How the Fad Index is built

The Fad Index is not a price guide or a forecast. It is a repeatable way to rank attention around toys and trading cards using countable evidence. The build starts with a defined universe, a fixed collection window, and rules that can be rerun. No single article sets the result. The index records where mentions cluster and how that cluster changes between builds.

The method below describes the mechanism. It does not list a current value because no measured figures are supplied in the data.

Set the universe and the window

Before any counting, define what belongs in the index. Use stable identifiers for each item: product line, release name, card set, character, or accessory category. Record common aliases so the same item is not split across spellings. Choose the collection window and category list in advance. The window is the only period that contributes to that build. Items outside it are not counted. If the census supplies a category count, use it as the denominator for that category, not as a score. This keeps large categories from dominating only because they contain more listings.

Count evidence once per article

The article library supplies mentions. For each item, count an article once even if the item appears several times in it. Separate a passing mention from a focused article by tagging the article type before counting; do not change tags after seeing results. If census data adds inventory counts, keep those counts in a separate field. Do not mix article mentions and census counts into one raw number. The index needs an audit trail: source, date collected, item identifier, and signal type. Missing data stays missing; it is not replaced with a zero unless the method defines zero as "checked and absent."

Normalize, weight, and publish

Raw counts are converted to shares within each category. Set weights in advance for the signals you use, such as mention count, recency within the window, and source diversity. Publish the weights and the normalization rule so a reader can reproduce the ranking. Show the rank, the raw supporting counts, and the category share. Do not present a single score without its inputs. If a figure is not supplied in the source data, leave it out rather than estimating it. The result is a comparative index, not a valuation. It tells you where attention is concentrated in the collected material and how that concentration changes between builds.

Is the Fad Index a price guide?

No. It records countable attention signals from the article library and census. It does not set or predict prices.

Does a high rank mean an item is a good purchase?

No. A rank shows concentration in the collected sources. Buying decisions need condition, budget, and market checks outside the index.

How often is it rebuilt?

Use the fixed schedule published with the method. This article does not state a schedule because no schedule is supplied in the data.

Can I rebuild it myself?

Yes, if you keep the universe, window, aliases, tags, weights, and normalization rule fixed. Changing those after counting breaks comparison.