Indexer results uniqueness score

The score tries to help answer the question: How good are my indexers and which ones do I really need?

Contrary to other (older) stats this score doesn't care about any search results that are not downloaded because, after all, you only care about the stuff you download, not all the results returned by an indexer but never used. Every time a result is downloaded (either by the GUI or via the external API) the number of indexers that were queried in the related search and the number of indexers which returned results with the same title are saved. We assume that any results with the same title are identical and that the same download could've been retrieved from any of those other indexers.

The score is then calculated via this simple formula: 100 * [Number of indexers involved in search] / [Number of indexers returning this result].

Examples:

Indexers involved	Indexers that returned the same result	Score for those indexers
10	1	1000
5	1	500
10	5	200
1	1	100
10	9	111

As you can see the first example produces a very high score because in this case the indexer from which the download was made was the only one out of 10 indexers which contained the result. Without this indexer the download could not have been made. In the last example the indexer gets only a low score because the downloaded result was also found on 9 other indexers.

These scores are then multiplied with the number of scores per indexer to make sure that indexers that are often used and would often return results are scored better than those which provide high scores but rarely.

If you have any ideas how to improve this calculation feel free to contact me.