How to Pick a Crypto Exchange After a Month Like This (2026)

Jump to a Section

• What the Two Failures Actually Teach
• Test One: Can You Get Out?
• Test Two: What Does It Really Cost?
• Test Three: Who Covers You When It Breaks?
• Test Four: Does Anything Bind Them Where You Live?
• The Part That Is Not About the Exchange

What the Two Failures Actually Teach

They failed in opposite ways, which is why taking both together is more useful than either alone.

CoinEx stopped on purpose. Nine years of operation, then a wind-down attributed to market conditions and compliance costs. Nothing was stolen. Users get a published window to withdraw, a token buyback, and a deadline. The risk here was never security — it was that a business can simply decide the economics no longer work.

Bitget was attacked. Private keys held, cold storage held, and an attacker still moved hundreds of millions by feeding forged transaction data through the exchange’s own approval system. Withdrawals froze for days. The loss was covered from a protection fund, and the schedule to reopen was published.

Put side by side: one shows that size and longevity do not prevent closure, the other that strong key management does not prevent loss. Neither is predictable from an exchange’s marketing, and no checklist would have flagged either in advance.

What a checklist can do is tell you how much damage each scenario does to you. That is a different and more answerable question, and it is the one worth spending time on.

Test One: Can You Get Out?

Both events turned into the same practical problem: a withdrawal you could not make on the day you wanted to make it. So check the exit before you check anything else.

• Withdrawal limits at your verification tier. Not the headline maximum — the one that applies to you, today, on the asset you actually hold. This is where people discover a daily cap mid-transfer.
• Which networks are supported for your asset. A venue that only supports one chain for USDT is a venue where one congested or halted network blocks your exit entirely.
• The fiat route, if you need one. Getting to stablecoin is not getting to money. Check whether the off-ramp works in your country before you need it.
• Whether they publish a status page. Sounds minor. During an incident it is the difference between knowing you are in a queue and assuming you have been robbed.

Run a small withdrawal early, before you have a large balance and before anything is wrong. It costs a network fee and it converts an assumption into a fact.

Test Two: What Does It Really Cost?

Cost is the only item on this list that affects you every single day rather than on the rare bad one, which is why it deserves more weight than it usually gets.

Published maker and taker rates are the starting point and not the answer. The number that matters is what leaves your account per round trip after discounts, volume tiers and any rebate, plus the slippage you actually pay. On a strategy with a modest edge, the gap between venues at the same nominal rate is frequently larger than the edge itself.

A concrete illustration of the spread: across venues we track, effective rates after rebate range from roughly 0.015% to 0.045% per side for comparable activity. On $10,000 of notional that is $1.50 against $4.50 — and repeated across a year of active trading it stops being a rounding difference.

Here is the spread as measured today, with the effective rate being what you pay per side after the rebate is applied:

Venue	Rebate	Effective fee	Cost on $10,000
Gate	70%	0.015%	$1.50
Bitget	50%	0.015%	$1.50
OKX	20%	0.04%	$4.00
Binance	10%	0.045%	$4.50
Bybit	20%	0.044%	$4.40
EdgeX	없음	0.045%	$4.50

Note what the table does to the obvious assumption. The highest rebate percentage and the lowest effective cost are not the same row — Gate’s 70% and Bitget’s 50% land at an identical 0.015%, because the rebate is applied to different starting fees. A rebate percentage on its own tells you nothing; only the effective rate does.

Two things people under-weight here. Funding, if you hold perpetuals across settlement intervals, can exceed trading fees on a multi-day hold. And order type usually moves your real cost more than switching venue does, because the maker-taker gap at one exchange is often wider than the taker gap between two exchanges.

We keep the rate comparison, including what each venue’s rebate actually nets out to, at TetherBase. The point of looking is the arithmetic rather than the headline percentage.

Test Three: Who Covers You When It Breaks?

The Bitget incident was covered from a stated protection fund larger than the loss. That is a meaningfully better position than having nothing, and it is worth checking whether a venue has one.

But be clear about what it is. A protection fund is the company’s own commitment, funded at its own discretion — not insurance, and not a legal guarantee. It can be drawn down, it can be smaller than the next incident, and its size is usually self-reported. Ask three things: does it exist, is its size published and verifiable on-chain, and is it bigger than a plausible loss.

The same scepticism applies to proof-of-reserves pages. An attestation is better than nothing and weaker than an audit. Check whether liabilities are covered too, because reserves alone say nothing about what is owed against them.

Test Four: Does Anything Bind Them Where You Live?

This is the test most people skip and the one that decides what happens when something goes genuinely wrong.

If an exchange is not registered in your jurisdiction, there is usually no local body to complain to, no domestic deposit protection, and no realistic legal route if funds are frozen or a dispute goes nowhere. Support is the only channel, and it is the company’s own.

That does not automatically mean avoid offshore venues — there are real reasons people use them, including product range and cost. It means the amount you keep there should reflect what recourse you would actually have. The threshold for “money I can afford to have stuck” is not the same at a locally regulated venue and an offshore one, and treating them as interchangeable is the mistake.

The Part That Is Not About the Exchange

Both September events were survivable for anyone holding a working balance and painful for anyone holding everything.

No amount of diligence on this list changes that. An exchange can close in an orderly way, be attacked despite good key management, restrict your country at short notice, or freeze withdrawals while it investigates — and all four have precedent. The variable you control is not which venue avoids those, it is how much of your position is exposed when one happens.

So the checklist has a final item that is about you: keep on any exchange the amount you need to trade with, and keep the rest where an approval system, a compliance decision or a corporate wind-down cannot reach it. That is boring advice, and it is the only item on the list that worked in both of this month’s cases.

FAQ

Is a bigger exchange safer?
Size helps with liquidity and with the ability to absorb a loss, and this month showed it does not prevent one. Treat size as one input rather than the answer.

Should I split across several exchanges?
It caps the damage from any single failure and multiplies the number of accounts you must secure and verify. For most people two is a reasonable compromise; beyond that the admin cost starts to exceed the benefit.

How do I check a protection fund is real?
Look for a published wallet address and verify the balance on-chain yourself. A figure quoted only in a blog post is a claim, not evidence.

Does a rebate programme make a worse exchange worth using?
No. Cost is the item that affects you daily, but it only matters among venues that already pass the exit and recourse tests. A cheap venue you cannot withdraw from is not cheap.

Sources

• Bitget Support — Withdrawal Service Update (September 24, 2026).
• Decrypt — CoinEx is shutting down after nine years.
• TetherBase — our own comparison of effective trading costs after rebate across connected venues, measured September 2026.

Written September 27, 2026 from public sources. Both situations referenced are developing and their details may change. This article is informational and is not financial advice. Cryptocurrency trading carries risk, and leveraged trading can result in the loss of your entire principal.