Quiz System: Generating the 4 Choices

When a Host uses the 4-choice quiz tool, Play Guess and Reveal doesn't just send the correct answer to the players. The app dynamically generates three additional options (the "distractors") to create a relevant and deceptive multiple-choice (4 answers) question.

This document explains how the app manages to find consistent wrong answers by relying on Apple Music data and a similarity scoring system.

How the Quiz Engine works overall

The heart of this feature relies on a "Quiz Engine" built into the app. Its goal is to build a balanced challenge: neither too obvious nor impossible.

1. Extracting the target: The app first identifies the correct answer (the title, artist, album, or year of the current track).
2. Searching for candidates: It then queries the Apple Music catalog to retrieve a large list of potential candidates (often around forty).
3. Filtering and cleanup: Candidates that are too similar to the target (for example, remixed versions of the same song or artists featured on the same track) are excluded.
4. Scoring: Each candidate receives a "plausibility" score based on similarity criteria.
5. Weighted selection: Finally, the algorithm randomly draws 3 candidates from among the best scores to ensure some variety if the same question is asked twice.

Generation logic by category

The method for finding the 3 wrong answers varies greatly depending on the type of information requested.

Title quiz

To offer credible fake titles, the algorithm favors:

• Other tracks by the same artist: Players might recognize the voice or style but pick the wrong song.
• Tracks from the same musical genre: The app looks for other popular titles within the current song's main genre.

Artist quiz

To throw players off about the performer, the quiz engine looks for:

• Artists from the same genre: Similarity of style is the primary criterion.
• The algorithm also uses tags related to the artist's perceived characteristics (whether it's a band or a solo artist) to offer alternatives of equivalent format when possible.

Album quiz

Generating fake album names requires staying consistent with the artist's world:

• The app first looks for other albums by the same artist.
• It rounds things out with albums from the same musical genre released by other similar artists.
• The engine makes sure that no album name gives the answer away too obviously (by removing "Deluxe" editions or standalone singles).

Year quiz

The logic for the year is completely different because it doesn't require the network. The app uses a mathematical method:

• It identifies the track's actual release year.
• It generates a list of neighboring years (between 1 and 15 years apart, before or after).
• The similarity score depends on proximity: The closer a generated year is to the actual date, the higher its score. As a result, the options are generally clustered very tightly around the correct decade, forcing players to be precise.

Intelligence and Variability

One of the strengths of this system lies in its "weighted" aspect. If the app always took the 3 candidates with the highest absolute similarity score, the options would always be identical for a given track.

Conversely, by randomly selecting the 3 options from among the top 15 or 20 suggestions, Play Guess and Reveal ensures that each game stays unpredictable.