Hi hi, Alex from Ultra Sound. We’ve been asked to share the state of an experimental change we’ve made to the relay. We call it “adjustments”. Let me set the context. Given:
- builder A bids against builder B, at relay X. builder A bids 19, builder B bids 21. relay X now has a 21 top bid.
- builder A bids against builder C, at relay Y. builder A bids 18, builder C bids 19, relay Y now has a 19 top bid.
Most relays are “simple”, and when the proposer asks, relay X would offer 21, Y would offer 19, and the proposer would pick the 21 bid. The ultra sound team realized the following. If you see the proposer asking the relays for their best bids as an auction, simply “max-bidding” as a relay is not always optimal. Relay X could, by slightly lowering its bid to 20, still win and keep a fraction of the bid for itself.
We’ve been running this experiment for some weeks now. Some points / questions that have come up:
- the impact so far is rather small. in feb ultra sound delivered 9,607 ETH in bids, and reduced bids by 55 ETH or 0.57%. March 6,266 delivered, 20 reduced or 0.32%. until last week, ultra sound has kept $0 for itself and returned 100% of superfluous bid value to builders.
- the ultra sound relay always provides the highest bid a proposer can possibly get, otherwise it simply loses the auction. as long as a proposer is connected to multiple competitive relays, this stays true.
- in practice the advantage one relay has over another is often latency based. it is possible one relay can no longer get a better bid to a proposer before the proposer’s deadline, whilst another can. the amount the bid appreciates in those milliseconds is value the relay adds and can capture. it would be impossible to receive the better bid if the relay wasn’t around or hadn’t invested in being fast.
- relays cost money to run
, we currently don’t charge proposers or builders. the more you charge proposer, and especially builders, the more attractive you make it for them to run their own. - if a builder runs their own relay, adjusting bids as outlined above becomes a no-brainer, and neutral relays which try to contribute to proposer-builder separation would go unused.
- relays have been in talks with various parties for months, including Lido! to see where relays could get funding. the ultra sound team is rather proud to say they may have found a way to fund themselves which asks neither builders or proposers to pay for their services and keep supporting Ethereum censorship resistance, the relay ecosystem, the ultra sound website, and maybe even based sequencing.
Happy to answer questions if anybody has them, you can also find me on TG as Telegram: Contact @smilingalex.