How We Calculate Your Face Preferences
The quiz compares the portraits you pick with the portraits you actually see. That gives each style context, especially when one style appears more often than another.
Take the free quizFree to start. Sign in to open your result.
- After Dark: 0% of picks, 23% of portraits shown.
- Art House: 23% of picks, 30% of portraits shown.
- Golden Hour: 77% of picks, 29% of portraits shown.
- Soft Focus: 0% of picks, 19% of portraits shown.
What the diagonal line means
Read across to see how much of the gallery a style accounted for. Read up to see how much of your selection it accounted for. The same simulated run supplies both the chart above and the table below.
Golden Hour accounts for 10 of 13 picks (76.9%) and 73 of 253 shown (28.9%). It sits above the diagonal.
Art House appears more often in this run, but accounts for only 3 picks.
| Style | Picked / shown | Share of picks | Share shown |
|---|---|---|---|
| Golden Hour | 10 / 73 | 76.9% | 28.9% |
| Art House | 3 / 75 | 23.1% | 29.6% |
| Soft Focus | 0 / 47 | 0.0% | 18.6% |
| After Dark | 0 / 58 | 0.0% | 22.9% |
How we choose your main style
For each style, we start with the fraction of shown portraits that you picked. To keep a style with only a few examples from dominating, we add the equivalent of four observations at your overall pick rate. As you see more portraits of that style, its own pick rate carries more weight.
For example, if you pick 25% overall and chose 3 out of 8 portraits of one style, the adjusted rate is (3 + 1) ÷ (8 + 4), or about 33%. The top adjusted rate guides the main style in your result.
What the detailed prediction score checks
When there are at least five picks and five passes, we hold out one of each, recalculate from the other portraits, and check whether the style model ranks the pick higher. Ties count as half. This score describes how well the styles explain the choices in that session.
If two held out portraits have the same style, the model cannot separate them using that label, so the pair contributes half a point. A score near 50% can mean the style labels offer little separation for your picks, even when you know exactly what you liked about an individual face.