Risk frameworks for active onchain strategies.
A note on conservative parameter setting, drawdown ladders, and operational rehearsal for leveraged onchain positions. Drawn from the June 2022 stETH-vs-ETH dislocation that followed the Terra collapse, and intended as a reference for treasuries evaluating these strategies.
01The unlevered baseline
Native ETH staking returns 2.8–3.4% APR at current participation. stETH and wstETH return roughly the same, minus protocol fees. That is the base rate. Everything above it is paid for by leverage and the discipline to maintain it. There is no free yield in a leveraged loop, only borrowed yield against a defined risk.
02Conservative LTV ceilings
Aave V3’s liquidation threshold for stETH-collateralised positions in standard mode is 79%. Naive sizing borrows a point or two below that, harvests the spread, calls it done. This note runs loops well below threshold, around 70%.
The nine-point buffer is sized for three coincident stresses:
- Up to 4 points for an intraday stETH-vs-ETH discount. The June 2022 episode reached roughly 7% at peak; a conservatively sized loop absorbs the bulk of that before automated logic engages.
- Up to 2 points for a stablecoin discount. USDC traded materially under par for several days during the SVB stress in March 2023.
- Up to 3 points for gas and oracle latency. Volatile windows can leave the unwind one or two blocks behind intent.
The yield drag from 70% rather than 76% is roughly 150 basis points. We pay it. Every quarter we re-evaluate; every quarter we keep paying.
03The unwind path
A leveraged loop is only safe with a defined unwind ladder you have rehearsed. Ours, abridged:
The unwind takes seven minutes on a normally functioning Aave market, 45 if Aave borrow liquidity is impaired or underlying DEXs are congested. We rehearse it monthly on a small position, in real-time, so the operations team has muscle memory rather than a runbook.
04Stress events: how the buffer earns itself back
In June 2022, the collapse of Terra and forced unwinds by 3AC, Celsius, and other concentrated holders produced a sustained widening between stETH and ETH. The peak intraday discount reached roughly 7% on 13 June; the secondary-market discount persisted at 3–5% for several weeks before normalising.
Positions at higher LTV ceilings (76%+) entered automated unwind cascades, deleveraging into thin order books at material slippage in the days that followed. Positions at the conservative 70% level held without triggering automated unwind logic; the operator could choose when to reduce. The differential shows the value of a wider buffer when liquidity is impaired for an extended window, not a 30-minute one.
The 600-basis-point gap between configurations turned a multi-week stress event into a passive watch for the conservatively sized mandate. The cost is the yield drag accepted up-front.
The buffer is what makes the trade durable. Yield drag from a conservative LTV is the cost paid to ensure an event like June 2022 passes without forced action.
05What this means for sizing
For a treasury new to leveraged staking, we recommend starting at LTV 60% with a notional cap of 10% of total treasury. The first two months are about operational rhythm (daily reports, alert flow, unwind rehearsals), not optimising yield. Optimisation comes after the committee has seen the position behave through one minor stress.
The full unwind playbook, including transaction templates, monitoring scripts, and mandate-level parameters, is available to mandate principals on request.