Decentralised News · 2027 planning series · Reviewed 22 November 2026
What Matters
A refund reduces an execution-cost record only when the amount belongs to the same transaction scope and economic beneficiary, and the record establishes that it was received. A promising refund setting, a pending amount and a provider’s received total answer different questions. Adding them together can make an expensive route appear free.
For Flashbots Protect, the practical task is to identify the refund mechanism, follow the configured recipient and reconcile the resulting payment with the costs actually paid. This article supplies that method and an exact-decimal ETH worksheet. Its example is synthetic: it demonstrates an accounting error, not a refund earned by DN or a forecast for your next swap.
The Signal
Flashbots documents gas-fee refunds separately from MEV refunds. A combined balance view can conceal that distinction. The useful record preserves which mechanism produced an amount, which address was credited and which underlying transaction or cohort it belongs to before calculating an execution-cost subtotal.
Two Refund Mechanisms Need Two Labels
Protect’s MEV-refund guide describes eligibility through MEV-Share when a transaction generates extractable value. Its privacy hints affect what searchers can learn and can influence refund opportunities. Eligibility does not specify the amount a particular transaction will generate or establish that a payment has arrived.
The gas-fee guide describes a separate mechanism for transactions that can land at a lower price. It covers priority-fee and coinbase-transfer payments and depends on BuilderNet landing the block. Congestion, competition, block economics, contribution and submission route affect the outcome. A large fee alone does not establish a refund amount.
| Record | What it contributes to the review | What it does not establish |
|---|---|---|
| MEV refund | A separately identified return associated with the transaction’s MEV | A percentage of the swap’s notional value |
| Gas-fee refund | A separately identified return associated with inclusion payments | Automatic repayment of the whole transaction fee |
| Pending refund entry | An amount reported as awaiting receipt | Cash already available to offset paid costs |
| Received payment record | A candidate credit to reconcile against the selected scope | Proof that every amount in a batch belongs to one trade |
Use those labels even when the amounts ultimately arrive in the same asset. ETH describes the unit; it does not identify the economic event. A token swap, a payment for a service and a refund can all affect a wallet while belonging to different parts of the analysis.
Eligibility Is About the Submitted Route
The gas-fee guide excludes public-mempool flow, independently non-exclusive submissions and submissions sent directly to MEV-Share. It separately documents Protect-managed sharing with other builders. Those paths should not be collapsed into a single “private transaction” category. Preserve the route used for the particular submission.
BuilderNet’s refund explanation describes remaining block value and the contribution of submitted orderflow. It also labels its published flat-tax formulation indicative, with implementation adjustments. A receipt’s gas price is insufficient to reconstruct that counterfactual block calculation. The worksheet below therefore reconciles supplied amounts rather than predicting an entitlement from gas spent.
That boundary matters when comparing routing experiments. If one group used different submission settings or landed through a different path, a difference in refunds may reflect the treatment itself. Record the difference before calling it a performance improvement. A meaningful comparison also needs transaction selection, observation cutoffs and costs defined consistently.
Follow the Recipient, Not the Wallet You Expected
The Protect settings guide distinguishes recipient rules. RPC MEV refunds can be split across configured recipients. Gas-fee refunds go entirely to the first specified refund address; without an override, the RPC default is tx.origin. The percentage attached to that first address does not split the gas refund.
For a synthetic illustration, suppose an RPC configuration allocates 20% of the relevant MEV refund to FIRST and 60% to SECOND. The remaining 20% goes to the validator under the documented RPC rule. FIRST receives the entire gas-fee refund. If the cost payer controls only SECOND, FIRST’s gas receipt cannot simply reduce that payer’s cost record.
The private-transaction API needs its own identity map. Its documentation describes validity.refund for backrun-refund addresses and percentages, with a transaction-signer default. The gas-fee guide instead identifies the signer used on the private-transaction request for gas refunds. Preserve the request-signing identity separately from the embedded transaction sender, and inspect any configured fee-refund delegation.
A recipient map should distinguish the technical address from the account boundary used for the cost comparison. An application may submit on behalf of a customer, or an organisation may control several addresses. Neither arrangement can be inferred merely from matching display names. Keep the relationship and any allocation agreement in the supporting record; the worksheet accepts aliases and verifies neither ownership nor authority.
Choose One Transaction or One Matched Cohort
Start with a scope sentence: “This record covers transaction TX-A, costs borne by OWNER-A and refund receipts attributable to that transaction as observed at CUTOFF-A.” For a cohort, replace TX-A with a defined set of transactions and preserve the membership list. The observation cutoff applies to the refund evidence as well as the costs.
DN proposes keeping four references together:
- The submitted transaction or cohort membership, with its network and inclusion evidence.
- The routing configuration and recipient identities applicable when those transactions were submitted.
- The paid cost and received payment references, including any justified allocation of a batch.
- The observation cutoff and the provider report’s coverage.
This is a matching method, not a security score. A missing allocation cannot be repaired by giving the record more points for having a transaction hash. If the refund is known only for an entire batch, retain a batch-level result until a narrower attribution is supported.
For example, a recipient report showing 0.006 ETH received across five transactions does not establish a 0.006 ETH refund for TX-A. Even if TX-A is one member, the other four remain in the numerator. Either isolate TX-A’s supported share or include the corresponding costs and payments for the full cohort. Do not change only the denominator to improve the apparent saving.
Build the Cost Side From Paid Fees
Ethereum’s transaction-receipt documentation distinguishes gasUsed for the particular transaction from cumulativeGasUsed, and defines effectiveGasPrice as the base fee plus tip paid per gas unit. For an ordinary Ethereum-mainnet execution-fee record, multiply the transaction’s gas used by its effective gas price. A pending transaction has no such receipt yet.
EIP-1559 distinguishes the maximum fee cap from the effective payment and specifies that the base fee is burned. A maximum authorisation is not itself a paid expense. The difference between that cap and the actual fee should not be entered as an additional Flashbots cash refund.
Consider a synthetic transaction using 120,000 gas at an effective 30 gwei: its execution fee is 0.0036 ETH. A 180,000 gas limit and 60 gwei maximum fee would describe an upper envelope of 0.0108 ETH, not another 0.0108 ETH payment. Subtracting the unused envelope from the actual expense would manufacture a benefit.
Add separately paid, attributable expenses only once. A service fee already included in another recorded debit needs no second line. The input asset exchanged in a swap is not an execution fee. This worksheet excludes token principal, output valuation and trading gains or losses; it also does not reconstruct additional fee components for other chains or transaction types.
An API Total Is a Report, Not Another Transfer
The Flashbots RPC reference gives recipient totals with pending, received and maxBlockNumber. Its detailed recipient method supplies paginated per-order records, including hash, amount in wei, block, status and recipient. The documentation describes weekly total calculations and a usual four-hour detail delay; these are not a payment deadline.
Keep the endpoint, returned coverage and observation time with the extract. The reference also states a historical block boundary for detailed records. A missing detail row is not automatically evidence of no lifetime refund. Where fee-refund delegation is configured through flashbots_setFeeRefundRecipient, preserve that mapping too.
For accounting, compare the report with the payment evidence; do not sum the two representations. If the API labels 0.0012 ETH received and the matched wallet movement is the same 0.0012 ETH, the credit is 0.0012 ETH. Two observations of one event are useful corroboration, not two events.
Nor does subtracting two recipient-total snapshots necessarily produce the refund for transactions submitted between those snapshots. A later distribution can relate to earlier transactions. Changes in processing coverage can alter which records are represented. Define the cohort using transaction membership and payment attribution, then compare reports at a common observation boundary.
If provider and payment records disagree, retain the difference as an investigation item. It could concern timing, scope, recipient mapping or an error. The number alone establishes neither a debt nor misconduct. The immediate useful action is to identify which record would explain the difference.
Worked Case: A False Zero-Cost Result
The following independent example assumes OWNER-A has documented both refund receipts as belonging to TX-A. Every number is synthetic, in ETH, and observed at one declared cutoff. It does not apply the FIRST/SECOND recipient arrangement above.
| Entered record | ETH | Treatment |
|---|---|---|
| Transaction execution fee | 0.0036 | Paid cost |
| Separate paid service fee | 0.0002 | Paid cost; no overlap with the execution fee |
| Gas-refund payment received | 0.0012 | Count once as a receipt |
| MEV-refund payment received | 0.0008 | Separate receipt |
| API-labelled gas refunds received | 0.0012 | Compare with the gas-payment record |
| API-labelled gas refunds pending | 0.0006 | Keep outside current receipts |
Paid costs total 0.0038 ETH. Received refunds total 0.0020 ETH. The recorded net execution cost is therefore 0.0018 ETH. The API received amount agrees with the gas-payment record, leaving a zero reconciliation difference.
The tempting error is to subtract the API’s 0.0012 ETH again, then subtract the pending 0.0006 ETH. That produces zero: 0.0038 − 0.0012 − 0.0008 − 0.0012 − 0.0006. The result combines a duplicated receipt with money not yet received.
If the pending gas amount subsequently arrives and belongs to exactly this scope, a separate conditional calculation would reduce the cost to 0.0012 ETH. That is a scenario, not the current result or a prediction of payment. When recording the actual later receipt, update the observation cutoff and remove its pending classification rather than preserving both statuses as credits.
Now change only the API received total to 0.0015 ETH. The 0.0003 ETH difference requires reconciliation. It does not increase the observed gas receipt to 0.0015 ETH. The worksheet retains the cost calculation from the supplied paid ledger but withholds the hypothetical pending-payment result until the report comparison agrees.
Another branch changes the beneficiary instead of the amount. Suppose the same transaction costs 0.0038 ETH and OWNER-A receives the 0.0008 ETH MEV payment, but the 0.0012 ETH gas payment belongs to an unrelated party. OWNER-A’s recorded cost is then 0.0030 ETH. The other party’s receipt is not a credit to OWNER-A. Preserve it as explanatory recipient evidence outside that cash ledger. A later transfer to OWNER-A would require its own receipt and attribution; an expectation that the other party will pass the money on is not an existing payment.
Use the DN Refund Receipt and Execution-Cost Ledger
Enter one line per paid movement using the displayed format. Gas costs, other paid costs, received gas refunds and received MEV refunds have separate kinds. Each line includes a unique reference, the selected scope, the account-boundary alias and an ordinary decimal ETH amount. Do not paste signed transactions, authentication headers or private account data.
Single-transaction mode accepts exactly one gas-cost row. Cohort mode permits multiple transaction receipts. A cohort still requires a defined membership and matching costs; the mode does not authorise combining unrelated activity. Duplicate references are rejected across all categories, including a payment entered once as a gas refund and again as an MEV refund.
The separate API section never supplies an extra cash credit. A mismatched report scope, beneficiary or cutoff disables that comparison while preserving the entered ledger’s calculation. Unknown or unused report fields can remain blank. For an applicable report, both amounts are required; an empty value is not silently converted to zero.
The net-cost line also requires declared alignment and a complete paid-record set. Otherwise, the tool shows entered subtotals and withholds that conclusion. No refund rows can represent reviewed zero receipts only when the complete-record declaration is justified. A missing payment history is not the same finding.
DN Refund Receipt and Execution-Cost Ledger
Change the illustrative assumptions below. This worksheet runs locally in your browser. It records your declarations; it does not verify evidence, connect to accounts, submit instructions or grant approval.
The initial numbers are synthetic. Unknown evidence remains unresolved. Downloads contain the values you enter; use aliases and avoid private or identifying information.
What a Lower Net Cost Can—and Cannot—Explain
A lower execution-cost subtotal is useful, but it does not establish a better trade. Output amounts, route prices, delays and unfilled or reverted attempts can change the economic comparison. DN’s two-pool split model addresses allocation and execution-cost tradeoffs; this ledger addresses which recorded refunds offset the selected paid costs. Their inputs need the same unit and scope before any combined comparison.
Similarly, changing privacy settings is a routing decision, not an accounting adjustment. The MEV-refund guide describes a tradeoff between shared hints and refund opportunities. The settings guide describes broader builder and TEE sharing in fast mode. A larger historical refund does not by itself establish that another route has acceptable information-sharing properties for the next transaction.
Use DN’s MEV route review for those boundaries. For a solver-mediated workflow, the intent-trading review helps separate the submitted constraint from completion evidence. Do not attribute every payment received after a swap to its execution route without the matching record.
DN Alpha Thesis
The useful advantage is an explainable difference between what was paid, what was received and what remains unresolved. A refund programme becomes decision-useful when its amounts can be assigned to a recipient and transaction scope without duplication. Until then, a larger dashboard number can be less informative than a smaller, fully reconciled receipt.
Methodology, Limits, Corrections and Disclosure
Primary documentation was reviewed on 22 November 2026. Named settings and API coverage can change; preserve the applicable version with each real record. This research did not connect an account, submit a transaction, verify an actual payout or measure refund performance. Official programme descriptions establish documented rules and interfaces, not a guarantee for a particular submission.
The reconciliation method and worked examples are original DN analysis. The worksheet uses integer wei internally and accepts decimal ETH to 18 places. It checks formatting and declared boundaries, not the truth of aliases, receipt evidence, beneficial ownership or batch allocation. A real mixed payment needs a justified allocation outside this worksheet; do not invent separate references to evade its duplicate check.
A negative net-cost result means supplied receipts exceed included costs. It is not a trading-return, tax or legal classification. No commercial referral links are added here. Corrections should identify the affected interface or rule, its date and primary evidence, so the article, source register and accounting assumptions can be updated together.
Sources and evidence register
Primary sources reviewed on 22 November 2026. The article distinguishes documented facts, DN analysis and synthetic examples. Recheck changeable rules and configurations before acting.
- Flashbots Protect: Gas Fee Refunds — Gas-refund mechanism, BuilderNet inclusion and routing conditions, and documented RPC/private-API recipients.
- Flashbots Protect: MEV Refunds — MEV-Share eligibility and information-hint tradeoffs; no per-transaction payout guarantee.
- Flashbots Protect: Settings Guide — Different MEV and gas recipient rules and current fast-mode information sharing.
- Flashbots Protect: eth_sendPrivateTransaction — Private-request interface and validity.refund backrun recipient settings, distinct from RPC configuration.
- Flashbots: JSON-RPC Endpoints — fee-refund totals and records — Pending/received categories, processing coverage, paginated detail and fee-refund delegation.
- BuilderNet: Refunds — Contribution and remaining-block-value explanation; the published refund formulation is explicitly indicative.
- Ethereum: JSON-RPC transaction receipt — gasUsed, effectiveGasPrice and the absence of a receipt for a pending transaction.
- Ethereum Improvement Proposal 1559 — Maximum fee versus effective payment and the burned base fee.
Frequently asked questions
Are gas-fee refunds and MEV refunds the same payment?
They are separate documented mechanisms. Preserve separate labels and payment references even when both are received in ETH.
Does a Protect RPC refund percentage split the gas refund?
No. The documented RPC rule sends the gas-fee refund to the first specified refund address. Percentage splits apply to MEV refunds.
Can a pending refund reduce current recorded execution cost?
No. Keep it outside received cash. The worksheet shows a separate conditional result if a matching pending gas amount later arrives.
Should I add the API received total to wallet refund receipts?
Not when they describe the same payment. Compare the records; count the received movement once. A difference needs investigation rather than another credit.
Can I use a recipient lifetime total for one transaction?
Only after the relevant records establish that transaction’s attributable amount. Otherwise keep a matched cohort or withhold the narrower comparison.
Does a negative net execution cost mean my trade was profitable?
No. It means entered refund receipts exceed the costs included in this ledger. The trade’s input and output values and other omitted costs are separate.