Starting on June 30., we are going to use a new metric for computing consistency rewards. The new metric will include quality of bids and settlement success rates. It replaces the current metric which just counts bids. This change of the consistency metric is done under the mandate from CIP-85.
Goals of consistency rewards
The goal of consistency rewards is to encourage active participation of solvers in all auctions. We have seen in the past that performance rewards alone cannot always provide enough of an incentive to bid, for example on small orders. Active participation is not a goal in its own right. We want more solvers to participate to be able to provide a robust service which reliably gives great prices and results in fast executions. The current metric used for consistency rewards does not sufficiently encourage good and fast executions.
The current metric
The current metric just counts which solvers bid on executed orders. In bidding data of the last months we do see an increase in average number of bids. This is especially visible for orders with little volume.
This increase in bids correlates with a decrease in average bid quality, . More bids do not directly translate into more good bids.
There is also an obvious way to manipulate this metric: a solver can submit bids with almost zero surplus which are not intended for execution. Such bids can increase a solver’s consistency reward while not providing any value to users or the protocol. We do observe such behavior by some solvers.
It is time to use a better metric.
The new metric
The new metric will combine bid quality and settlement success rate.
For each auction and each order which got executed in that auction, we define relative surplus of a bid as
relative_surplus(order, solver) = proposed_surplus(order, solver) / proposed_surplus(order, winner)
Here we only consider the best bid of each solver from any of the fair solutions they submitted. Surplus is the difference of amounts proposed for execution and limit amounts.
The settlement success rate per accounting period is defined as
success_rate(solver, period) = count(orders settled by solver in period) / count(orders won by solver in period)
If an order is part of multiple auctions, it is counted multiple times. If a solver does not win any order in any auction, the success rate is set to zero.
The new consistency metric is defined as
new_metric(solver, period) = success_rate(solver, period) * sum(relative_surplus(order, solver) / sum(relative_surplus(order, solver’) over solver’) over order in period).
The new metric has two main components.
relative_surplus(order, solver) / sum(relative_surplus(order, solver’) over solver’): Instead of just counting bids like the current metric, the new metric rewards solvers with larger bids more. Only large bids will contribute to a robust service with good prices. Dividing by the sum of relative surplus of all solvers bidding on the order normalizes the contribution per trade to 1. This means that orders with few solvers bidding are more profitable for these solvers and there is a larger incentive for other solvers to join such competitions.success\_rate: We want to encourage bids but only if they are credible. A fast service requires trades to get settled and thus high settlement success rates.
The full metric is the product of these numbers, discounting bid quality by success rate.
Example
Suppose 4 solvers bid on an order in some auction
- Solver A bids 0.1ETH, wins and executes the trade
- Solver B bids 0.08 ETH
- Solver C bids 0.019 ETH
- Solver D bids 0.001 ETH
The relative surplus compared to the winner is
- Solver A: 1.0
- Solver B: 0.8
- Solver C: 0.19
- Solver D: 0.01
The sum of relative surplus for this trade is 2.
If these solvers have success rates 0.8, 0.9, 1.0, 0.5, respectively, in the relevant accounting period, this trade will contribute as follows to the consistency metric
- Solver A: 0.8 * 1.0 / 2.0 = 0.4
- Solver B: 0.9 * 0.8 / 2.0 = 0.36
- Solver C: 1.0 * 0.19 / 2.0 = 0.095
- Solver D: 0.5 * 0.01 / 2.0 = 0.003
The proposal mitigates the issue of solvers submitting bids with almost zero surplus and no intention to settle. The first term implies that submitting such bids does not increase payments. The second term, which might decrease when bidding without the intention to settle, means that submitting such bids can even decrease payments.
The distribution of rewards on mainnet would roughly change as follows (considering the period 2026-03-01 to 2026-05-31).
Alternatives to the new metric like discounted surplus from CIP-85 and weighted averages of different metrics have been discussed internally and with solvers. The core team considers the proposed metric as a good compromise between simplicity and encouraging the two important factors of quality of service: quality of prices and time of execution.