Zigzag Pattern
Ranking
Network
Retrieval Score
Caveats:
1. Filecoin retrievals are made across all FIL+ deals, not just those from Filecoin datacap allocators that demand retrievability. We will soon have the fidelity to remove certain allocator deals from the calculation which will likely improve the Filecoin score.
2. Walrus retrieval requests are currently made to the aggregator nodes, not to the root storage nodes. A better comparison here to Filecoin and Arweave would be to test the retrieval success rate of fetching the erasure encoding blobs on Walrus nodes. However the Walrus Network doesn't yet expose public interfaces for this.
3. As this leaderboard develops, we would like to make a distinction between L1 storage nodes and L2 nodes, such as aggregators, gateways, CDNs.
Zigzag Pattern
How are networks ranked?

Networks are ranked based on their retrieval score: the higher the score, the higher the position on the leaderboard.

How is the retrieval score determined for Filecoin?

The Filecoin retrieval score represents the average of all Storage Provider (SP) retrieval success rate (RSR) scores over a 24-hour period. Storage Provider RSR is calculated by dividing the number of successful retrievals by the total number of valid retrieval requests for that provider within a given timeframe. Only requests that pass fraud detection are included in the calculation. For more detail please refer to Spark protocol docs.

How is the retrieval score determined for Arweave?

The retrieval score for Arweave is determined by the percentage of successful retrievals over a 24-hour period. Retrievals are conducted by selecting a random Arweave transaction and performing a retrieval from a random Arweave node, and the results are reported back to the API. For more details, refer to the Arweave Checker repository.

How is the retrieval score determined for Walrus?

The retrieval score for Walrus is calculated based on the percentage of successful retrievals over a 24-hour period. Retrievals on Walrus are performed by selecting a random blob and fetching it from one of the public aggregator nodes. The result of a retrieval check is reported back to the API. Note that this approach may result in a higher retrieval score than Filecoin and Arweave as it is fetching the data from an aggregator rather than from the storage nodes. For more details, refer to the Walrus Checker repository.