Crypto Futures Trading Costs by Order Size

The 8 October capture closes a reproducibility gap in the original baseline. It retains the actual bid and ask responses for Bybit BTCUSDT, OKX BTC-USDT-SWAP and Hyperliquid BTC, together with the OKX contract-size specification. A reader can rebuild the quantities and VWAP from those levels instead of trusting only the collector’s final numbers.

The compact table focuses on $100,000. Downloads also retain $1,000 and $10,000 instructions. The money column values the filled quantity’s price difference from the best quote on its own side of the book. It does not contain commission, a future exit or movement between venues.

The new books and the broader 4 October baseline have different cohorts and capture times. They are not pooled into one median or a monthly ranking. A difference between two captures can reflect market movement, book depth or retrieval scope; two windows cannot establish a persistent execution advantage.

The collection ran from 02:41:44 to 02:50:32 UTC on 4 October 2026. Ten venues were requested for each of six assets: BTC, ETH, SOL, XRP, DOGE and SUI. Each of the 60 exchange–asset captures produced results for $1,000, $10,000 and $100,000 orders. That yields 180 size scenarios and 360 directional estimates, because buying and selling are evaluated separately.

Every provider–asset observation in this preserved baseline returned an available result at collection time. That historical status does not say its API is responding now. More importantly, the 60 observations were not simultaneous. Prices and depth could change while the collector moved between requests, so small cross-venue differences cannot be interpreted as a synchronized arbitrage opportunity.

The route retains its monthly-execution name to keep existing links stable. The analysis itself is labelled by its actual window. One capture per pair cannot establish a monthly median, a weekend pattern, the worst hour or the depth normally available to a trader. A monthly average would require repeated observations distributed through the month, with an explicit sampling and outage policy.

The edition dated 8 October reorganizes the preserved results and adds reproducible summaries. It does not insert new October observations into the 4 October sample. This separation lets a later reader distinguish when the order-book response was collected from when the report’s presentation and analysis were revised.

The overview below groups the results by asset and requested amount. Each cell reports the median of buy impact plus sell impact across venues where both directions completely filled in the captured calculation. It is a paired depth indicator: the two sides come from the same provider–asset observation. The sample count is visible so a partial result cannot disappear behind an apparently stronger median.

A basis point is 0.01%, or $1 per $10,000 of exposure. At $100,000, 1 basis point represents about $10 of price difference against the calculation’s reference. This is a more readable scale for small depth costs than a long row of percentage decimals. The directional detail retains additional precision for readers checking the original outputs. The collector measures impact from the bid–ask midpoint, so the spread component on that side is already included. Do not add half the spread again.

An increase between size columns indicates that larger orders reached less favourable prices in those captured responses. The pattern is useful for comparing how sensitive an asset’s displayed depth was to the requested amount. It does not establish a stable execution advantage for an exchange, and an unchanged tiny value may simply mean the requested size stayed within the best quoted level.

DOGE at $100,000 has nine fully filled pairs rather than ten. Its tenth result remains in the asset detail with the fill percentage. Excluding it from a full-fill median avoids comparing the cost of buying or selling the whole amount with the cost of executing only part. The exclusion is a property of that scenario, not a claim that the venue has no DOGE liquidity.

At $100,000, KuCoin XRP has a captured buy impact of 6.3897 basis points and sell impact of 1.16. The difference is 5.2297 basis points, the largest directional gap among fully filled scenarios at this amount in this edition. Collapsing the result into a single average would hide which action encountered the thinner displayed side.

Gate BTC provides a different example: the same $100,000 amount shows 0.0059 basis points on the buy calculation and 0.7238 on the sell calculation. Both are complete fills in the captured model, but the sell-side cost is materially larger relative to the buy-side reading. This is a depth asymmetry at that observation, not a forecast that BTC will rise or fall.

A trader opening a short consumes bids; buying to close consumes asks. A trader opening a long consumes asks; selling to close consumes bids. The direction-specific result is therefore more relevant than an exchange-wide liquidity adjective. It still does not predict the second leg later in time, when both the price and the book may be different.

The two directions represent independent orders of equal dollar notional applied to the captured response. They are not a recorded trade followed by an actual exit, and equal dollars on both sides need not correspond to exactly the same token quantity. That is why this report uses buy and sell labels rather than promising a measured round-trip cost.

The Hyperliquid DOGE $100,000 sell calculation reaches 94.9019% of the requested amount; the corresponding buy reaches 100%. In dollar terms the sell covers about $94,901.90 and leaves $5,098.10 unfilled within the collected depth. Any reported price impact for that partial side describes the filled portion, not a guaranteed cost for the full $100,000.

There are several reasons a collected depth response can stop before an order is complete. The endpoint may return a limited number of price levels, the adapter may aggregate depth, or the visible response may genuinely contain less size than requested. This archive does not contain the complete upstream book needed to distinguish every possibility. It supports the partial-fill finding without supporting a wider claim that the order could never fill.

Treating the missing portion as zero cost would make the incomplete venue look artificially attractive. Dropping the entire asset would throw away the other 29 size scenarios for DOGE, most of which are complete. The report instead preserves the individual result, excludes only that pair from the relevant full-fill median and shows the missing share beside the directional value.

The practical next step for a reader considering that amount is a fresh depth inspection at the intended venue and time. Reducing an order, using a limit order or changing execution timing are different choices with different fill trade-offs. This historical report cannot choose among them from one incomplete response and does not present order splitting as a guaranteed remedy.

SUI at $100,000 illustrates the spread of observed outcomes across otherwise identical dollar requests. MEXC shows 2.0509 basis points to buy and 1.5591 to sell. Gate shows 13.9301 and 11.0158. KuCoin shows 12.0659 and 10.4111. All three pairs fully fill in the captured outputs, so the difference is not explained by one row completing only a fraction of the amount.

That does not make MEXC the universal SUI winner. These are separate request times and exclude fees, funding, account restrictions and execution latency. The useful finding is narrower: the sampled SUI depth differed considerably across venues at this size. A decision about that contract should examine its own current book instead of assuming that a venue’s BTC depth describes its altcoin markets.

For BTC, several small impact readings round to the same four-decimal value. A rank built from tiny residual differences would imply more precision than the collection design supports. The detail therefore presents measurements without a best-exchange badge. Meaningful differences in size coverage and direction deserve more attention than a contest between almost identical tiny values.

Each asset panel includes the exact venue symbol and timestamp with the three requested sizes. Those fields let a reader check that a familiar ticker really refers to the intended futures contract and that comparisons come from the stated capture. The panel keeps complete and partial scenarios together; the overview is a summary of this evidence, not a replacement for it.

The depth estimates exclude commissions. A hypothetical 0.05% taker charge is 5 basis points, or $50 on $100,000 for one fill. That can outweigh the sub-basis-point BTC impact seen in several captured responses. Comparing impact alone answers a book-depth question; it cannot identify the lowest all-in trading cost.

We deliberately do not attach today’s account fees to a 4 October observation and label the sum historical. A fee comparison needs the relevant channel, account tier, order class and effective date. An API tariff can differ from an app promotion, while a maker order introduces uncertain fill timing that a market-order depth estimate does not model.

Funding adds a different time dimension. It depends on holding through the applicable settlement and on the rate and direction then in force. A short-lived execution example has no specified holding period, so adding an assumed funding bill would answer an invented trading scenario. The Fee Break-even Calculator and Funding Rate Calculator handle those separate inputs explicitly.

The same separation applies to permitted leverage. A deep-looking book does not establish that a particular account can open the requested leveraged position. Margin brackets, existing exposure and order limits belong to the capacity check. The BTC capacity study complements this execution dataset, but its September inputs are not merged into the October observation as if collected together.

There are now two evidence scopes. The 4 October baseline preserves normalized collector outputs and can reproduce their summaries; its full upstream books were not retained. The 8 October BTC edition preserves full upstream response bytes, contract conversion inputs and a script that recalculates every displayed result offline.

Each package has its own version and checksums. A new observation is appended, never written over an older file. The 8 October script rejects a response whose bytes do not match the archived hash. It converts OKX contract quantity using its captured 0.01-BTC multiplier; treating one contract as one BTC would exaggerate available depth by a factor of 100.

The collector records the original response body, request, retrieval start and finish, returned status and checksum. Re-running it creates a new timestamped directory. A failed request stays a failure; it is not replaced with a prior book carrying a new date.

Before publishing a month-wide comparison, sample the same contract cohort at predefined UTC times across multiple days and retain failures as part of coverage. The publication must state the number of days, windows and fully covered directions for each size. Until that sampling exists, the navigation and title call these observations snapshots.

A later benchmark should use a fixed shared cohort for trend claims and a separate panel for newly added contracts. Version the normalization method as well as the data so that a unit correction does not look like a change in exchange liquidity. No missing historical window is reconstructed from current APIs.