping"). Before that it was all algo and profiles... and OKC didn't have a rep for being any better than all the other dating sites.


I like the way this is written. Did you use anything to generate the models and JsonProperty annotations?


Vim macros mostly :D




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: