Home › Learn › Lesson 5

Choosing a trustworthy exchange — the three questions that actually matter

Judge an exchange on three testable things, in this order: can you get your money out, what does a round trip really cost you, and what happens to your balance if the company fails. Brand recognition, app design and referral bonuses answer none of those. This lesson turns all three into checks you can run yourself in an afternoon, with the arithmetic done so you can see how much the wrong choice actually costs.

What you are actually agreeing to when you deposit

Start with the legal reality, because every other consideration hangs off it. When you send coins to a centralised exchange, you no longer own those coins. The exchange owns them, and you own a claim against the exchange — an entry in its internal database saying it owes you that balance. This is not a technicality that only matters in a crisis. It is the entire reason exchange selection is a risk-management decision rather than a shopping decision.

The practical translation: you are an unsecured creditor of a private company, usually incorporated somewhere with limited disclosure requirements, holding customer assets that may or may not be legally segregated from the company's own. In a normal week that claim behaves exactly like ownership — you can trade it, withdraw it, watch it move. On the one abnormal day, it behaves like every other unsecured claim: it queues behind secured creditors and gets settled at whatever the estate can pay.

Nobody deposits expecting that day. But the arithmetic of it should shape how much you deposit, which is a point we'll make concrete further down. First, the three questions.

Question 1 — Can you get your money out?

This is first because it is the only question you can test directly, cheaply, and before you have anything at stake.

Run the withdrawal test. Deposit a small amount — enough that it is a real transaction, small enough that losing it would be annoying rather than painful. Complete identity verification fully. Place one small trade so the account has activity. Then withdraw most of the balance and time the whole thing: from clicking withdraw to the funds arriving in your wallet or bank.

What you're looking for is not just success but friction. Did the venue suddenly request documents it never mentioned at signup? Did a "security review" appear only on the way out? Is there a withdrawal limit that only becomes visible when you try to exceed it? Was support reachable by a human? A venue that processes a small withdrawal in minutes with no surprises has told you more than any review site can.

The part most guides miss: repeat this monthly, even when you don't need the money. Withdrawal behaviour is the earliest externally visible symptom of a venue in difficulty, and it degrades before any announcement is made. Processing times stretch from minutes to hours. A network is "temporarily under maintenance." A limit appears. Individually each has an innocent explanation, and each will be given one. What you have, if you've been running the test all along, is a baseline — you know what normal looks like on that specific venue, so a change registers as a change instead of as bad luck. Traders who only ever test the exit on the day they urgently need it are discovering the fire escape during the fire. The mechanics of doing this safely are in our guide to withdrawing crypto safely.

Question 2 — What does a round trip really cost?

Almost every beginner compares exchanges on the advertised commission, which is the smallest and most visible of four costs. The full bill for one round trip is:

commission (both sides) + the spread you cross + slippage on your size + the withdrawal fee, amortised.

Only the first is printed on your receipt. The rest are embedded in your fill price, which is why they feel free and aren't. If that mechanism isn't yet obvious, how the crypto market actually works takes the order book apart in detail.

Here is the comparison worked out. Two venues, one $10,000 round trip in BTC — buy and later sell, market orders both ways.

The "free" venue is 1.8× more expensive, and the $8,640 annual difference never appears as a line item anywhere. It shows up only as slightly worse entries and slightly worse exits, thousands of times, which the trader experiences as being marginally unlucky rather than as paying a bill. Note the mechanic in the third row: because a market order pays roughly half the spread above the mid on the way in and half below on the way out, a round trip costs you the full quoted spread. That's why a spread quoted as a fraction of a percent can dwarf a commission quoted in basis points.

Two refinements before you go and re-rank everything on spread alone. First, the spread you should measure is the one on your pair at your hour, not the headline BTC/USDT number. Illiquid pairs and quiet weekend sessions are where spreads quietly triple. Second, size matters as much as spread: if your order is large relative to the resting depth, you pay slippage on top, and the venue with the better headline spread but thinner book can end up worse for you specifically.

The fourth cost is the one people ignore because it feels like a rounding error. A flat withdrawal fee of $25 on a $500 withdrawal is 5% — larger than every commission difference you just spent an evening comparing. Check the withdrawal fee per network before you deposit, not after, and check whether the venue supports a cheap network for the asset you'll actually be moving.

Question 3 — What happens if it fails?

You cannot audit an exchange. You are not going to read its books, and if you did you would be reading what it chose to show you. What you can do is check a short list of structural features that change the odds and the recovery, and then size your exposure as if the checks might be wrong anyway.

Start with the most over-trusted signal. A proof-of-reserves attestation typically publishes a Merkle tree of customer balances plus wallet addresses the exchange claims to control, letting you verify your own balance was included in the total. That is genuinely useful and worth more than nothing. But solvency is assets minus liabilities, and proof of reserves speaks only to the asset side. The liabilities figure — the total the venue owes its customers — is generally the exchange's own number. Nor does an address prove unencumbered ownership: coins can be borrowed for the snapshot, and a point-in-time attestation says nothing about the following Tuesday.

So the honest reading is: proof of reserves is half a balance sheet. Its real value is asymmetric — the absence of one from a large venue is a meaningful negative, while the presence of one is a weak positive. Treat it as a filter, never as a guarantee, and be sceptical of anyone who presents it as an audit. It is an attestation, which is a much smaller word.

Then work through the structural checks in the table below. None is decisive alone; together they separate venues that have built for the bad day from venues that have built for the marketing page.

That sixth row deserves a note, because it is where the incentives of the venue and the incentives of the trader can genuinely diverge. On derivatives venues, when a liquidation cannot be filled in the market, some of the loss is absorbed by an insurance fund — and when that is exhausted, by auto-deleveraging: forcibly closing profitable traders on the opposite side. It is the correct engineering answer to an unfillable liquidation, and it means your winning position can be closed without your consent because someone else's leverage failed. A venue with a thin insurance fund and an opaque ADL policy is quietly transferring that risk to you. The market conditions where this bites are exactly the ones described in the anatomy of a liquidation cascade, and it is one more reason the spot versus futures decision is not just about leverage.

Size the exchange like a position — a worked example

Here is the idea that reframes this whole lesson. Your balance on an exchange is not "your money, stored." It is an unsecured position in a private company, and you would never open a position without deciding its size first. So decide this one first too.

Work it from your risk rules, not from convenience. Take a $5,000 account running a 1% risk per trade:

Thirty percent of the account is on the exchange; seventy percent is not. Nothing about your trading changed — same trades, same sizes, same stops — but a total loss at that venue costs you 30% of capital instead of 100%. That is the cheapest risk reduction available to a beginner, and it costs one extra transfer per month. If any of the numbers above are unfamiliar, work them yourself in the position size calculator, and check the leverage arithmetic with the liquidation price calculator.

One caution on the buffer, because it is where people quietly cheat: the 2× is there so an ordinary adverse move doesn't liquidate you at exactly the wrong moment. If you find yourself topping up mid-trade to avoid a liquidation, that is not a buffer problem — it is a sizing problem wearing a disguise, and it will be dealt with properly in the risk-management stage of the curriculum.

Two accounts beats one, and it isn't close

A final structural point that costs almost nothing. Keeping accounts at two venues means you are never fully halted by one venue's outage, one venue's maintenance window, or one venue's decision to pause withdrawals on the network you use. It also gives you a live comparison: run the same pair side by side for a month and you will see, in your own fills, which venue is actually cheaper for the trades you actually make — which is better evidence than any fee table, including the one above.

The cost of the second account is a few minutes of verification and slightly more admin. The benefit is that no single company's bad week can stop you trading or trap your capital. Nearly every trader arrives at this eventually; the only question is whether they arrive before or after the week that forces it. Our current exchange comparison lays out the trade-offs venue by venue.

Common mistakes at this stage

Choosing on the sign-up bonus. A $50 bonus is one month of the spread difference we calculated above; it is a customer-acquisition cost, not a feature. Reading proof of reserves as an audit. It attests assets at a moment in time, not solvency. Comparing headline BTC fees while trading altcoins. Your costs live on your pairs, and the venue that is cheapest on BTC is frequently not the cheapest on the pair you actually trade. Testing the withdrawal only when you need the money. By then you are gathering information you can no longer act on. Keeping the whole account on the venue because it is convenient. Convenience is real, but it is worth a fraction of a percent, and you are pricing it at 100% of your capital. Treating regulation as a safety guarantee. It improves the odds and the process; it has never prevented a failure.

FAQ

Does proof of reserves mean an exchange is safe? No. It attests to assets at one moment; the liabilities it owes customers are usually self-reported. Half a balance sheet cannot demonstrate solvency. Its absence at a large venue is a meaningful negative; its presence is a weak positive.

Is a zero-fee exchange actually cheaper? Frequently not. Commission is one of four costs. In the example above, a venue charging 0.10% per side with a tight spread costs $22 per $10,000 round trip while a zero-commission venue with a 0.40% spread costs $40 — the difference hidden in the fill price.

How much should I keep on an exchange? The margin your open and planned positions need, plus a buffer — in the worked example, $1,500 of a $5,000 account. The rest has no job on the venue.

What's the fastest way to test an exchange? Withdraw before you need to. Deposit small, verify fully, trade once, withdraw, and time it end to end. Then repeat monthly so you have a baseline to compare against.

← Curriculum · Previous: Lesson 4 — How the crypto market actually works