AllDefi Research · DeFi Yield Strategy Series (3/N)
TL;DR
- A depeg is not a default. A stablecoin briefly trading at $0.87 doesn't mean it failed — but for a leveraged loop, a temporary pricing event equals a permanent loss of equity.
- What decides whether a loop survives isn't the final magnitude of the depeg, but the moment it first crosses the liquidation threshold. That moment often requires only 3–6% of deviation.
- Minimum depeg that triggers liquidation = (LT − LTV) / LT. The more aggressive the leverage, the less depeg the position can absorb.
- Different depeg mechanisms have very different speeds and recovery curves. A risk system that only watches "price" cannot tell them apart.
1. An Uncomfortable Fact
In March 2023, after Circle's exposure to Silicon Valley Bank became public, USDC traded down to $0.87.
That was a depeg. But USDC did not fail. Circle's reserves were intact. The market simply repriced bank counterparty risk within 48 hours. Two trading days later, USDC was back at $0.999.
For an address holding a 10x USDC loop that weekend, the recovery meant nothing. The position had already been liquidated the moment USDC touched $0.97. USDC came back. The principal did not.
This is the most counterintuitive thing about depegs and leverage: a depeg is not defined by whether the asset eventually recovers. It is defined by whether you are still there at the bottom.
2. Leverage Translates Depeg Into Default
Mainstream narratives often conflate depeg and default. In research, they are two distinct event types.
Default is structural — reserves run short, contracts fail, the asset disappears. Value is unrecoverable.
Depeg is market-driven — at some moment, in some venue, an asset's quoted price diverges from its theoretical value. Most depegs recover.
But under leverage, the difference between the two disappears. Liquidation is based on real-time price, not final price. Once the moment arrives where your collateral mark-to-market × liquidation threshold < your debt, you are liquidated — regardless of what happens a second later.
Without leverage, depeg and default are two different things. At 10x leverage, they are the same thing.
3. How Big a Depeg Triggers Liquidation?
Let LTV be the maximum loan-to-value the protocol allows, LT be the liquidation threshold, and d be the depeg magnitude on the collateral. When a loop is maxed out, the liquidation condition simplifies to:
Minimum depeg triggering liquidation: d > (LT − LTV) / LT
Plug in typical parameter sets and a sharp stratification appears.
An aggressive stablecoin pair with 90% LTV and 93% LT is liquidated by a depeg of just 3.2%. An extreme stablecoin loop (92% LTV / 93% LT) is liquidated by 1.1%. A typical LST loop (85% / 90%) tolerates 5.6% before liquidation. A conservative ETH / stablecoin loop (75% / 85%) needs 11.8% to be wiped.
This produces a counterintuitive conclusion: leverage and depeg tolerance are inversely related.
A 10x stablecoin loop sounds aggressive, but its depeg tolerance is only 3.2%. The March 2023 USDC drop to $0.87 would have liquidated any 10x USDC loop the moment the price touched $0.97 — before the market had time to understand what was happening. A 5x stETH loop sounds safer, but its tolerance is 5.6% — the 7% deviation in June 2022 was enough to wipe out every 5x stETH loop on Aave.
Part 1 introduced 1 / (1 − LTV) as the leverage formula — the "front side." (LT − LTV) / LT is its back side — the depeg threshold. Together, the two define the survival boundary of any loop.
4. Four Mechanisms That Produce Depegs
Different depegs have different speeds, recovery profiles, and signal patterns. Lumping them together obscures the strategic differences that matter.
Type 1 — Reserve / Counterparty Risk. The reference case is USDC-SVB in March 2023. The reserve structure itself is intact, but the counterparty holding the reserves is being repriced. Onset within hours; recovery within days (USDC took roughly 3 days). For a leveraged loop, the damage approaches 100% — even if the asset recovers, liquidation has already occurred.
Type 2 — Redemption Queue Stress. The reference case is stETH-Celsius in June 2022. The asset can theoretically be redeemed 1:1, but the redemption mechanism has time delays — validator exit queues, protocol settlement windows. When redemption isn't immediately available, the secondary market discounts. Slow widening over days to weeks; convergence once the queue clears. Damage to loops is moderate-to-high — discounts erode the health factor over time, but the reaction window is far longer than in Type 1.
Type 3 — Algorithmic / Hedging Mechanism Failure. The reference case is the UST collapse in May 2022. The peg-maintenance mechanism itself fails under extreme market conditions — algorithmic arbitrage, delta-neutral hedging, reserve rebasing. The failure self-accelerates: depegs trigger redemptions, redemptions trigger more depeg. This is the only category of depeg events that regularly leads to permanent asset destruction. Damage to loops approaches total loss regardless of leverage level.
Type 4 — Oracle / Price Source Divergence. Different price sources (CEXs, different DEXs, aggregators) quote the same asset at different prices, and the lending protocol's oracle is using a stale or thin-liquidity source. Instantaneous; recovery within minutes — but not necessarily above your liquidation threshold. Liquidations from this category are "false liquidations" — the asset has not actually depegged, but your position is already gone.
The key difference among the four is not severity — it is signal pattern. Type 1 has counterparty news leading the price. Type 2 has redemption queue metrics. Type 3 shows imbalance in on-chain flows and hedges before the break. Type 4 only appears as cross-venue spread. A risk system that watches only "price" will miss the early signals from the first three and the false signals from the fourth — simultaneously.
5. The LST / LRT Exception
Yield-bearing assets like stETH, wstETH, weETH, and ezETH have a structurally unique property in depeg analysis: their depegs are not continuous drifts. They are step jumps.
Ordinary stablecoin depegs are usually smooth — the market gradually reprices new risk, and price moves continuously.
LSTs and LRTs behave differently. In normal conditions, their secondary market prices track theoretical value within a few basis points. Then at some moment — a large unstaking request, a validator slashing event, a wave of points farming capital exiting at once — the main Curve pool tilts instantly, and the price jumps from 1.000 to 0.96 in a single block.
The impact on loops is non-linear. The frequency of depegs is lower than with stablecoins, but the immediate impact of a single event is larger. Any strategy that depends on "see deviation, then react" fails on LSTs and LRTs — by the time deviation appears, the reaction window has already closed.
6. How AllDefi Handles Depeg
Continuing Part 2's three-layer architecture, with the specific variables adjusted for depeg events.
Layer 1 — Continuous Monitoring. The system tracks, every block: real-time prices for the collateral asset across multiple liquidity venues, the deviation between price sources, and the rate of change in that deviation (continuous vs. step-jump). Cross-source price comparison is the core signal for identifying oracle-divergence-type depegs.
Layer 2 — Threshold-Tiered Response. Deviation crosses preset alert threshold (e.g., 0.5%) → halt new loop additions. Deviation continues widening without converging within a reasonable window → actively unwind part of the position. Deviation approaches the pair's liquidation threshold (per the formula in Section 3) → exit fully. Significant divergence appears between price sources → switch execution path to avoid pools that feed the oracle. Each rule is pre-defined. No human decision required in the loop.
Layer 3 — Execution Optimization. Unwinding is split across multiple transactions and DEX venues to avoid single-point impact. Execution pace adjusts dynamically based on observed slippage and gas. Deeper pools are preferred even when the path is longer.
In a moderate-severity depeg event, the system typically completes risk reduction within 30 minutes — which is precisely the window where most depeg events still have an open reaction window.
7. Back to the Question
How big a depeg wipes out a loop?
Not 5%. Not 10%. Not any single number.
It depends on the leverage, the type of depeg mechanism, and how fast the position responds.
A 10x stablecoin loop liquidates at 3.2% depeg — an event that happened in March 2023, and will happen again. A 5x LST loop is wiped at 5–7% deviation — happened in June 2022, and in several recent LRT incidents. When an algorithmic mechanism breaks, no leverage level and no reaction speed gets you out in time.
Part 1 showed how looping leverage is calculated. Part 2 showed how interest rates can give back the profits. This part shows how a depeg can take the principal in a single moment.
The three reports together form the complete mathematical skeleton of looping risk.
A skeleton is not a tool. Turning the math into executable real-time risk management is what AllDefi is building.
Series Roadmap
- (4/N) The cost of exiting under liquidity stress: real-world slippage and borrow rates during crisis events
- (5/N) On-chain looping concentration: a backtest using Aave V3 mainnet data
- (6/N) From retail looping to automated looping: the AllDefi system architecture
AllDefi is a Multi-chain Institutional Yield Layer, delivering structured DeFi yield strategies to institutional capital — along with the automated risk infrastructure required to run them. This report is for research purposes only and does not constitute investment advice.
