At 2:40 a.m. your phone wakes you: health factor 1.04. Twelve hours earlier the same loop looked bulletproof — 3x exposure, three clicks, an APY that finally beat plain staking. Then ETH prints a 12% candle down, the stETH/ETH peg slips another few points, and by sunrise the position is gone. Nobody called. A liquidator repaid your debt in one transaction, took your collateral at a discount, and moved on.
That loop was not a bad idea executed badly. It was a good idea executed without arithmetic. The number that decided your fate — health factor — was computable before you opened the position, in about ninety seconds, with one division. Almost nobody runs it.
This playbook gives you that arithmetic: the exact formula, HF = Σ collateral × liquidation threshold ÷ debt, and why it is not LTV; a step-by-step build of a 3x loop and the buffer table showing how far price can fall before HF hits 1.0; how e-mode and isolated markets rewrite every parameter; why the borrow rate can liquidate you with prices flat; a de-risking ladder for HF 1.8, 1.5, 1.3 and 1.1; and the mistakes that turn a dip into a forced sale. One $30,000 example keeps the numbers honest.
None of this removes risk. It removes surprises.
TL;DR. Health factor = (collateral value × liquidation threshold) ÷ debt, and below 1.0 you are liquidatable. LTV caps what you may borrow at the door; the liquidation threshold is the cliff that actually feeds HF, so the gap between them is your entire cushion. A true 3x loop runs at a 66.7% effective LTV and, on an 80% threshold, opens at HF 1.20 with only a 16.7% price buffer; build to a 65% LTV and it widens to 18.8%, while a 4x loop leaves 6.25%. E-mode lifts thresholds to ~93–95%. Your borrow rate is a second liquidation trigger — above the utilization kink it can eat the position with prices flat. Open loops at HF ≥ 1.8, watch at 1.5, act at 1.3, treat 1.1 as an emergency. Model yours on the DeFi yield calculator.
Health factor is not LTV, and the gap between them is your margin of safety
The protocol cares about one equation. For a single collateral type it collapses to HF = collateral value × liquidation threshold ÷ debt. Across several deposits it is the sum of each collateral's value times its own threshold, divided by total debt including accrued interest. Above 1.0 the loan is yours to manage; at or below 1.0 the position is permissionlessly liquidatable, with no grace period and no phone call. Aave, Compound, Morpho and Spark define it the same way; Morpho calls its threshold LLTV. If you want the borrower-side walkthrough, the Aave, Morpho and Spark borrowing experience covers it line by line.
LTV is a different animal. It is the maximum you may borrow when you open the position — say 65% or 75% per dollar of collateral. The liquidation threshold is the higher ratio at which liquidators may seize collateral, say 80%. Only the threshold enters the health factor. That means the space between LTV and threshold is the entire amount of price noise the protocol will tolerate before it acts, and it is thinner than it looks: on a 65% LTV with an 80% threshold, you start at HF 1.23 and a routine drawdown does the rest. Borrow interest compounds the problem from the other side: it grows the denominator, so HF drifts down even on a flat chart.
Building a 3x loop, one tranche at a time
Every loop is bounded by the same ceiling: max leverage = 1 ÷ (1 − max LTV). At a 75% LTV cap that is 4x — no sequence of clicks gets past it, because the final borrow cannot exceed the collateral behind it. To land on 3x rather than the ceiling, you borrow roughly two-thirds of each new tranche instead of the maximum, and the series converges. The mechanics below turn $30,000 of stETH collateral into 3x exposure; the same pattern appears in our liquid staking looping breakdown.
| Pass | Borrow & redeposit | Cumulative collateral | Cumulative debt | Effective LTV |
|---|---|---|---|---|
| Start | — | $30,000 | $0 | 0% |
| 1 | $20,000 | $50,000 | $20,000 | 40.0% |
| 2 | $13,333 | $63,333 | $33,333 | 52.6% |
| 3 | $8,889 | $72,222 | $42,222 | 58.5% |
| 4 | $5,926 | $78,148 | $48,148 | 61.6% |
| Converged (3x) | → | $90,000 | $60,000 | 66.7% |
Stop when effective LTV reaches 66.7% and you have a 3x loop; push every pass to the 75% cap and the same series converges on 4x. That choice — how hard to lean on the cap — separates a loop that survives a 20% week from one that does not.
How far the collateral can fall before HF hits 1.0
Liquidation arrives when collateral × threshold = debt, which rearranges to liquidation price = entry price × (debt ÷ collateral) ÷ liquidation threshold. The buffer — how far price can fall from entry — is simply 1 − (debt ÷ collateral) ÷ threshold. With an 80% threshold, the whole table is arithmetic on how much you borrowed.
| Target leverage | Effective LTV | Debt ÷ collateral | Fall from entry to HF = 1.0 (LT 80%) |
|---|---|---|---|
| 2x | 50.0% | 0.500 | −37.5% |
| 2.9x (a 65% LTV build) | 65.0% | 0.650 | −18.8% |
| 3x | 66.7% | 0.667 | −16.7% |
| 4x | 75.0% | 0.750 | −6.25% |
Read the last row against history. A 4x loop tolerates only a 6.25% drop — thinner than the roughly 6–7% discount stETH traded at during the June 2022 depeg, which alone could have put a 4x position at the edge before ETH moved. A full 3x is better but still tight at 16.7%; a 65% LTV build buys 18.8%. The buffer depends on how hard you lean on the LTV cap, not on the APY the dashboard shows. The chain reaction once liquidations begin is covered in our liquidation cascade explainer.
The $30,000 example, priced at seven ETH levels
Take one loop. You bring $30,000 of ETH, stake it to stETH at ETH = $3,000, and build a 3x loop: $90,000 of collateral against $60,000 of stablecoin debt, a 66.7% effective LTV, and an 80% liquidation threshold. The starting health factor is $90,000 × 0.80 ÷ $60,000 = 1.20. Now walk the price down.
| ETH price | Change | Collateral value | Health factor |
|---|---|---|---|
| $3,000 | entry | $90,000 | 1.20 |
| $2,850 | −5% | $85,500 | 1.14 |
| $2,700 | −10% | $81,000 | 1.08 |
| $2,550 | −15% | $76,500 | 1.02 |
| $2,500 | −16.7% | $75,000 | 1.00 (liquidation line) |
| $2,400 | −20% | $72,000 | 0.96 |
| $2,340 | −22% | $70,200 | 0.94 |
The position is already liquidatable at a 16.7% drop, and by the −22% candle that opens this article it sits at HF 0.94 — the aftermath, not the warning. Two honest caveats make this worse, not better. If you borrow ETH rather than a stablecoin, ETH/USD moves cancel out of numerator and denominator, so your real trigger becomes the stETH/ETH rate and the borrow rate rather than the ETH chart. And every swap in the build costs a spread, so the collateral you deposit is slightly smaller than the model.
E-mode and isolated markets rewrite every parameter
The numbers above use a conservative 80% liquidation threshold. Efficiency mode changes them by design. Aave's ETH-correlated e-mode runs roughly 90–93% LTV with a 93–95% liquidation threshold. At a 95% threshold the same 3x loop opens near HF 1.42 with roughly a 30% buffer instead of 16.7%. At the same size e-mode is genuinely safer — but it also raises the leverage ceiling to about 14x, so it is dangerous only in the wrong hands. In April 2026, attacker accounts on Aave V3 looped rsETH in e-mode at up to 93% LTV and sat at health factors of 1.01 to 1.03; one collateral reprice was all it took.
Isolated markets invert the logic. Morpho Blue uses one immutable parameter per market — the LLTV — as both the borrow cap and the liquidation threshold, with no shared pool to spread losses through. That is safer for contagion: a bad market fails alone. But because LLTV is a single number, borrowing to the cap leaves HF at 1.00 by construction, so on a 65% LLTV market you must build well under the cap, to 50% or below, just to reach HF 1.30, and shallow liquidity plus curator risk shift onto you. The venue you pick — Aave, Morpho or Spark — is a parameter choice, not just a brand choice.
Your borrow rate can liquidate you with prices completely flat
Price is only the loudest way to lose a loop. Rates are set by utilization, and above a pool's kink the curve bends five to ten times steeper, so debt can compound faster than collateral earns. A position at HF 1.1 can be pushed under 1.0 by interest alone over days, with the chart doing nothing. April 2026 showed the shape: Aave's WETH available liquidity fell from $689 million to $1.5 million in two hours, utilization pinned at 100%, and reported WETH borrow rates moved from about 2.3% to roughly 8.7% APR. At 8.7% on a $60,000 debt that is about $14 a day of pure HF leakage — trivial at HF 1.8, decisive near 1.1. Before you borrow, check the current utilization, the distance to the kink, and the rate at 100%; the full curve is in utilization and the kink.
The de-risking ladder: act at 1.8, 1.5, 1.3 and 1.1
Because HF binds price and rate together, you monitor and act on HF rather than on price. Here is the ladder, with the price drop each level can still absorb at an 80% threshold. The uncomfortable truth: a freshly built 3x sits at 1.20, already below the 1.3 action line, which is why serious loopers start lower.
| Health factor | Zone | Price drop left before HF 1.0 | What to do |
|---|---|---|---|
| 1.8 | Green | −44% | Hold; optional top-up only if the thesis is intact. |
| 1.5 | Monitor | −33% | Daily checks; alerts on HF, price and the borrow rate. |
| 1.3 | Warning | −23% | Stage repayment or added collateral, on the same chain, in calm markets. |
| 1.1 | Act now | −9% | Repay or top up immediately; bots are already watching the address. |
Two rules make the ladder real rather than decorative. Keep rescue funds on the chain where the position lives — an L2 withdrawal can take days, and a top-up on L1 is useless mid-liquidation. And de-risk in calm markets: partial repayment at HF 1.5 costs nothing, while at HF 1.1 with high utilization the transaction may not land. Run the cleanup early with the yield calculator.
Six mistakes that turn a dip into a forced sale
- Confusing LTV with the liquidation threshold. Borrowing to the LTV cap pockets you just above the cliff; the gap is not a margin, it is the protocol's noise allowance.
- Looping to the ceiling. 4x leaves a 6.25% buffer — smaller than a known historical depeg. The maximum the UI allows is not the maximum you should take.
- Correlated collateral and debt with no peg awareness. Borrowing ETH against stETH is neutral to ETH/USD but fully exposed to the stETH/ETH rate; ignoring the peg makes it two bets, not one.
- Ignoring the borrow rate. A cheap variable loan near the kink is a spike waiting to happen, and interest alone can push a low-HF position under 1.0.
- Treating a high e-mode LTV as free headroom. A 93% LTV is efficient and lethal in equal measure when the collateral misbehaves.
- Holding rescue capital on the wrong chain, or none at all. The moment you need it is the moment it is least accessible.
Sources and further reading
- Aave documentation — the official health-factor definition, LTV and liquidation-threshold parameters, and efficiency-mode configuration.
- Morpho docs — Liquidation — how isolated markets use a single immutable LLTV as both borrow cap and liquidation threshold.
- ethereum.org — Decentralized Finance — the overcollateralization, oracle and lending mechanics underneath every loop.
- L2BEAT — bridge and rollup risk data, relevant when rescue capital sits on a different chain than your collateral.
Frequently asked questions
What is the health factor formula for a DeFi loan?
Health factor = the sum of each collateral's value times its liquidation threshold, divided by total debt including accrued interest. In short, HF = (collateral value × liquidation threshold) ÷ debt. Above 1.0 the position is safe; at or below 1.0 anyone can liquidate it. Borrow interest lowers HF even when prices are flat, because it grows the denominator every block.
Is LTV the same as the liquidation threshold?
No. LTV is the maximum you may borrow when you open the position, for example 65% or 75%. The liquidation threshold is the higher ratio at which the protocol allows liquidators to seize collateral, for example 80%. The health factor uses the liquidation threshold, never the LTV, so the gap between the two is your entire cushion. Borrowing at the LTV cap parks you just above the cliff.
How far can ETH fall before a 3x loop is liquidated?
At an 80% liquidation threshold, a full 3x loop runs at a 66.7% effective LTV and its collateral price can fall about 16.7% before HF reaches 1.0. Building the same loop to a 65% LTV widens the buffer to 18.8%. A 4x loop leaves only 6.25%, which is thinner than the roughly 6–7% discount stETH traded at during the June 2022 depeg. Liquidation price equals entry price × (debt ÷ collateral) ÷ liquidation threshold.
Does e-mode make leveraged looping safer?
E-mode raises both the LTV and the liquidation threshold for ETH-correlated assets, which is exactly what makes deep loops possible. ETH-correlated e-mode has run around 90–93% LTV and 93–95% liquidation threshold. At a 95% threshold the same 3x loop starts near HF 1.42 with roughly a 30% price buffer instead of 16.7%, so at the same size it is safer. It is more dangerous only because it also permits far higher leverage, up to about 14x at a 93% LTV, so the temptation to oversize grows.
Can a borrow-rate spike liquidate me if the price does not move?
Yes. Borrow rates are set by utilization, and above the pool's kink they jump five to ten times steeper, so debt can compound far faster than the collateral earns. During the April 18, 2026 rsETH incident, Aave's WETH available liquidity collapsed from $689 million to $1.5 million in two hours, utilization hit 100%, and reported WETH borrow rates moved from roughly 2.3% to about 8.7% APR. A position already at low health factor can be pushed under 1.0 by interest alone, with prices completely flat.
Size your loop before you open it
Model leverage, net APY and your health factor at each price level — then decide the de-risking ladder you will actually run.
Open the DeFi Yield CalculatorRelated: build the loop safely with liquid staking looping; understand the chain reaction in DeFi liquidation cascades explained; see how rates and utilization move in the utilization kink; and read the borrower's field notes in Aave, Morpho and Spark borrowing. Keep the Aave parameters and the yield calculator handy.