Start at 3-5x. Reading 100x as "tougher leverage" gets it backwards — what it actually means is "BTC moves 1% and your margin is gone". At 3x BTC has to run about 30% against you before liquidation, at 5x about 18%, and either of those leaves room to make mistakes and learn from them. Leave 10x and above alone until you can say out loud what each individual trade is betting on.
Attached, always. You sleep, work, take meetings — the market does not wait. The 2024-08-05 flash crash hit during Asian midday; the "I will close it by hand" crowd woke up already liquidated.
Isolated. Your worst case on isolated is the margin you put on that position. Worst case on cross is your full account balance. Beginners should not touch cross.
It can. Funding is charged on position value, not on your margin: +0.05% once, three settlements a day, three days running comes to 0.45% of position value; to convert that to margin you multiply by your actual leverage, so at 5x those three days cost about 2.25% of margin. Settlement frequency also varies by coin — every 8, 4, 2 or 1 hours — with BTC and the other majors usually on the 8-hour cycle. Check the next settlement time shown on the contract page before you open.
Crypto trades 24/7, so strictly speaking there is no such thing as a gap. But the order book is at its thinnest in the small hours of Saturday and Sunday, and one large order there can knock the price down a stretch and let it bounce straight back. Carry full size at high leverage through that window and a move like this is enough to punch through your liquidation price — at an hour when you are most likely asleep.
1. Mistake one: going cross-margin all-in
The setup: you saw a clean breakout. You felt confident. You moved to OKX, switched to cross margin (because "more capital backing the trade"), and put 80-100% of your account into one position at 10x leverage.
The break: cross margin uses every USDT in your account as collateral for that one position. A 5% adverse move at 10x wipes 50% of your account. A 10% adverse move at 10x wipes 100%.
The fix: isolated margin for the first three months of live trading. Wall off each position. Even if you misread direction, the worst case is the wall, not the whole account.
For the textbook framing of liquidation mechanics, Investopedia's liquidation entry walks through the engine logic in its general form.
2. Mistake two: no stop, "I'll close it manually"
The setup: you opened a 10x long without setting an SL on the ticket. Your plan was to watch the trade and close if it moved against you.
The break: a 10x position liquidates around an 8.5% adverse move. BTC routinely moves that much in 6-12 hours. If the move happens overnight, you wake up to zero.
The fix: every order has SL/TP attached at submission, not as an afterthought. Set the SL at a sensible 1-2R distance (matching the position-size paradox piece). If the chart moves against you, the order goes off automatically while you sleep.
The argument against stops — "I'll close it manually so I don't get stopped on noise" — only works if you are watching the chart 24/7. Almost nobody can do that. The stop is your insurance policy against your own absence.
3. Mistake three: the wrong leverage — 50x is not aggressive, it is suicidal
The math (covered in detail in the leverage / liq price piece) is really just the inverse of the leverage, pulled in a little further by the maintenance margin. 3x leaves you about 33% of room before liquidation. 5x leaves about 18%. 10x leaves 8-9%. 50x leaves about 1.7%. 100x leaves under 1%.
"1.7% of room" sounds like a number until you put it next to a BTC chart: that is an ordinary intraday range, reachable on a quiet day with no news attached. At 100x even the opening taker fee is meaningful relative to what is left between you and the liquidation engine.
The fix: fix the leverage at 3x to start with, and stop flipping between 3x and 20x to see what happens. Whether you move it up later is a question about your risk budget, not about how many days or how many trades you have behind you. Answer two things first: at this leverage, how much margin does an adverse move you consider "unlikely" take out — and if that money were gone, would your life or your sleep change? If you cannot answer, leave the dial alone. Win rate and trade count are not qualifications; seven wins out of ten can just mean the tape suited you, and one liquidation erases all of it.
Make it concrete with a hypothetical. A 300 USDT account, first perp trade ever, copies a KOL screenshot captioned "100x BTC long, all-in" — BTC quoted at $98,420, isolated margin, 100x, 30 contracts long. Lay out the three formulas first so contracts and coins do not end up in the same bucket: one BTC-USDT perp contract has a face value of 0.01 BTC, so 30 contracts = 0.30 BTC; contracts × face value × price = notional, 30 × 0.01 × 98,420 ≈ $29,526; notional ÷ leverage = initial margin, 29,526 ÷ 100 ≈ 295 USDT — which is basically the whole 300 USDT account. None of the figures below come from a real fill; they are here to carry the arithmetic.
At 100x the liquidation buffer is under 1%: the notional is a hundred times the margin, so a 1% adverse move produces an unrealised loss equal to the entire margin, and the engine takes over before you get all the way there, because maintenance margin and fees come out first. One percent below $98,420 is $97,436, and the actual liquidation sits a little above that — and BTC covering that distance in a few minutes is an ordinary occurrence that needs no news to trigger it. On the books it reads like this: price down 1%, notional loss ≈ $295, exactly the 295 USDT you posted; down 0.5% and half the margin is already gone. Opening to zero can be a matter of a quarter of an hour, and there is nothing to do in between — with 295 of 300 USDT committed, the free balance starts near zero, so there is no topping up and no room to move.
The honest description of 100x: not meaner leverage, but a standing instruction to the exchange to settle your account the moment BTC moves 1%.
That is the reasoning behind the repeated "leverage ≤ 5x" line. The only sensible use of 100x is "short-hold + tight stop + tiny size", never a beginner's all-in. If you are just starting perp and have under 500 USDT, lock the max at 5x and do not touch anything above 10x. That does not guarantee you will not be liquidated; it lowers the odds that an ordinary adverse move alone punches through your margin, which is what buys you room to make mistakes and catch up on the theory.
4. Mistake four: not budgeting funding across settlements
The setup: you open a long and ignore the funding countdown. The next settlement fires while you are still in, and the position is charged position value × funding rate — where a USDT-margined position value is face value × contracts × mark price. Note the settlement interval is set per coin, anywhere from every 8 hours down to hourly, with BTC and the other majors usually on 8 hours. Rather than memorising a timetable, read the "next settlement" clock on the contract page.
The break: if funding is +0.05%/8h, a 10,000 USDT notional position pays 5 USDT per settlement. On an 8-hour cycle that is three a day, so 30 days at that rate is 30 × 3 × 5 = 450 USDT. The trap is reading that against the wrong base: it is 4.5% of the position value but, on 1,000 USDT of margin behind it, 45% of your margin. Those two numbers are a leverage multiple apart.
The fix: before holding a position across any settlement, check the funding rate. If it is high (>0.02%/8h against you), recalculate whether the trade still pencils. The full math is in the funding rate piece.
5. Mistake five: holding leveraged into the weekend
The setup: you carry a 10x position into Friday US close, planning to hold it through to Monday because you have a directional view.
The break: weekend liquidity on crypto perps is thin, and the small hours of Saturday and Sunday are the thinnest window of all — the US desks have gone to bed and Asia is not up yet. A large order landing there moves price further than the same order would during an active session, and it can come back up just as fast. On 10x leverage a wick like that can take out your stop or your liquidation price even though the candle closes back where it started.
The shape of a thin-book drop is not hard to find, and it is not exclusive to weekends: on 2024-12-09, BTC fell from around $98,400 to around $94,200 in a short span, a little over 4%, with market-wide liquidations spiking to the high end of the range for that period. On the way down, high-leverage liquidation prices are simply taken out, and anyone away from the screen has no chance to react. The weekend is not a different mechanism — it is the same mechanism with a thinner book and a higher chance you are asleep for it.
The fix: enter the weekend with no leverage above 3x, or flat. If you have a strong view that requires weekend exposure, take it on spot, not on a leveraged perp.
6. The four-number checklist before every order
Before you click Buy or Sell, type the answer to four questions into a notes app. If any answer is missing, do not place the trade.
Leverage. What multiplier am I using? (Answer should be 3-10x.)
Liq price. What is the actual liquidation price for this position? OKX prints it directly on the order ticket. Is the distance to liq sensible relative to current volatility?
Stop price. Where is my SL set? Is the distance to stop a reasonable percentage of margin (0.5-2%)?
Max loss in USDT. If the stop hits, how much USDT do I lose? Is that a number I can lose without anger?
Four numbers. Thirty seconds. The number of rookie blowups this prevents is almost embarrassing.
7. An optional way in (better adapted than copied)
What follows is one possible sequence, not a calendar to tick off. It is written as steps to make the order clear — take twice as long over it, half as long, or stop somewhere in the middle, and it still holds. The whole run can also be done on the demo account: if you would rather not put real money down, swap the live step for more demo and you can still get to the end of it.
Step one. Read this piece, the leverage piece and the funding piece. If you are going to fund the account, keep the amount inside what you could watch go to zero — start small rather than pushing a large sum in on day one.
Step two. Set every contract setting to isolated margin with a 3x default. Open the "demo trading" mode in the OKX app.
Step three. Run a dozen or so BTC trades on demo. Log every one: entry, stop, position value, leverage, exit, PnL in USDT. The point is not what you made — it is whether each trade went through the four-number checklist.
Step four (optional). Small live size. Position value at or under 100 USDT, leverage 3x. This step exists for exactly one reason: to get used to the emotional pressure of live trading, which demo cannot reproduce. If you are not ready for live yet, keep building the demo log instead — skipping this does not break what comes after.
Step five. Take the position value up to around 200 USDT, leverage still 3x. Keep a journal per trade: why did I open it, what was I expecting, what actually happened, what do I take from it.
Whether you then go to 5x, or up to 500 USDT of position value, is not a question about a win-rate number. It is whether you can spread your own log out and answer "what was each of these trades actually betting on", and whether you accept the worst case at the new setting. If the answer is vague, do not move up — there is no rule anywhere that says leverage has to go anywhere.
From here, the natural next pieces are the stop-loss paradox and Kelly position sizing.