Crypto Leverage Risk: Can a Correct Trade Still Be Liquidated?

How much margin and leverage should I use for a fixed loss budget?

Choose the maximum acceptable whole-account loss and invalidation distance before choosing leverage. For a linear contract, a rough pre-cost position notional is loss budget divided by adverse price fraction; required initial margin is notional divided by leverage. Raising leverage does not make the same notional safer.

Illustrative sizing only: a $10,000 account with a $100 price-loss budget and a 2% adverse move implies $5,000 notional before fees, funding and worse fills. That notional uses $1,000 initial margin at 5x or $500 at 10x. Its dollar price loss for the same filled quantity is unchanged; the higher-leverage version has less collateral to survive the path. Account exposure is not the margin number shown beside the order.

Reserve costs inside the loss budget and compare the stop with the actual mark-based liquidation boundary and maintenance tier. If the proposed stop is beyond that boundary, or execution uncertainty consumes the budget, reduce notional or decline the trade; this page cannot certify a safe setting. Minimum order size and venue rules must also be checked.

Verify: Calculate linear-contract profit and cost denominators

Does a stop loss guarantee I will avoid liquidation?

No. A stop and a liquidation engine can use different price references, and a triggered stop is not a guaranteed fill. This historical margin test checks hourly adverse mark extremes; it does not simulate a queue, stop execution, protective order or second-by-second venue behavior.

Compare your stop’s trigger reference with the venue’s liquidation reference, whether the order reduces the intended position, and the available depth at expected size. A mark-price move can breach maintenance while a last-price chart appears above a stop. Gaps, latency, outages or partial fills can leave exposure after the trigger.

Confirm the protective order was actually accepted in the venue account—not just entered in a local app—and monitor its quantity when adding or reducing positions. If account state, price reference or order status is unknown, the protection cannot be treated as verified. Do not add collateral solely to keep a failed thesis alive.

Verify: See the historical adverse mark-price path evidence

Which frozen price and funding dates does the leverage study cover?

The price-only exhibit is frozen at September 19, 2026 UTC, while the complete funding-inclusive experiment stops in August 2026. Neither is a live liquidation quote. The table below preserves the original price observation; the October 3 content update does not extend its input range or fill missing settlements.

The September 20 edition could not obtain the September monthly or requested daily funding file, and the public futures funding REST endpoint was unavailable from this server location during that recorded check. That evidence explains why the frozen combined experiment stops in August; it does not establish that September settlements are still unavailable today. Current rates or zero cannot replace the actual settlement paths missing from this edition.

The newest complete observations within that frozen input range appear next. The explanatory breach charts were selected by latest eligible historical date, not earliest date or best return. September price-only observations do not convert the preserved August funding-inclusive sample into a September backtest.

Verify: Recent mark inputs and funding availability checks

Most recent completed cohorts in the verified funding sample

The newest complete cohort in the frozen funding sample starts July 27, 2026; all available BTC/ETH/SOL long and short rows are retained, including negative outcomes. A complete 30-day path needs its endpoint, so this is a historical comparison rather than the latest market quote.

Can I be right about Bitcoin and still be liquidated at 10x?

Yes—you can be right about the eventual price direction and still run out of margin first. In our 10x, 30-day BTC/ETH/SOL replay, 19/129 (14.7%) directionally correct observations crossed the modelled margin boundary before the finish. Across both directions, 128/258 (49.6%) crossed it. These are a fixed-assumption margin stress test on historical mark prices, not real exchange liquidation records or a forecast of your liquidation probability.

This is not a test of a profitable entry signal. At each prescribed Monday open we examine both a hypothetical long and a hypothetical short. Their eventual directions are complementary; counting them as independent trading opportunities would be misleading.

We use historical Binance mark-price highs and lows, not spot candles. The maintenance-margin rate is a fixed research assumption, not a reconstruction of Binance’s changing historical risk tiers. Exchange-specific account balances, positions, fees and liquidation engines can produce different outcomes.

Why “10x means I can survive a 10% move” is only a rough shortcut

Leverage relates position notional to initial margin. The account must still retain maintenance margin and allowances for closing; entry fees and funding reduce or increase its available balance. That is why a simple reciprocal of leverage is not your exact liquidation distance.

For the generic model, position quantity q=1/entry price and initial wallet W=1/leverage−entry cost. At mark M, equity=W+direction×q×(M−entry). The risk boundary is reached when equity is no greater than (maintenance rate+closing reserve rate)×q×M. Funding changes W during the path.

With a 0.5% maintenance assumption and 0.05% closing reserve, the model boundaries shown below already differ from 1/leverage before any funding. They are illustrative account equations, not tradable stop prices.

Why can the chart look safe while the position is liquidated?

Your displayed chart may use the last traded price, while the venue monitors a mark-price reference. Those are not the same series. A stop can also use a different trigger reference from liquidation; a last-price stop may not activate before the mark-based risk limit.

A historical hourly low or high tells us whether a model threshold was crossed inside the hour, not the exact second, fill price or queue position. We report the first affected hour. We do not reconstruct slippage, insurance-fund settlement or auto-deleveraging.

Cross-margin is a different account problem: other positions and collateral change available equity. We model one isolated position with no rescue deposits, not a cross-margin wallet. Review your own venue’s position panel and risk rules rather than using this page as a live liquidation calculator.

Verify: Official example: Bybit margin and mark-price calculations

BTCUSDT long: the eventual direction did not save the position

Entry 2026-03-16T00:00:00Z; the first model breach occurred in the hour starting 2026-03-27T14:00:00Z. The terminal futures-price move was +1.82%—favourable for this long. This is the most recent eligible entry for this symbol among 10x/30-day correct-endpoint cases that breached; ties use observation ID.

The top panel plots hourly mark closes and the model boundary. A wick can cross even when the plotted close does not; the replay checks the full hourly high/low. The middle and bottom panels continue the account equation after the first breach only to explain the counterfactual. They do not show money an actually closed position could recover.

ETHUSDT long: the eventual direction did not save the position

Entry 2026-03-16T00:00:00Z; the first model breach occurred in the hour starting 2026-03-27T13:00:00Z. The terminal futures-price move was +6.59%—favourable for this long. This is the most recent eligible entry for this symbol among 10x/30-day correct-endpoint cases that breached; ties use observation ID.

The top panel plots hourly mark closes and the model boundary. A wick can cross even when the plotted close does not; the replay checks the full hourly high/low. The middle and bottom panels continue the account equation after the first breach only to explain the counterfactual. They do not show money an actually closed position could recover.

SOLUSDT short: the eventual direction did not save the position

Entry 2026-05-04T00:00:00Z; the first model breach occurred in the hour starting 2026-05-08T17:00:00Z. The terminal futures-price move was -11.54%—favourable for this short. This is the most recent eligible entry for this symbol among 10x/30-day correct-endpoint cases that breached; ties use observation ID.

The top panel plots hourly mark closes and the model boundary. A wick can cross even when the plotted close does not; the replay checks the full hourly high/low. The middle and bottom panels continue the account equation after the first breach only to explain the counterfactual. They do not show money an actually closed position could recover.

Does the result depend on the maintenance-margin assumption?

Yes: higher maintenance requirements can move the margin boundary closer. The frozen sensitivity table changes only the maintenance rate to 0.4%, 0.5% and 1.0% in the same 30-day paths; it keeps symbols, starts and costs fixed. These are assumed rates, not verified historical venue tiers for your position.

A lower leverage setting provides more starting margin for the same notional, but a zero count in one sample is not proof that it is safe. Longer adverse moves, higher costs, another token or a different margin regime can change the outcome.

What should you check before opening a leveraged position?

Check the price reference used for both your stop and liquidation; margin mode; total notional rather than only deposited margin; maintenance tier; both-side fees; next funding settlement; and what other positions can consume collateral. Stress-test an adverse move before a favourable finish.

Size a potential loss against money you can afford to lose, rather than choosing a leverage number first. Reducing notional and adding margin are not equivalent to predicting direction better. This research does not recommend 2x, 5x or 10x.

CoinNudge can provide supported market and leverage-context alerts. It does not monitor your exchange account, maintain your margin, submit stops or guarantee delivery before a liquidation.

Verify: Understand supported alerts · Current funding and cost context

Is isolated margin safer than cross margin—and can adding margin save a trade?

The isolated-versus-cross choice changes which collateral can support or lose money on a position, not whether the price direction is correct. This experiment models one isolated position with no rescue deposits. It does not reconstruct cross-margin netting, multiple positions or an exchange’s account-specific liquidation calculation.

Adding eligible margin at the same quantity can widen the modeled collateral buffer; increasing quantity while retaining the same leverage increases dollar exposure instead. Reducing a position changes both exposure and remaining margin. Funding, other positions, collateral value and risk tiers may change an actual boundary even without a new market order.

Before relying on a changed margin figure, verify the updated account equity, maintenance requirement, position quantity and displayed liquidation reference in the venue. Cross margin can expose collateral you intended to reserve for other trades; isolated margin still can lose the allocated capital. No margin mode or rescue action is established as profitable by these historical counts.

Are the 49.6% and 14.7% liquidation figures the same probability?

No. The 49.6% figure is 128/258 modeled boundary crossings across both directions at 10x over 30 days; 14.7% is 19/129 among the observations whose endpoint direction was favorable. They answer different descriptive questions and neither is a calibrated chance of liquidation for a live strategy.

The replay examines both long and short at prescribed Monday entries in BTC, ETH and SOL, rather than a validated entry signal. The paired directions and overlapping weeks share prices and shocks. A zero count at lower leverage in one subgroup does not establish safety, and continuing an account curve after breach is counterfactual, not realized recovery.

To reproduce a row, filter the observation CSV to horizon, leverage, direction and fixed maintenance rate before counting complete paths and endpoint-correct flags. Retain missing and pending candidates separately. If your count differs, check those fields and the funding boundary convention rather than assume the page predicts a different market.

Verify: All frozen leverage observations · Candidate paths and missing-hour exclusions

Methods, field definitions and reproducible evidence

BTC/ETH/SOL Binance USD-M mark-price and trade-price hourly archives, 2025-09-01 through 2026-08-31. Each symbol has 8,736 mark hours versus 8,760 trade hours; June 29, 2026 UTC is absent from the frozen monthly mark files. Any candidate touching missing data is excluded, not interpolated. The shared 7/30/90-day candidate ledger contains 48 missing and 57 pending windows; this article uses only 7/30-day rows.

Weekly Monday 00:00 entries, 7/30-day holds, both sides, 2/5/10x. At internal funding timestamps subtract direction×q×mark open×actual settlement rate; exclude settlements exactly at entry and exit. Then check hourly adverse mark extreme against equity and maintenance+closing reserve. The fixed 0.05% entry cost and closing reserve and 0.5% maintenance rate are modelling assumptions; actual historical tier reconstruction, partial liquidation and actual fills are not claimed.

breach_time is the first affected hour or null. endpoint_direction_correct compares the final trade-price open with entry without requiring survival. counterfactual_net_return continues the wallet equation even after breach and must not be read as achieved P&L. funding_paid_per_initial_notional is signed; receiving funding is negative. No probability forecast or inference from stop-loss spot data is used.

All displayed percentages are historical descriptions, not calibrated probabilities of your next trade. Starts overlap and assets share market shocks. Counts are not independent trials; no causal effect, statistical significance or execution guarantee is claimed. Missing and pending outcomes remain separate from zero. Inputs were acquired retrospectively, not preserved as real-time investment knowledge.

The complete files load only when you choose a download. Download spot.json.gz and derivatives.json into one folder, save four_decision_studies.py, then run: python four_decision_studies.py INPUT_FOLDER OUTPUT_FOLDER. Python standard library only; no exchange key or live trading is used. The input directory contains 591 discovered target markets, not 591 complete markets for every analysis. Per-market coverage and per-window eligibility are published.

JSON carries original float precision; CSV contains the same observation fields. UTC timestamps, missing values, cost scenarios and model variants must be retained when reproducing a table. Hashes protect this frozen edition against accidental changes; they do not prove economic completeness.

Verify: Full study JSON · Observation CSV · Summary CSV · Candidate and missing ledger · Daily source bundle · about 19 MB gzip · Hourly derivative inputs · about 5 MB · Per-market daily coverage · Reproduction program · Preset protocol · Hashes and source coverage · Chart coordinates

Frequently asked questions

Sources and verification

• Binance public-data documentationPrimary data format and checksum documentation; our simulation is not a Binance trading record.
• Binance spot market-data fieldsDaily OHLCV; candles are not execution, order-book or personal cost-basis evidence.

Use evidence without outsourcing your decision.

Inspect the market tools and see exactly which alerts CoinNudge delivers.