First compare the same instrument
DZPTO presents a teaching estimate for linear contracts. An exchange displays an estimate produced by its live risk engine for a specific account and position. Coin-margined versus stablecoin-margined, isolated versus cross, and one-way versus hedge mode are not directly interchangeable.
Confirm that the comparison uses mark price rather than last trade. Liquidation engines generally reference mark price, which can differ from the latest candle or fill shown elsewhere in the interface.
Maintenance margin is not always 0.5%
Larger positions can enter a higher notional bracket, changing both maintenance requirements and permitted leverage. A single DZPTO rate reproduces one assumption; it cannot infer an authenticated account bracket.
Binance's notional and leverage bracket endpoint is an account API. A public site cannot know your tier without credentials, and DZPTO deliberately does not request or store exchange API keys.
Where the differences come from
Once 'the numbers do not match' is broken into a checklist, most gaps land in one of the six rows below. The left column is the cause, the middle says which way it pushes the estimate, and the right says which screen settles it.
The order matters. Rule out the cases where the two sides are not describing the same position at all (the first two rows), then parameter differences, then account state. Working backwards wastes time tuning parameters before noticing the comparison was never like for like.
| Source of difference | Effect on the estimate | Where to check |
|---|---|---|
| Contract type (USDⓈ-M vs coin-margined) | Different formula; not comparable | Settlement asset on the contract page |
| Margin mode (isolated vs cross) | Different definition of usable equity; cross is usually further away | Mode flag in the position list |
| Maintenance tier | Real r exceeds the smallest tier for large positions; liquidation is nearer | Venue risk-limit or tier table |
| Liquidation fee omitted | Slight underestimate, amplified at high leverage | Venue fee and liquidation documentation |
| Last traded price used instead of mark | Looks touched or untouched when the opposite is true | Switch the chart to mark price |
| Unrealised P&L and order holds | Changes usable equity; largest effect under cross margin | Available balance on the assets page |
A worked reconciliation
Suppose your calculation gives 45,248.87 and the exchange shows 44,910. The 338.87 USDT gap is about three quarters of a percentage point — at that magnitude it is normally a parameter difference rather than an error.
Working through the list reveals that the position's notional has crossed into the next tier, where the maintenance-margin rate is 0.65% rather than 0.50%. Changing r from 0.0055 to 0.0070 and recomputing moves the estimate towards the venue's figure. Whatever remains usually comes from how the platform handles the liquidation fee and mark price.
You do not need the two numbers to agree exactly. You do need to be able to say where each percentage point of difference comes from — and if you cannot, you have not yet found your real risk line.
- Confirm contract, direction, quantity and margin mode match before comparing anything else.
- Enter the maintenance-margin rate the venue displays instead of keeping the default.
- Verify you are reading the mark price, and note the mark-to-last spread at that moment.
- If more than a percentage point remains, treat the venue result as authoritative and keep a timestamped screenshot.
Account state changes the answer
Cross-margin equity, unrealized PnL, other positions, open-order locks, funding debits and margin transfers can all affect available equity. In isolated mode, the margin assigned to the position and later adjustments matter.
Liquidation fee allowances, staged position reduction and partial-liquidation rules also vary. Omitting them from a simplified model creates larger differences at high leverage or high notional values.
A five-step reconciliation
Record the exact contract and settlement asset, direction and quantity, average entry and mark price, margin mode and available equity, then maintenance bracket and liquidation fee.
Replace every default with the live account value. If a difference remains, use the exchange estimate and preserve its timestamp. Updated platform rules and live order previews supersede old articles and spreadsheets.
Official sources and calculation boundary
This article is about where differences come from, so the sources carry more weight than usual: OKX documents the trigger condition and Binance's endpoint documents how tiers move with notional value. Neither contains your account data — which is precisely why a public estimate and an account figure diverge.
Next checks in this series
Liquidation and margin risk
BTC at 5x, 10x, 20x and 50x: how liquidation risk changes
Reproducible scenario guide
Liquidation and margin risk
How adding margin changes a BTC long liquidation estimate
Reproducible scenario guide
Venue comparison
Binance versus OKX margin models: what must be aligned
Reproducible scenario guide
Frequently asked questions
Can DZPTO read my Binance or OKX account?
No. A public estimate should not require an API key. Enter the bracket and margin inputs displayed in your account manually.
Which liquidation price is authoritative?
The signed-in exchange position and risk information is operationally authoritative. DZPTO is a pre-trade estimate and cross-check.
How large a difference is normal?
There is no fixed threshold, but magnitude is informative. Under one percentage point usually reflects maintenance-tier or liquidation-fee handling. More than a few percentage points usually means the two sides are not describing the same thing — a different margin mode or contract type.
Why not connect to the exchange API and show the real liquidation price?
Reading account tiers and positions requires account credentials. Asking a visitor for exchange API keys so that a comparison looks more precise is a disproportionate risk. DZPTO therefore uses public data only and labels the result as an estimate.