Sources
Written up in November 2026 from the lab’s February 2026 data research on HYPE staking tokens, carried out alongside the liquid staking marketplace. Thresholds and timings are as observed then.
A liquid staking token trades at a small discount to the asset it redeems for. The discount is the fee a buyer charges for waiting out the redemption queue. A separate note works out the floor for that fee. This one is about where, above the floor, the fee should sit at a given moment.
During research on HYPE staking tokens in early 2026, three measurements did most of the work. All of them come from public on-chain data, visible to every participant at the same time.
1. How long the queue will take
Hyperliquid’s unstaking queue is seven days, and one address can have at most five withdrawals pending. A protocol that stakes for many users has to batch their requests, and the way it batches decides how long a given request waits.
The two tokens we studied behave differently. Kinetiq’s kHYPE files unstaking requests on a regular cycle run by a keeper. A request submitted just before the keeper runs spends about a day less in the queue than one submitted just after. stHYPE settles pending withdrawals against new deposits, as Lido does for stETH: when someone stakes, their HYPE pays out the oldest request and nothing has to leave the validators. While deposits are steady, a request can settle in under four hours.
A staked token earns nothing while it sits in the queue, so the number that matters is the expected wait at the moment of filing. Its inputs are the size of the queue ahead, the recent rate of new deposits, and the protocol’s processing schedule. The result is a forecast. It is wrong when flows change suddenly, and it should be published with the time it was made.
2. Leverage that may have to unwind
Much of the selling in a staking token comes from one strategy. A holder deposits the staked token as collateral, borrows HYPE against it, stakes the borrowed HYPE, and repeats. The position earns the staking yield on the whole amount and pays the borrowing rate on the borrowed part. It makes money while borrowing costs less than staking earns.
When the borrowing rate crosses the staking yield, the position loses money every day it stays open, and holders unwind by selling the staked token. In February 2026 that line was near 2.2% a year for kHYPE and 2.4% for stHYPE. The HYPE borrowing rate on HyperEVM lending markets, plotted against those two levels, was the most useful single chart in the research.
Forecasting the rate before it moves is harder. Rates jump when large lenders withdraw, and their reasons vary: a move in the HYPE price, a rise in volatility, flows in and out of vaults that lend. We found many correlations and a great deal of noise. We have no reliable leading indicator to report.
The signal is used for sizing. If large unwinds look likely, a buyer holds capital back, because better prices are coming. If they look unlikely, the buyer commits more at the current discount.
3. Whether sellers would sooner wait
A holder who wants out can sell now for a fee or join the queue for nothing. The ratio of redemption requests to sales over a week shows how sellers are making that choice. When requests rise relative to sales, sellers have looked at the market price and decided the fee is too high.
The ratio changes with the reason for selling. Holders unwinding because the HYPE price is moving fast pay the fee with little resistance. Holders unwinding because of a brief jump in borrowing rates are much more likely to queue. A fee that suits the first case is too high in the second.
From signals to a fee schedule
In the marketplace, the fee for selling a staked token is a schedule of steps, keyed to how much of the pool’s liquid reserve has been used once the trade is done. A small sale into a full pool pays the lowest step. A sale that would leave the pool short pays more. Steps never decrease, so splitting a large sale into pieces does not earn a lower fee.
| Reserve used after the sale | Fee |
|---|---|
| Up to 2,000 | 1 bp |
| 2,000 to 4,000 | 2 bps |
| 4,000 to 12,000 | 4 bps |
| 12,000 to 50,000 | 10 bps |
| Above 300,000 | 100 bps |
An illustrative schedule with intermediate steps omitted, in units of the staked token. One basis point (bp) is 0.01%.
The three signals move the steps. A long queue raises the upper steps, since a large sale would wait longer to be redeemed. Borrowing rates approaching the staking yield raise them too, ahead of the selling. A rising share of redemption requests lowers the bottom steps. If updates stop arriving, the schedule still responds to the pool’s own reserves, so the pool keeps a workable price without the outside inputs.
What the same data can be misused for
Anyone who can read these signals can also lean on them. A buyer could withhold capital to deepen a discount, push up the borrowing rate on one lending market to force unwinds, or fill the redemption queue so that selling is the only quick exit. Each produces a single profit and makes the leveraged strategy less safe to run, which removes the selling the market exists to serve. A venue that expects to be operating next year has a reason to refuse these trades. A firm passing through has none.