Tinder and Matching Algorithms

Have you ever wondered how Tinder and its matching algorithm works? As one of the most popular online dating platforms, Tinder has significantly changed the way people meet and start relationships. I was curious about how its algorithm shapes our future, so I decided to dig into it. In this article, I will briefly share what I discovered about “matching algorithms” and their inner workings.

But first, let’s look at some statistics. According to Pew Research, 30% of US adults have used a dating site or app including 9% doing so in the past year. This number increases to 53% for individuals under the age of 30 with Tinder being the most popular among them. Furthermore, 10% of married couples met via dating sites or apps.

However, 50% of users report somewhat negative experiences. Women who have used online dating are more likely to feel overwhelmed (54%) by the number of messages they receive while men are more likely to feel insecure about a lack of messages (64%).

Given that online dating usage is increasing every year, with 53% of individuals under 30 using it and 10% of total marriages starting this way, dating algorithms now have a huge impact on our daily lives.

I suggest we do the simple thing here and blame the algorithms for not being good enough. Of course, there are two other factors that impact algorithm choice as well:

1. The male to female ratio is around 3 to 1, which is no different from my high school and engineering school.
2. Women and men wanting different things from dating sites and apps could also have an impact here, but let’s ignore it for now. Check out the percentage of dating users who want different things from dating sites and apps below, especially in terms of wanting casual sex.

But how does Tinder work?

Tinder is a dating platform that allows users to create a personalized profile with photos, a bio, and their interests. The platform uses a swiping system where users swipe right to show interest in a potential match and swipe left for disinterest. If both users swipe right on each other, the app matches them, and they can start chatting within the app.

Right for like, Left for nope

It’s super straightforward. But given the sheer number of users, how does Tinder’s algorithm decide to show users to each other in the first place?

Tinder previously used an Elo score for each profile, but the platform moved away from it in 2019. Here is the public announcement. In the Elo algorithm, Tinder assigns a numerical value that represents how desirable or popular a user is to other users based on the number of right swipes they receive. So if you are more “desirable,” meaning that you have a higher score, you are more likely to be shown to other high-scoring users, increasing your chances of getting a match. On the other hand, if you have a lower Elo score, you may be shown to low-scoring users, which makes it difficult for you to get matches. Again, the rich get richer.

If you’re interested in learning more about the Elo score, which was initially developed for chess in the 1960s, I suggest you read another Medium article I wrote that explains Elo in detail. Enjoy it!

Now let’s continue with the current algorithm. Unfortunately, Tinder is very private about how its current algorithm works. However, there is nothing stopping us from making educated guesses.

For sure, the app now uses a more complex algorithm to figure out which profiles are show to each user given their public communications.

1. Tinder Activity
   Being active is crucial. Tinder matches active users with active users instead of solely focusing on some arbitrary score. Again, this is something I shared in details with Elo score vs. Glicko score in my Medium article. Imagine a scenario where a user has the absolute average score on the platform but he/she was inactive for a while. Maybe they found someone who they enjoyed spending time with it. There is no point in showing this profile to active users as it increases frustration. Matching might be the first step but olsa having meaningful conversations — hopefully in real life eventually- is much more important.
2. Distance, Age and Gender
   These are way to obvious factors algorithm takes into account but here we go. You want to meet someone who you can conveniently meet in real life based on your sexual orientation and desired age span. Tinder provides functionality to limit your potential matches.
3. Interests
   Tinder is an endless journey that offers flexibility to its users. For individuals who want to reveal more about themselves, Tinder considers the hobbies and lifestyle descriptions that members include in their profiles. For instance, if someone is passionate about hiking or desires to find someone who loves dogs, they can narrow their search to those with similar interests.
4. Photos
   Tinder utilizes anonymous cues from photos to customize suggestions beyond the information that users provide. The platform recommends profiles that have similar pictures to those that users have previously shown interest in, and displays their profiles to more individuals who have also liked members with similar photos. For instance, if users have shown interest in individuals who enjoy outdoor activities, attend festivals, or spend time at the beach, Tinder will take note and suggest similar profiles. Maybe this could be an article of its own I can work later.
5. Liked or Noped
   Likes and Nopes are valuable indicators of members’ preferences. Tinder uses this information to refine potential matches by analyzing how frequently profiles, including those in the same geographical area, receive likes or nopes.

What is excluded?

Religion and ethnicity exclusion of religion and ethnicity on user preferences is likely driven by a desire to create a platform with a more inclusive and equitable environment. Allowing users to filter potential candidates matches based on their religion or ethnicity might inadvertently promote discrimination and prejudice. Excluding these could also help to reduce bias in the matching process. Nice move Tinder for a progressive approach. Studies have shown that people sometimes have unconscious biases when it comes to selecting potential partners based on various factors such as religion. By removing the filtering option, Tinder creates a more level playing field and gives more people the chance to find a match based on their personality and interests. And of course, based on their appeal, but this is another topic.

By positioning Tinder as a platform that promotes diversity and inclusivity, management may be able to attract users who may have been hesitant to use the app due to concerns around social biases.”

Best Guess: Gale-Shapely Algorithm

Given that Tinder is secretive about its internal algorithm, the most likely option is something close to Gale-Shapely. Here’s why I think this is the case: Tinder’s parent company Match acquired Hinge in 2019, around the same time that Tinder announced its move away from the Elo algorithm. In 2018, Hinge rolled out a feature using Gale-Shapely to recommend one potential candidate per day for the best pairing.

Gale-Shapely is an algorithm that solves the stable marriage problem where an equal number of men and women each have preferences for their potential matches. The objective is to match each user with a partner such that no two people would prefer anyone else over their current partners. Shapley and Roth were awarded the 2012 Nobel Memorial Prize in Economic Sciences for “the theory of stable allocations and the practice of market design,” but Gale passed away in 2008 before seeing his work recognized at the highest level.

The algorithm starts by having each man propose to his most preferred woman. The women choose between their candidates, accepting the proposal of the man they prefer the most and rejecting all others. The rejected men then propose to their next most preferred woman who has not already rejected them, and the process continues until every man is either matched with a woman or has been rejected by all women.

This is where it gets interesting. If a woman receives a proposal from a man while she is already “engaged,” she evaluates alternative proposals to her current one and chooses the one she prefers more. Then, the rejected partner proposes to their next most preferred woman, and the process continues.

However, stability is based on an equal number of men and women, which we know is not the case.

Here is an example how it would work when there are more men than women.

Consider a scenaria where there are 4 men and 3 women.

Men’s preferences:

M1: W1 > W2 > W3 (Men 1 prefers W1 most, then W2 then W3)
M2: W2 > W1 > W3
M3: W1 > W2 > W3
M4: W2 > W3 > W1

Women’s preferences:
W1: M4 > M1 > M3 > M2
W2: M1 > M2 > M3 > M4
W3: M1 > M2 > M3 > M4

Algorith initiates with all men proposing their most prefferd woman. In this scenario;

M1 proposes to W1,
M2 to W2,
M3 to W1,
M4 to W2.

Next, the women accept the proposals of their most preferred suitors so W1 accepts M1,
W2 accepts M2,
W3 has no proposal (yet).

Sorry for M3 and M4 since they are rejected and left without a partner and W3 has no proposal yet.

Now, rejected M3 and M4 go for their next preferred candidates.

M3 proposes to W2,
M4 to W3.

M3 is rejected again because W2 is already engaged to M2 and W2 prefers M2 over M3.
W3 accepts M4 proposals.

Next round,
M3 proposes to W3.

Since W3 prefers M3 over M4, she changes her decision to M3.

Final state is (M1, W1), (M2, W2), (M3, W3) and good luck for M4. Given that there is no preference over their current partners the solution is stable.

Now all women is engaged but in the next round M4 proposes to W1 as M4’s second preference was W1. However, W1 is already engaged to M3 and W1 preferes M3 more than M4 so M4 is again single. Then, M4 goes to his next most prefered candidate which is W2.

I believe that moving from the Elo score to maximizing matching is better on many levels.

1. The success of Tinder is not about ensuring that the “desirables” get more matches. Love is not a competition; it is collaboration. Using the Elo score does not ensure compatibility, but rather it ensures that there are “winners” and “losers,” as seen in other areas where Elo is or was used, such as in chess and Dota 2. The psychological impact on users is not insignificant, and I am sure we all have a “friend” who has suffered from this problem.
2. Like all businesses, Tinder wants satisfied and happy users. Their algorithm was not perfect, and it is not perfect now. But staying competitive requires constant work on their matching algorithms. Ultimately, this is what Tinder aims to provide.