EIP-8068 Neutral effective balance design

No call decision · Consensus & staking · EIP status Draft · created 2025-10-28

Changes how the effective balance is calculated so that skimming (0x01) validators and compounding (0x02) validators receive the same yield. It removes the structural disadvantage for compounding validators, encouraging validator consolidation.

Why it is needed

Ethereum's fast finality roadmap assumes that staking service providers will consolidate multiple 32 ETH 0x01 validators into compounding (0x02) validators. But under the current design, compounding validators lose out. The idle balance not captured by the effective balance, which is the basis for reward calculation, is nearly zero for 0x01 validators but averages 0.75 ETH for compounding validators. As a result, a 32 ETH 0x01 validator earns about 2.3% more consensus layer yield than a compounding validator. Separately, there is a loophole where a compounding validator can partially withdraw down to a balance of 32.75 ETH while keeping an effective balance of 33 ETH once the effective balance exceeds 33 ETH, then redeploy the withdrawn capital to another validator for extra yield. Both problems hinder validator consolidation, so the proposal is to make the effective balance calculation neutral so that every configuration receives the same yield.

What changes

Today

After EIP-8068

Who is affected

In depth

Changes to the hysteresis thresholds

HYSTERESIS_DOWNWARD_MULTIPLIER is changed to 3 and HYSTERESIS_UPWARD_MULTIPLIER to 2. As a result, the threshold for rising from effective balance E to E+1 becomes E+0.5, and the threshold for falling to E becomes E+0.25. Under steady-state compounding, the effective balance becomes the balance rounded to the nearest integer, so compounding 0x02 validators also maintain E=b on average. Under the current design, the idle remainder r=b-E lies in the range [0.25, 1.25), averaging 0.75 ETH, so a 0x02 validator compounding from 32.25 ETH to 34.25 ETH over about two years earns roughly 2.3% less CL yield than a 32 ETH 0x01 validator (whose average idle balance is 0.0235/2=0.01175 ETH).

Two new fields in the beacon state

temporary_upward_threshold: List[Gwei, VALIDATOR_REGISTRY_LIMIT] and reset_eb_flags: List[boolean, VALIDATOR_REGISTRY_LIMIT] are added to BeaconState. At the upgrade boundary, both lists are initialized with length len(state.validators), to 0 and False respectively. temporary_upward_threshold stores the one-time upward threshold applied right after a partial withdrawal, and reset_eb_flags marks validators whose effective balance must be reset during the next epoch processing.

Reset logic on partial withdrawal

After decrease_balance() is called in process_withdrawals(), if the validator has a compounding withdrawal credential (has_compounding_withdrawal_credential) and exit_epoch is FAR_FUTURE_EPOCH, reset_eb_flags is set to True. During epoch processing, the new helper initiate_effective_balance_and_threshold() is called to reset the effective balance to the smaller of balance - balance % EFFECTIVE_BALANCE_INCREMENT and get_max_effective_balance(v). Then offset = (EFFECTIVE_BALANCE_INCREMENT + (balance - effective_balance)² // EFFECTIVE_BALANCE_INCREMENT) // 2 is added to the effective balance to store temporary_upward_threshold, and the flag is reset to False.

Modifications to process_effective_balance_updates()

Validators with reset_eb_flags set first call the reset helper and skip the regular update for that epoch. If the downward condition (balance + DOWNWARD_THRESHOLD < effective_balance) holds, the effective balance is lowered to the floor value (balance_floor) and temporary_upward_threshold is cleared to 0. An upward update requires both effective_balance + UPWARD_THRESHOLD < balance and balance >= temporary_upward_threshold. In that case, if the fractional part of the balance is at least half of EFFECTIVE_BALANCE_INCREMENT, the effective balance is raised to the floor value plus 1 INCREMENT (rounding up); otherwise it is raised to the floor value, and temporary_upward_threshold is reset to 0.

Integral derivation of the temporary upward threshold

Resetting the effective balance to E = ⌊b0⌋ after a partial withdrawal creates an initial idle remainder r0 = b0 - E. The temporary threshold t is set so that, while the balance grows by compounding from b0 to the next integer E+1, the average idle balance becomes zero. This is expressed as a 'zero area' condition: the positive idle balance accumulated over the range where EB=E (from b0 to t) must exactly cancel the negative idle balance over the range where EB=E+1 (from t to E+1). Writing this as the integral equation ∫(b-E)db + ∫(b-(E+1))db = 0 and solving it yields the formula given in the specification, t = E + (1+r0²)/2. This way, even a partial withdrawal is treated, on average, as the neutral state E=b.

Analysis of the yield gap under the current design

The original text compares relative yields across configurations under the current design. A compounding 0x02 validator earns 100×0.75/b % less yield than the baseline due to an average idle balance of 0.75 ETH (2.33% at a 32.25 ETH baseline). By contrast, a validator that repeatedly withdraws down to 32.75 ETH using hysteresis earns 0.56% more than the baseline even after accounting for a partial withdrawal gas cost of 84500, a base fee of 0.5 gwei, and a 4.5-day delay before the withdrawn ETH becomes usable on the execution layer (assuming a 256-epoch wait plus an average 3.4-day sweep wait, based on 770,000 active validators). Without the delay it is 0.60%, and with a zero base fee plus the 4.5-day delay it is 0.72%. At current base fee levels, this strategy is more profitable than a 2048 ETH skimming validator.

Alternative specification: a minimal fix that only prevents the loophole

A minimal spec is also proposed that leaves the yield gap as is and only blocks the workaround. It adds only pending_eb_resets: List[ValidatorIndex, PENDING_PARTIAL_WITHDRAWALS_LIMIT] to BeaconState, and on a partial withdrawal the validator index is placed in it. In process_effective_balance_updates(), the effective balance is reset to the minimum of three values, the floored balance, the maximum effective balance, and the existing effective balance, so the effective balance never increases during a reset. In this case, the idea is to offset the lower yield of 0x02 validators with the withdrawal fee increase from EIP-8062.

Remaining issues and current status

This EIP causes the effective balances of many validators to rise all at once at the hard fork boundary, and the source states that this effect needs further analysis. The source's status is Draft, so the spec may still change. The call notes provided contain no decision on whether this EIP is included in a fork.

Original EIP · ← EIPs