> ## Documentation Index
> Fetch the complete documentation index at: https://docs.forum.market/llms.txt
> Use this file to discover all available pages before exploring further.

# Basket Indices

> Some indices are derived from a basket of other indices, such as the most-watched horror films right now.

## Purpose

A basket index tracks attention to a whole theme rather than one keyword. Interest in horror moves from film to film, so a horror index that only searched for "horror" would miss most of it. A basket follows whichever films are drawing attention now.

A basket index has two parts:

| Part | Weight | What it tracks |
| :- | :- | :- |
| Basket | 65% | The films currently in the basket, each through its own index |
| Keyword | 35% | The theme's keyword (e.g. "horror"), as in the [general methodology](https://docs.forum.market/guide/mechanics.md) |

## Membership

Each day from 07:00 UTC the basket is rebalanced:

1. Take every film that Wikidata files under the basket's genre (including its subgenres) and has an English Wikipedia article.
2. Rank them by average daily English Wikipedia views over the previous week.
3. Keep the top 10, leaving out films we have reviewed and excluded (e.g. comedies tagged with a horror subgenre).

Each member has its own index, listed on the [Index Explorer](https://docs.forum.market/guide/index-explorer.md), which searches for the film's title and its Google Trends topic. A member counts only once its index has published in the past 24 hours, so a new entrant joins with its first data point and a film that leaves stops counting straight away.

## Calculation

For each source metric (e.g. YouTube video count or Google Trends interest), the basket takes the **plain mean of the members' raw values**, before any transformation. It then transforms and weights that mean exactly as a single index would: the $\ln(x+1)$ transformation, the 15-minute normalization and the metric weights from the [general methodology](https://docs.forum.market/guide/mechanics.md).

Means are taken on raw counts because the members' index values are already log-transformed: averaging them would understate a single film breaking out.

**Example**

Suppose three members report YouTube video counts over the past 24 hours of `120`, `40` and `20`.

1. Mean: `(120 + 40 + 20) / 3 = 60`
2. The basket's YouTube video count is `60`, treated from here on like any index's YouTube video count.

The mean is uncapped: a single film breaking out moves the basket.

The basket and keyword parts are then combined with their weights into the Raw Index, and smoothed into the final index as in the [general methodology](https://docs.forum.market/guide/mechanics.md#final-index-ema).

## Current basket indices

| Index | Basket |
| :- | :- |
| `HORROR-IDX` | Top 10 horror films |


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.