Crypto Leverage Statistics: Fees, Markets & Limits

The main distribution uses 377 confirmed contracts; 341 have a retrieved leverage ceiling, so percentages of ceilings use 341 unless explicitly stated otherwise. Fee panels are a different sample of 20 schedules, separated by execution channel. The historic BTC capacity panel contains eight contracts. Those three denominators should never be divided into one another.

Two useful findings survive that separation. First, 192 of 341 retrieved maxima sit between 50× and below 100×: the most common captured range is not the 200× headline at the top of a marketing page. Second, the historic BTC median falls from 150× at $100,000 to 50× at $10 million, a two-thirds reduction across the same eight contracts. The latter describes a size effect, not the average leverage traders actually use.

At equal $100,000 fill value, the public CEX median taker rate of 0.0525% means $52.50 of commission; the public DEX median of 0.045% means $45. The $7.50 difference is small enough for funding or execution impact to reverse a fee-only choice. The groups are small and unweighted, so this is a scale comparison rather than evidence that every DEX is cheaper.

Leverage was retrieved for 341 of the 377 confirmed contracts, or 90.5% of the sample. Of those 341 values, 192 fall from 50× to below 100×: 56.3% of the retrieved subset. Another 63 are at least 200×. The distribution therefore contains some very high ceilings, but those outliers do not describe the typical record in this particular collection.

A maximum is the advertised or documented ceiling captured for an exact contract, usually associated with its smallest permitted risk bracket. It is not the leverage being used by customers. This dataset has no account balances, liquidations or user position histories, so it cannot estimate average trader leverage or tell us how many people select 100×.

The practical value of this distribution is a screening question: is a very high leverage requirement common enough in the covered contracts to justify comparing it? Once an exact market is selected, the more useful question becomes whether that limit survives at the intended position size. That is why the size result appears alongside the headline distribution rather than being treated as a footnote.

A retrieved leverage ceiling exists for 341 contracts, but a usable normalized margin ladder exists for only 302. That is 80.1% of all confirmed contracts. The difference matters because a single ceiling does not explain the additional collateral required as a position grows. A trader comparing a six-figure order needs the applicable bracket, not just the largest multiple on a product page.

Eight providers have ladders for every contract they contributed to this confirmed sample. Binance and WEEX have zero normalized ladders in this particular collection, accounting for the other 75 records. This is a collection limitation. It does not mean either exchange operates without margin rules, or that no separately sourced ladder exists elsewhere on this site.

Coverage also depends on the field. All 377 confirmed records have a retrieved funding interval in this edition, while margin ladders are missing from 75. A single completeness badge would hide that difference. The Contract Explorer lets a reader inspect the dates and sources for leverage, fees and limits separately instead of assuming that one successful price request refreshes every rule.

The frozen fee collection contains 20 starting maker/taker rows used in the comparison. A combined median taker rate is 0.05%, but it is a mixed-channel descriptive number. It includes public exchange schedules, an API-specific schedule, institutional execution and a builder base cost. Calling that the median retail crypto futures fee would misdescribe the inputs.

Separating channels leaves 12 public CEX schedules with a median maker rate of 0.02% and median taker rate of 0.0525%. There are three public DEX schedule rows with a median taker rate of 0.045%. These are small, unweighted groups. They exclude special account quotes and do not prove that one business model is inherently cheaper after funding, borrowing, network or execution costs.

MEXC is isolated as an API observation at 0.06% maker and 0.08% taker. Its official June update explicitly separates API fees from web and app promotions. For scale, a 0.08% charge on a $100,000 fill is $80 for that execution. A website promotion cannot be substituted into an API cost estimate without evidence that it applies to that channel.

The Coinbase row concerns institutional execution through Coinbase Bermuda to Deribit. It is not a quote for a retail Coinbase perpetuals account. Orderly is a different case again: its recorded base cost is not necessarily the final amount a builder charges an end user. Both rows remain visible with their correct scope instead of being allowed to make a retail comparison look cheaper.

Lighter and Paradex are separated as order-class-dependent entries. A zero starting commission in those records does not establish zero cost for every order type or classification. Funding and price impact remain separate even when the trading commission is zero. Pool-based fee schedules are outside the maker/taker median because their charging model is not an equivalent pair of order-book rates.

The all-channel median and the CEX median excluding MEXC remain in the download for continuity with earlier editions. The channel panels are the preferred interpretation. A reader can reproduce either measure and see exactly which rows have been included, rather than having a cleaner-looking number replace the original calculation without explanation.

The confirmed sample contains 247 eight-hour intervals, 75 hourly intervals and 55 four-hour intervals. That corresponds to 65.5%, 19.9% and 14.6% of 377 contracts. These are interval counts, not the share of trading volume paying funding on each schedule. They do not describe the amount or direction of the next payment.

An interval changes how a quoted rate should be compared. Under a deliberately constant-rate example, 0.01% every hour would create eight charges over eight hours, while 0.01% every eight hours creates one charge over the same duration. Real funding rates move, and a position must satisfy the venue’s holding-time or accrual rule. The example explains why interval labels belong beside rates; it is not a forecast of actual carry.

A provider can contribute more than one interval across its assets. TAO appears among the four-hour records, but the finding is about the exact captured contracts rather than the token having one permanent funding schedule everywhere. A reader planning a trade should carry the contract-specific interval into the funding calculator instead of borrowing the interval from a different market on the same exchange.

An interval alone also cannot identify the next settlement timestamp. Nor does it fully describe a venue that accrues continuously and settles through a separate process. This study counts the documented field present in the contract source. The Funding Rate Calculator handles event timing and continuous models as distinct choices.

The eight-contract BTC cohort keeps its own September observations and is not a new October liquidity sample. At $100,000 of total exposure its median documented ceiling is 150×. At $1 million it is 87.5×, at $10 million 50×, and at $100 million 10×. All eight contracts remain comparable at these checkpoints; the falling median is not caused by removing the inconvenient rows.

The ordering also changes. Hyperliquid has the smallest headline ceiling in that cohort at 40×, yet the same 40× remains documented at $100 million and exceeds the other seven study results there. That reversal is a useful reason to compare at size. It is not a recommendation to place a $100 million order, because a margin rule does not measure the executable book.

At the largest checkpoint seven of the eight records also exceed a documented single-order limit. The study retains the total-position margin result and flags the order constraint separately. Removing those records from the margin median would silently turn a position comparison into an order-entry comparison. The dedicated BTC study develops those cases without repeating every ladder here.

The separate access index covers 201 countries and territories through 2,782 platform evidence records in the compiled summary. Those records are documentary access findings, not approved user accounts. One jurisdiction can contribute many platform records, and one platform can have several different product routes.

A technically documented BTC contract does not automatically become available to every resident whose country is absent from a restriction list. Equally, a platform that supports spot purchases in a jurisdiction does not necessarily offer leveraged derivatives there. The appropriate next step is the Country Access Checker, which distinguishes affirmative product evidence from inference and conditional access.

The access totals are frozen into this statistics edition so later country edits cannot quietly change the cited denominator. They summarize the source dossiers and do not constitute a new legal survey of all 201 jurisdictions on 8 October. Current decisions need product-specific evidence rather than a broad country label.

There are three different questions behind the numbers on this page. The broad registry asks where our research has an exchange–asset record. The confirmed-contract sample asks whether an exact contract and its public fields were retrieved. The BTC size study asks how the same position amount interacts with eight preserved margin ladders. The samples overlap, but their totals are not interchangeable.

The narrower contract collection requested 40 assets from ten providers, creating 400 possible provider–asset combinations. It retained 377 confirmed contracts. The remaining 23 requested combinations are outside that confirmed sample; they are not entered as zero-leverage markets. Providers contribute between 36 and 39 records, so the collection is controlled by the requested asset list but is not an equal 40-contract sample per provider.

Every confirmed contract receives one observation weight in the distributions below. A small altcoin contract counts once, just as BTC does. We do not have a suitable account-position dataset with which to weight these limits by trader usage, and reported turnover would answer a different question. Read a percentage as a share of the stated records, never as a share of all crypto trading.

The broader internal registry contains 1,000 research rows and 41 market labels, while the confirmed-contract collection requested 40 assets. Registry totals remain in the fixed download for provenance; they are not substituted into the market distributions above.

Use version 2026-10-08.1 when quoting a figure on this page. The version identifies a compilation, not a simultaneous market capture. Fee reviews, contract observations and the BTC size cohort have different source dates, retained in the package. An October analysis date cannot make a September margin ladder current.

The prior 6 October edition reported 997 registry rows. The new compilation contains 1,000. That difference reflects the state of the compiled registry, not three proven exchange listings arriving during the intervening period. Without a controlled listing history, the change should not be used as a market-growth statistic.

The earlier manifests remain available. Corrections are issued as new editions instead of overwriting their data. The JSON includes the arrays behind the charts and channel breakdowns; the CSV contains named headline metrics and denominators. The compressed input package preserves the schedules and records used to compile them. A citation should identify the specific sample as well as the version.