Computation Methods Explained

Harmony offers three ways to rank your tracks, artists, and albums. Each method serves a different purpose depending on the insights you’re after. The table below gives a high-level comparison before we dive deeper.

Method	Data Source	Accuracy	Time Range Support	List Length
Machine-Learning (Spotify)	Spotify Top API	⭐️	3 presets (4w, 6m, 1y)	≤ 50
Play Count	Streaming-history package	⭐️⭐️	Any (custom)	Unlimited
Time Played	Streaming-history package	⭐️⭐️	Any (custom)	Unlimited
Custom (SOON)	Streaming-history package	⭐️⭐️⭐️	Any (custom)	Unlimited

Tip: Uploading your Extended streaming history unlocks the Play Count and Time Played methods.

1. Machine-Learning (Spotify API)

Spotify’s Web API returns a list it calls “Top Artists/Tracks” which is generated by proprietary models. The exact recipe is not public, but factors likely include:

Total play time per item
Skip rate
Recency (when you listened)
Historical presence in previous lists

Formula (conceptual)

\text{score} = w_1 \cdot \text{play\_time} + w_2 \cdot (1 - \text{skip\_rate}) + w_3 \cdot \text{recency} + w_4 \cdot \text{streak}

where the w coefficients are internal to Spotify.

Pros & Cons

✔ Advantages	✖ Limitations
Works out-of-the-box	Only 3 fixed ranges (4 weeks, 6 months, 1 year)
No data import required	Capped at 50 items
-	Inaccurate for niche content

2. Play Count

This method orders items purely by the number of times you played them.

Formula

\text{score} = \text{total\_plays}

If you filter by date range, we count only plays whose timestamp lies within the window.

Example

Imagine two tracks over the last year:

Track	Plays
Song A	120
Song B	95

Play-count ranking → Song A #1, Song B #2.

Pros & Cons

✔	✖
Extremely accurate	Requires package import
Custom time ranges	-
Faster than API (local DB query)	-

3. Time Played

Here we rank by milliseconds listened rather than count. Longer tracks (podcasts, DJ mixes) get fair representation.

Formula

\text{score} = \sum \text{listen\_duration\_ms}

Example

Track	Plays	Avg. Length	Time Played
Symphony (10 min)	20	600 000 ms	12 000 000 ms
Pop Hit (3 min)	40	180 000 ms	7 200 000 ms

Despite fewer plays, Symphony outranks Pop Hit because total listening time is higher.

Pros & Cons

✔	✖
Accurate & bias-free	Requires package import
Rewards long-form content	-
Custom time ranges	-

Choosing a Method

Casual glance → Machine-Learning is fine.
Precise stats & deep dives → Play Count or Time Played.
Long DJ sets, podcasts → Time Played gives the fairest ranking.

You can switch methods from the Sort dropdown on the settings page (SOON).

Now you know exactly how those numbers are produced, happy analysing!

Computation Methods

On this page