On January 28th, we uploaded every photo we had of ourselves to the internet to mathematically make the perfect dating app profiles. 230,000 votes later, our photos have filtered and guided us to the definive answer to what makes for the best dating app photos. Notably, there’s one type of photo that you must never include. We’ll save that for last.

To those who don’t trust this analysis, we would argue that polling people online is ecologically comparable to a dating app and that a quarter million pairwise votes ensures statistical robustness. To those who do trust this analysis, we ask: seriously? This report is tongue-in-cheek, but we have found several fascinating takeaways and have updated our dating profiles to our top performing photos. (Wish us luck).

We use a softmax-weighted Elo system inspired by chess, averaged at 1,500. You can check it out here. A match consists of a photo facing off against another photo, with the user judging which is better. Our random algorithm shows photos with stronger Elos at a far higher rate, providing more feedback on the photos in real contention to make it onto our Hinge profiles.

At scale, Elo scores can help predict the probability that one picture will “win” over another picture, even if they haven’t actually faced off.

As our goal here is to strengthen our dating profiles, we’ll filter for results from female voters unless specified otherwise. Voting behaviors between men and women track relatively similarly but do diverge in key categories which we will discuss.

We also introduce an Ick factor. Five women have gone through all of Josh’s photos and manually selected photos that give them “the ick”, giving us a 0-5 Ick score which can help us ground what these Elo ratings in an emotion. A big thank you to Maia, Maura, Sarah, Mia, and Daphne.

Lastly, while we recognize that these rating mechanisms are highly subjective, we do believe that there is truth to the methodology. A few exact duplicates accidentally ended up in the dataset, but maintained shockingly similar ratings over the course of thousands of votes.

Now onto the findings.

While celebrities always seem to flex their photogenic smolder, smiling photos do significantly better than their non-smiling counterparts. The median smiling photo (1499 Elo) beats the median non-smiling photo (1472 Elo) 53.9% of the time. Drilling down, we find more expression is favored: open smiles (1503 Elo) beat closed smiles (1486 Elo) 52.4% of the time.

The first piece of advice a fresh bachelor receives on dating photo selection is “no women.” We find this folk wisdom holds true. Among female voters, photos with one woman (Elo 1458) lose to photos with one man (Elo 1480) 53.2% of the time. Perhaps unsurprisingly, this is higher than the outcome of male votes, which sits at a 51.4% loss rate.

The one thing Jonathan loves more than DSLR photography is being photographed while on his DSLR. These candids are more dynamic than the baseline photo and sneak a peek into Jonathan’s favorite hobby. We found photos featuring his camera (1568 Elo) reign over non-camera photos (1472 Elo), winning 63.5% of the time. Our takeaway is that you should show off your skills and interests.

Not everything’s better together. While solo shots (1491 Elo) sit just below the Elo baseline, each additional person runs down the rating, as displayed in the graph below. A key point to make here is voters only deliberated on individual photos, not full profiles where a group photo might thrive in third or fourth position.

Let’s discuss the 2010s-era elephant in the room: selfies. Traditional selfies (1484 Elo) lost to non-selfies (1503 Elo) 52.7% of the time. But the biggest loser is a third contender: the mirror selfie, which sits at a horrendous 1452 Elo, losing to non-selfies 57.3% of the time.

This disadvantage is further highlighted by the Ick Factor. The average non-selfie scores 0.42, traditional selfies double that at 0.88. And mirror selfies leap to 1.38. Incredibly, 78 out of 112 mirror selfies induced an ick from at least one of the five rankers

Furthermore, the lowest ranked photo in Josh’s entire photo roll was a mirror selfie (1343). Shield your eyes.

In retrospect, our biggest findings (smile genuinely, no women, showcase hobbies, and avoid mirror selfies) may seem intuitive, but finally having empirical evidence adds significant credibility to these assertions. And more importantly, these findings make life a tad better for us: Josh has always been insecure about his natural smile, but the empirics say it’s a winner. And since updating his profile two days ago to match his top photos, Jonathan is scheduling his first second-date in a year.

To conclude, flying solo on Valentines is tough. But we’re all in this together. We sincerely hope you can leverage this data to upgrade your own profile and feel a bit more hopeful for the year ahead. Go buy yourself flowers!