The metric for consistency rewards will be changed to only reward orders which are in-market at order creation. Orders which are out-of-market will not contribute to consistency rewards but still count towards performance rewards as usual. The change is planned to take effect with the accounting week 2026-09-15 to 2026-09-22.
Context
With the current consistency metric, the protocol distributes one share of the budget for consistency rewards to every executed trade. The share is distributed among all solvers proposing a trade in proportion to the surplus they propose. This means that order size or profitability of the order do not impact consistency rewards for that order. Consistency rewards therefore act as a form of subsidy, which can be exploited.
Solvers reported, and our monitoring confirmed, that a large fraction of rewards in the last few weeks went towards orders which would not have been profitably executed without such a subsidy. This affected up to 18% of consistency rewards on mainnet and up to 47% of consistency rewards on BNB.
As an example, an order might trade 1 USDC for at least 1.01 USDT. A solver might execute such an order by giving it 1.02 USDT and paying $0.03 worth of the native token to cover network costs. Normally, the solver would have made a loss. But with consistency rewards of $0.06 worth of the native token, solvers executing such orders make a profit. While such orders have a place to provide liquidity for other traders, they are not expected to continuously be executed by solvers against public liquidity.
This dynamic has been used by third parties for arbitrage trading, with solvers facilitating these trades offsetting losses with rewards.
We therefore propose a small change to the metric for computing consistency rewards: we only reward orders which are in-market.
Change to metric
The consistency metric is changed such that only orders which are in-market at order creation result in consistency rewards.
An order is in-market if its limit price is no better than the price implied by the quote which led to the creation of the order net of all fees. This definition is already used for quote rewards.
This change is done within the mandate of CIP-85 on changing the metric for consistency rewards. All other aspects of consistency rewards as well as rules for performance rewards stay as is. Due to the urgency of the matter, we plan to apply this change retroactively to the current accounting period.
The general idea for this change is that we do not want to subsidize order executions which would clearly not have been possible without subsidies. This is not simple to detect as it requires knowledge of exact market prices. We have therefore tested several approaches.
Backtesting and expected impact
We have tested different approaches to adapt consistency rewards to prevent subsidizing arbitrage trades. Besides reusing the in-market classification already in use for quote rewards, we have also tested the use of native prices to detect arbitrage trading, restricting rewards based on the number of trades per wallet, as well as other variants of these rules.
We measured the changes against impact on rewards towards wallets trading tokens back and forth at a profit, see this Dune query, compared to impact for other traders. The rule on using the in-market classification performed best in these experiments.
The change is expected to reduce consistency rewards for arbitrage orders on mainnet and BNB by around 60% to 80% while consistency rewards earned on other flow are affected by around 10% to 20%. The total budget for consistency rewards stays the same.
There are also disadvantages to excluding out-of-market orders, as solvers might not execute them as well anymore for legitimate flow. Limit orders are still an important part of the protocol and result in significant protocol revenue and performance rewards. Since the proposed rule performed best in tests and since the problem of subsidizing arbitrage flow is currently disrupting consistency reward payments, we settled for this solution now.