Time-weighted staking rewards pay on duration as well as size, so the unit being rewarded is token-seconds rather than tokens. A holder who staked 100 tokens six months ago outranks one who staked 100 tokens yesterday, and capital that arrives just before a snapshot earns close to nothing. The mechanism is a direct answer to just-in-time staking, and its behaviour is decided almost entirely by two implementation choices: whether the weight accrues continuously or in tiers, and what a partial withdrawal does to the accrued weight.
Tiered multipliers create cliffs, and cliffs get farmed. If jumping to the next tier is worth more than the accrual you lose by unstaking and restaking, participants will do exactly that, and the mechanism built to reward patience will have taught people to churn.
Scroll to see the full diagram
The unit being rewarded is token-seconds
Strip the branding off and every implementation computes the same thing: an amount multiplied by a duration. Bestla VC's design writeup states the accounting form directly, using stake age measured in token-seconds, where 50 tokens held for one day produces 4,320,000 token-seconds; new deposits pull the average stake age down while withdrawals leave the age of the remaining balance intact.3
Ethereum's base layer does none of this, and the contrast is instructive. A validator's consensus reward is computed from effective balance and total active balance with no tenure term at all, so a validator active for three years and one active for three epochs earn the same reward for the same duties.4 Time-weighting is an application-layer choice, not a property of proof-of-stake.
That is worth saying because founders often import the mechanism assuming it does something for security. It does not. It changes who holds the token and for how long, which is a distribution and liquidity objective. If the underlying business does not generate a reason to hold after the multiplier stops growing, time-weighting has purchased a delay, not an outcome.
Continuous accrual and tiered multipliers behave differently
A continuous curve pays for each additional second held. A tiered curve pays nothing extra between thresholds and then jumps. Bestla's writeup names the failure mode that follows: tiered multipliers such as 1x at 30 days, 1.5x at 90 and 2x at 180 create cliff effects where participants unstake and restake to reach the next tier, and it cites Synthetix as a system where this behaviour has been observed.3
Published curves are usually a hybrid, which is where the trouble hides. Giza Protocol documents boosts of 25 percent after one week, 37 percent after one month, 82 percent after three months, 175 percent after six, 360 percent after twelve and 400 percent after twenty-four.1 Look at the last step. Twelve additional months of lockup buy an increase from 4.6x to 5x. Anyone doing that arithmetic at month twelve is reading a curve that has stopped paying for patience.
Design consequence. Plot your own multiplier curve and mark where its slope goes flat. That point is where your unlock pressure concentrates, and it is knowable at design time rather than discoverable at month twelve. Pairing the curve with rolling exits rather than a single cliff is the standard fix, and the vesting cliff entry covers why a single date is the wrong shape for anything.
Per-deposit accounting is what makes stake-splitting neutral
The subtlest implementation decision is whether a user's stake is tracked as one blended position or as a set of individual deposits. It sounds like bookkeeping. It decides whether the mechanism can be gamed.
One documented quadratic implementation scores each deposit as its amount multiplied by the square of the time it spent at risk, and maintains the sums per deposit rather than blending a user's total into a single weighted-average entry time. The stated reason is Jensen's inequality: with a blended average, a user who topped up an existing position would lose part of their bonus, whereas per-deposit accounting means splitting a stake across several addresses and topping up a single address produce the same mathematical result.2 Same implementation also clamps each deposit's start time to the opening of the observable risk window and fixes the terminal time at the moment risk ended, so nobody gains weight by delaying their claim.2
Giza reaches the same conclusion from the product side, stating that both the time-weighted boost and its unstaking fee are calculated at the individual deposit level, so two deposits made six months apart are treated independently.1 Two very different systems, one shared answer: account per deposit, and Sybil-splitting stops being profitable without any identity check.
The reset rule is the whole mechanism
Everything above is preamble to one parameter: what happens to accrued weight when a holder takes some of their stake back. Giza's answer is blunt. Withdrawing any portion of staked tokens resets the time-weighted boost associated with that stake, stated explicitly as a measure to encourage consistent long-term participation.1 They pair it with a withdrawal fee of 5 percent inside the first month, decaying linearly to zero over 30 days.1
A hard reset is the strongest version and it has a cost. It converts a partial exit into a total loss of accrued position, which means a holder facing a liquidity need withdraws everything rather than the amount they need, since the marginal cost of the rest is now zero. Softer variants scale the reset to the fraction withdrawn, at the price of being easier to model and therefore easier to optimise against.
Neither is correct in the abstract. What we ask for is that the choice be deliberate and stated in the same paragraph as the boost table, because a holder reading a 5x multiplier without the reset rule next to it has been shown half a contract.
What time-weighting does not fix
It does not fix the exit. A curve that concentrates weight at a duration boundary concentrates unstaking at that same boundary, and the aggregate release is now synchronised rather than spread out. If the reward is denominated in the same token being locked, you have also correlated the sell pressure with the reward payout. Pair duration incentives with rolling unlocks so exits distribute across time.
It does not fix the funding source either. A multiplier changes how an emission budget is divided, not where that budget comes from. Bestla's framing puts this well by treating the maximum multiplier as a governance parameter to be set conservatively and adjusted, rather than a launch-time headline.3 A 5x boost on a rate paid entirely from issuance is a 5x share of dilution, and the APR versus APY entry covers how those rates get quoted.
Underneath sits the question we anchor every engagement to. Time-weighting is a tool for aligning holders with a business that is worth holding. It is not a substitute for one. This page is reference material for design work, not investment advice, and not a recommendation about any protocol or token.
Common questions
What are time-weighted staking rewards?
Rewards that scale with how long capital has been staked, not only how much. The unit being rewarded is effectively token-seconds: an amount multiplied by a duration.3 Capital that arrives just before a reward snapshot earns close to nothing, which is the point. Giza Protocol publishes one example curve running from a 25 percent boost after a week to 400 percent after twenty-four months.1
Can time-weighted rewards be gamed by splitting a stake?
Only if the implementation blends deposits. Systems that track each deposit separately make splitting a stake across several wallets mathematically identical to topping up one, which removes the incentive entirely without any identity requirement.2 Giza states that both its boost and its unstaking fee are calculated per deposit.1 Blending everything into one weighted-average entry time is what creates the arbitrage.
Do tiered staking multipliers work better than continuous accrual?
Tiers are simpler to communicate and easier to farm. Because nothing accrues between thresholds and then jumps at one, participants have a direct incentive to unstake and restake to reach the next tier, a pattern Bestla VC's design writeup names as a cliff effect and attributes to systems including Synthetix.3 Continuous accrual removes the jump, at the cost of a boost number that is harder to put on a landing page.
Does Ethereum use time-weighted staking rewards?
No. A validator's consensus reward is computed from its effective balance against total active balance, with no term for how long it has been active, so a three-year validator and a three-epoch validator earn the same for the same duties.4 Time-weighting is an application-layer design choice used by protocols to shape holder behaviour, not a feature of proof-of-stake consensus.
See Tokenomics Design for how this applies in practice.
Sources
- Time-Weighted Staking
Giza Protocol Docs, 2026
The published boost schedule from 25 percent at one week to 400 percent at twenty-four months, the rule that withdrawing any portion resets the boost, the 5 percent unstaking fee decaying to zero over 30 days, and per-deposit calculation of both. Read 3 August 2026. - k=2 Formula Mechanics, ConfidencePool
DeepWiki, incjanta/confidence-pool, 2026
A documented quadratic time-weighting implementation scoring each deposit as amount times time at risk squared, the Jensen inequality rationale for per-deposit rather than blended accounting, and the clamping of entry and terminal times to the observable risk window. - How to Design Staking Rewards for Token Longevity
Bestla VC Research, 2026
Named practitioner writing, published 28 April 2026. Token-seconds accounting, the cliff effect produced by tiered multipliers with Synthetix cited as an observed case, and the argument for setting the maximum multiplier conservatively through governance. - Proof-of-stake rewards and penalties
ethereum.org, Ethereum Foundation, 2026
The base reward formula, which contains no tenure term, establishing that Ethereum's consensus layer does not time-weight validator rewards.
Last reviewed 2026-08
Know the terms but not sure how they apply to your project? That is what an engagement is for. We design, document, and stress-test the whole token economy inside the Tokenomics Data Room.
100+ projects advised. Complete tokenomics in 4 to 6 weeks.