Skip to main content

Weigh attestation correctness in Raver Score

Feature RequestMethodologies

The attestation correctness score is currently calculated by taking the average of the source, target and head score. These scores already include lateness metrics.
To calculate the attestation effectiveness this average correctness is multiplied with the participation rate and divided by the inclusion delay:


Correct target votes receive double the rewards of head or source votes. I propose reflecting this in the attester effectiveness rating by changing the formula to:

This would better reflect the total attestation effectiveness as it places the weight intended by Ethereum on the specific parts of the attestation.

Posted by Florian Winkler ·
2 comments

Log in to comment and vote

Comments2

  • e

    •

    May 28, 2025

    thank you for the proposal @Florian Winkler. in the spirit of extending the discussion, we have chosen to maintain the formula in its present form to respect the “permanence” and “autonomy” principles we started out with in the first place (

    https://docs.rated.network/documentation/methodologies/introduction-to-methodologies).

    we designed the RAVER in Phase 0, with a view of having it be “blunt” enough to withstand the test of time without needing surgery with every upgrade and rule change, but “specific” enough to be able to tell a story about an operator’s useful life across time and across upgrade eras.

    drawing the exact lines is more art than science. so for example, “too blunt” would be not being able to tell a performant operation from a less performant one. “too specific” would be recreating the rewards spec in a different format (it doesn’t offer any new information).

    specifically to the point you’re raising, we have historically found that the correlation of the 3 metrics is high enough to not warrant any material distinction—hence we are where we are today.

    we’re due for a re-run, we’ll get back to you here with findings.

  • Robert Tiong

    •

    May 26, 2025

    Thank you for the proposal, Florian. Definitely something worth looking into. Will have the team review give feedback here.

    Discussion is also open to everyone else reading this and wants to chime in.