Collateral Haircuts: The Math Behind Your Liquidation Price

A collateral haircut is the discount a lender slaps on whatever you deposit. It decides two things: how much you can borrow, and how far the price can fall before you're liquidated. If you borrow against crypto to fund a trading or gambling bankroll, this number matters more than the interest rate. Full stop.

Most people read the headline, something like "borrow up to 75%", and stop. Big mistake. The haircut is a safety margin built for the protocol, not for you, and it can change while your loan is open. Once you see how it's sized, you can work out your real buffer to the dollar.

What a haircut actually is

Say you deposit $10,000 of a token. The protocol won't give you $10,000 of borrowing power. It applies a haircut, say 25%, and counts your deposit as $7,500. That's the most you can borrow. Flip it around and the maximum loan-to-value (LTV) is 75%.

So: max LTV = 1 minus haircut. One idea, two names.

Most lending protocols run three related numbers, and mixing them up is how people get burned:

• Max LTV: the ceiling when you open or add to a loan. You can't borrow past it.
• Liquidation threshold (LT): a slightly higher ratio. If your debt divided by collateral value crosses it, your position can be liquidated.
• Liquidation bonus (or penalty): the discount a liquidator gets when they repay your debt and take your collateral. Guess whose pocket that comes out of. Yours.

The gap between max LTV and the liquidation threshold is a small cushion, so you aren't liquidated the second after borrowing at the limit. It's thin. Don't call it a safety margin.

Why protocols haircut collateral at all

Decentralized lenders don't make margin calls. There's no risk desk phoning you at 3am. When your position goes underwater, the only defence is a liquidator, usually a bot, who repays part of your debt, takes your collateral at a discount, then sells it.

That only works if the liquidator can make money. Plenty eats into their margin:

• Price gaps. Crypto doesn't fall smoothly. It can drop 15% in minutes and skip right past your liquidation price.
• Slippage. Dump a big block of a thinly traded token and you push the price down. Bigger position, worse fill.
• Oracle lag. The protocol reads prices from an oracle that updates on a schedule or when price moves past a set deviation. The market can move before the oracle does.
• Network congestion. In a crash, transaction fees spike and blocks fill up. Liquidations get delayed or become unprofitable to send.

If the collateral is worth less than the debt by the time anyone acts, nobody liquidates. The protocol is stuck with bad debt: loans that will never be repaid. Bad debt drains lenders' deposits, lenders panic and withdraw, and if that collateral token is used across many protocols, forced selling in one spills into others. That's the cascade the haircut is there to stop.

The buffer math the protocol is running

Here's the core equation. Sit with it, because it explains why haircuts are the size they are.

Liquidation triggers when debt equals collateral value times the liquidation threshold. For the liquidator to profit, the collateral they receive has to be worth at least the debt they repay times (1 + bonus). Liquidations stop being profitable once collateral value falls to debt times (1 + bonus).

Put those together and the price drop between the trigger point and the point where liquidation stops making sense is:

Protocol buffer = 1 minus LT × (1 + bonus)

With an 80% threshold and a 5% bonus, that's 1 minus 0.84, or 16%. The protocol is betting that once a position hits its trigger, liquidators can clear it before the price falls another 16%. If the token routinely gaps further than that, the parameters are too loose and bad debt follows.

That's why stable, deep, well-oracled assets get high thresholds and small haircuts, while small-cap or thinly traded tokens get brutal haircuts or can't be used as collateral at all.

How protocols calibrate haircuts

Risk teams, governance votes or automated systems weigh roughly the same inputs:

• Volatility. How far does the asset typically move in an hour or a day, and how ugly are its worst days?
• Liquidity depth. How much can be sold on-chain before slippage blows through the liquidation bonus? This often matters more than volatility.
• Concentration. If a few wallets hold most of the deposited collateral, one liquidation could flood the market.
• Oracle quality. How many sources feed the price, how often it updates, and how easy it is to manipulate.
• Correlation with the debt. Borrowing a liquid-staked token against its underlying asset is low risk because the two move together. Some protocols offer special high-LTV modes for correlated pairs.

The "dynamic" part shows up in a few ways. Governance can lower thresholds when on-chain liquidity dries up. Supply caps limit how much of a risky token can be deposited, so total liquidation volume stays within what the market can absorb. Isolation modes let a new asset serve as collateral only for limited borrowing. Some designs adjust parameters automatically as measured volatility or liquidity changes. Any of these can move your liquidation price without you touching your position.

Worked example (hypothetical numbers)

Every figure here is invented for illustration. Use your protocol's actual parameters.

You deposit 10 units of token XYZ, priced at $2,000. Collateral value: $20,000. The protocol sets max LTV at 75%, liquidation threshold at 80%, liquidation bonus at 5%, and lets liquidators repay up to 50% of the debt in one go.

You borrow $12,000 of a stablecoin. That's 60% LTV, comfortably under the 75% ceiling.

Health factor = (collateral value × LT) ÷ debt = (20,000 × 0.80) ÷ 12,000 = 1.33. Drop below 1.0 and you get liquidated.

Liquidation price = debt ÷ (units × LT) = 12,000 ÷ (10 × 0.80) = $1,500. XYZ can fall 25% before you're hit.

Now XYZ falls to $1,500. Collateral is worth $15,000. A liquidator repays $6,000 of your debt and receives $6,000 × 1.05 = $6,300 of XYZ, which is 4.2 units. You're left with 5.8 units ($8,700) and $6,000 of debt. Your health factor resets to (8,700 × 0.80) ÷ 6,000 = 1.16. You've lost $300 to the bonus, plus exposure to any rebound on those 4.2 units.

Now the part people miss. Suppose, before any crash, governance cuts XYZ's threshold from 80% to 70% because liquidity has thinned. Your liquidation price becomes 12,000 ÷ (10 × 0.70) = $1,714. Your buffer shrinks from 25% to about 14%. You did nothing. Nothing at all.

Look at the bottom row. Borrow at max LTV and you've got a 6% buffer under the original parameters. A threshold cut liquidates you on the spot.

A step-by-step walkthrough before you borrow

1. Find the three numbers. Look up max LTV, liquidation threshold and liquidation bonus for your exact collateral asset in the protocol's docs or market page. They differ by asset and sometimes by market.
2. Check for special modes. Note whether you're in isolation mode, a correlated-asset mode or a standard market. Each has different limits.
3. Pick a stress drop. Decide the largest price fall you want to survive without acting. For a volatile token, a 40-50% drop isn't extreme. Write the number down.
4. Back out your max safe loan. Safe debt = units × current price × (1 minus stress drop) × LT. For 10 XYZ at $2,000, a 45% stress and an 80% threshold: 10 × 2,000 × 0.55 × 0.80 = $8,800.
5. Cut it again for parameter risk. Re-run step 4 assuming the threshold drops 5-10 points. Borrow the lower figure.
6. Account for interest. Debt grows every block. Over months, accrued interest pushes your liquidation price up. Add the expected interest to your debt figure.
7. Keep repayment funds off-protocol. Hold enough stablecoins in a separate wallet to repay or top up quickly. Don't park them on an exchange you'd have to withdraw from mid-crash.
8. Set alerts on health factor, not price alone. Many dashboards and bots can ping you when health factor falls below a level you choose, say 1.5.
9. Watch governance. Follow the protocol's governance forum or risk updates for proposed changes to your collateral's parameters. Changes are often announced before they take effect.

Common mistakes

• Borrowing to max LTV. Max LTV is a ceiling, not a target. It assumes you'll manage the position actively, which most people don't.
• Mixing up max LTV and liquidation threshold. Your liquidation price uses the threshold. Plug max LTV into your math and you get a false number.
• Borrowing a volatile asset. If you borrow a token that rises while your collateral falls, you get squeezed from both sides. Stablecoin debt keeps one side fixed, unless the stablecoin depegs upward against your collateral.
• Watching exchange prices instead of the oracle. The protocol liquidates on its oracle price. A wick on one exchange may trigger nothing; a stale oracle may fire later than you'd expect, or catch up all at once.
• Looping. Depositing, borrowing, buying more collateral and depositing again multiplies your exposure. Each loop pushes your liquidation price closer. A 10% move can wipe out a heavily looped position.
• Assuming partial liquidation is harmless. Each round costs you the bonus. A falling market can trigger several rounds in a row.
• Planning to top up during the crash. That's exactly when fees spike and transactions stall. Act at a health factor where you still have hours, not minutes.

FAQ

Why did my liquidation price move when I didn't do anything?

Three usual causes. Interest has added to your debt. The protocol changed your collateral's liquidation threshold. Or, if your debt isn't a stablecoin, the borrowed asset's price rose. Check all three on your position page.

Is the haircut the same thing as the liquidation penalty?

No. The haircut decides how much you can borrow and when liquidation starts. The penalty, or bonus, is what you lose when it happens. A protocol can have a small haircut and a large penalty, or the reverse, and both hit your real cost of being wrong.

Is a lower LTV always safer?

Lower LTV gives you more room before liquidation, yes. But you're still exposed to oracle failures, smart contract bugs and stablecoin depegs that no LTV choice protects against. Low LTV handles price risk. It doesn't handle protocol risk.

The haircut tells you what the protocol thinks your collateral is worth in a fire sale. If your own stress number is gentler than theirs, ask yourself why you think you know better than the people who'd eat the bad debt.