Crypto collateral haircuts reduce the value a risk engine recognizes from an eligible asset. Start with the asset quantity, apply the engine's valuation price and currency conversion, then apply the documented recognition ratios to the correct tiers. Recognized collateral is distinct from market value, borrowing capacity and a liquidation threshold; multiplying every displayed percentage together can count the same restriction twice.
Identify the value being reported
A haircut is a valuation adjustment. With a flat 10% haircut, an eligible asset worth $1,000 contributes $900 of recognized collateral under that simple rule. The holder still owns the original quantity. No $100 payment, sale or withdrawal has occurred merely because the risk engine recognizes less value.
A recognition ratio states the other side of the same adjustment: 90% recognition corresponds to a 10% haircut. Read the label carefully. A displayed “90% collateral ratio” can mean 90% recognized, whereas a 90% haircut would leave only 10%. The percentage needs a definition before it needs a calculator.
Keep these amounts in separate columns:
| Amount | Meaning in the worksheet | What it does not establish |
|---|---|---|
| Asset quantity | Units actually held | Whether the asset is eligible collateral |
| Reference market value | Quantity translated into the risk currency | Proceeds available from selling the whole balance |
| Recognized collateral | Value admitted under the collateral rules | Maximum borrowing or permission to withdraw |
| Borrowing capacity | Amount allowed under the applicable borrowing rules | A permanent limit after prices change |
| Liquidation measure | Account state under the specified liquidation rules | A guaranteed execution price or final loss |
Bybit's Unified Trading Account documentation provides one implementation: collateral is valued in USD, with its USD index price incorporating a USDT conversion rate. It then applies collateral value ratios. This makes the price conversion and the recognition discount separate steps in that account's methodology. Bybit's UTA collateral valuation guide.
A price quoted in USDT is not automatically the same number in USD. Conversely, applying another conversion when the input is already a USD valuation price repeats a step. Record the units on both sides of every conversion.
Reconstruct a tiered balance before stressing it
The following example is hypothetical. It uses a fictional collateral token X and quote token Q, with no borrowing, positions, orders, fees or accrued interest. All 120 X are eligible. The prices and ratios are invented to demonstrate the accounting, not describe a current product.
Assume X is valued at 100 Q per token, and each Q converts to $0.98. The reference dollar price is therefore $98 per X. Market value is 120 × $98 = $11,760.
The fictional engine recognizes the first 100 X at 90% and the next 20 X at 75%. These are marginal quantity tiers: each ratio applies only to its slice. The recognized quantity equivalent is 100 × 0.90 + 20 × 0.75 = 105 X. Multiplying 105 by $98 gives $10,290 of recognized collateral.
The balance has an effective haircut of 12.5%, calculated as one minus $10,290 divided by $11,760. That blended number summarizes this balance under these rules; it is neither tier's individual haircut.
Applying the final tier's 75% recognition to all $11,760 would produce $8,820. That would understate the correct amount by $1,470. Applying the first tier's 90% to the whole balance would also be wrong, in the other direction.
Bybit's UTA guide illustrates separate quantity slices, and its official tiered-collateral API reference labels the tier boundaries in coin units. These documents support checking the tier denominator; they do not make this fictional schedule a Bybit schedule. Bybit's tier boundary schema.
Attribute the decline to price, conversion and rule changes
Now change one input at a time while keeping the 120-token holding fixed. First the X price falls from 100 Q to 90 Q. Then Q's USD conversion falls from $0.98 to $0.96. Finally, the recognition ratios change to 85% for the first 100 X and 65% for the remaining 20 X.
The final rule change reduces the recognized quantity equivalent to 100 × 0.85 + 20 × 0.65 = 98 X.
| Snapshot | Q per X | USD per Q | Recognized X equivalent | Market value | Recognized value |
|---|---|---|---|---|---|
| Initial | 100 | $0.98 | 105 | $11,760 | $10,290 |
| Asset price falls | 90 | $0.98 | 105 | $10,584 | $9,261 |
| Conversion then falls | 90 | $0.96 | 105 | $10,368 | $9,072 |
| Recognition ratios then fall | 90 | $0.96 | 98 | $10,368 | $8,467.20 |
The asset-price step removes $1,029 of recognized support. The conversion step removes another $189. The rule change removes $604.80. Together they reconcile the $1,822.80 decline from $10,290 to $8,467.20.
Market value does not change in the last row. The effective haircut nevertheless increases from 12.5% to approximately 18.33%. That is a change in how much value the fictional engine admits, not another token-price loss.
This attribution depends on the stated sequence. Applying shocks in a different order can allocate their interaction differently, even when the final recognized value is the same. Retain the sequence beside the result instead of describing each contribution as a universal sensitivity.
Collateral rules can change independently of price: Bybit says it reviews and updates its collateral value ratios with market conditions. A saved ratio therefore needs an observation time. Bybit's collateral-ratio update notice.
The stablecoin collateral stress worksheet extends the conversion problem to depeg and access risks. The calculation here isolates valuation; it does not establish that either asset can be converted promptly at the reference price.
Quantity tiers and dollar tiers can diverge
A tier boundary is part of the formula. Bybit's separate Crypto Loans documentation describes tier amounts as portions of an asset's USD value, rather than the coin-quantity boundaries described above for UTA. Even within one provider, the product matters. Bybit's Crypto Loans tier methodology.
To see the consequence, compare a second hypothetical engine with the same starting recognition. It recognizes the first $9,800 of X's market value at 90% and the remaining value at 75%. Its dollar boundary stays fixed. Initially, this also produces $9,800 × 0.90 + $1,960 × 0.75 = $10,290.
After the price and conversion moves, but before any rule change, market value is $10,368. The dollar-tier engine recognizes $9,800 × 0.90 + $568 × 0.75 = $9,246. The quantity-tier engine recognizes only $9,072 at that same snapshot.
The $174 difference arises because more of the smaller dollar balance falls inside the higher-recognition dollar tier. The quantity-tier engine still puts exactly 100 X in its first tier. Two models that agree on day one can disagree after a move without either calculation being internally inconsistent.
A spreadsheet that permanently stores the starting 87.5% blended recognition would reproduce the quantity-tier result here. It would miss the dollar-tier result. Rebuild the slices whenever the input controlling the tiers changes.
Check eligibility and percentage denominators
Eligibility comes before valuation. An asset in a wallet can have market value while contributing no support to the account being examined. Bybit's API documentation distinguishes platform collateral eligibility from the user's collateral switch. It also marks its former single collateral-ratio field deprecated following the introduction of tiered logic. A parser can retrieve a number successfully and still use the wrong field. Bybit's collateral information reference.
Next, identify whether borrowing limits operate on raw collateral value or an already adjusted value. Both arrangements are possible; the product's formula decides.
For a hypothetical rule that explicitly combines 90% recognition with a 70% borrowing limit on recognized collateral, $10,000 of market value becomes $9,000 recognized and then $6,300 of borrowing capacity. Both multipliers belong because the fictional rules define separate layers.
For a different rule that directly permits borrowing 70% of $10,000 collateral value, capacity is $7,000. Recasting that same restriction as a 30% haircut and then applying 70% again gives an incorrect $4,900. The error is counting one restriction twice.
These are total modeled debt limits, not additional amounts available to an account that already owes money. Existing debt and other product constraints need their own check.
Aave V3's reserve documentation defines LTV as maximum borrowing relative to collateral value and describes a separate liquidation-threshold parameter. It also explains that parameters differ between markets and change over time. Do not import an exchange's additional haircut into that framework without evidence that the relevant implementation includes it. Aave V3 reserve parameters.
The LTV versus liquidation-threshold ledger examines those distinct limits. A valuation haircut also differs from collateral actually transferred during liquidation; the liquidation bonus and penalty accounting guide follows that later event.
Make the valuation reproducible
Use one record for the precise account, product and observation time. A practical worksheet needs the following fields:
| Field | Evidence to retain |
|---|---|
| Asset identity and eligibility | Token, chain or venue asset code, account mode and collateral status |
| Quantity | Positive balance admitted to the calculation; separate liabilities |
| Price | Risk engine's reference source, timestamp and quote currency |
| Conversion | Quote-to-risk-currency rate, source and direction |
| Tiers | Boundaries, units, marginal versus whole-balance treatment and ratios |
| Result | Market value, recognized value and blended haircut, separately labeled |
| Further rules | Borrowing denominator, liquidation measure and applicable obligations |
Reconcile this worksheet with the platform's reported collateral before using it for another decision. Investigate a difference by checking the asset identity, price timestamp, conversion direction, eligible balance and tier slices in that order. Keep full precision through the calculation and apply the platform's documented rounding convention at the appropriate step.
Keep liabilities outside the positive-asset shortcut. For example, Bybit's UTA guide applies collateral ratios to positive balances while negative balances receive 100% treatment. Discounting an obligation as though it were collateral would understate what the account owes. Bybit's positive and negative balance treatment.
A correct valuation can still belong to the wrong account. The collateral fragmentation guide explains why support recognized elsewhere cannot automatically satisfy a local obligation. Likewise, a recognized dollar is not necessarily a withdrawable dollar or an executable sale dollar.
Before relying on a collateral figure, reproduce it from the actual units and dated rules, then reconcile the next borrowing or liquidation calculation using its own denominator. The useful output is a traceable amount of recognized support, with every discount applied once.