Decentralised News · 2027 planning edition · Evidence reviewed 22 November 2026
What Matters
A market order can execute immediately and still leave part of the requested quantity unfilled. Some venues convert it into a protected immediate-or-cancel instruction; others apply spending or whole-order checks. A price guard limits a type of execution risk while potentially leaving exposure unresolved. Verify both the fills and the remaining position before sending a replacement.
Evidence and method
Evidence checked 22 November 2026 for the 2027 planning edition. The comparison uses official Bybit, Coinbase, Kraken and OKX materials and preserves differences between products, interfaces and protection references. DN did not trade accounts, capture live books or benchmark fill rates. The original calculator walks a synthetic three-level order book under one declared per-level price band; it is not an emulator for every named venue. It excludes queue competition, cancellations, latency, hidden liquidity, fees, spending limits and eligibility checks. A displayed model result cannot establish a real fill. Source inconsistencies and product-specific conditions are identified rather than replaced with a universal “market orders always fill” rule.
DN Alpha Thesis
A useful execution instruction specifies two things: an acceptable price boundary and what the strategy should do with unexecuted quantity. Tightening the boundary may improve the worst acceptable price while reducing completion. Widening it may increase completion in a given book while authorizing more adverse execution. Neither choice removes the need to reconcile the result. For a hedge or exit, the residual quantity can matter more than the average price on the filled portion. The right comparison therefore includes completion, cost and the explicitly authorized response to a partial result.
DN Protected Order-Book Depth Worksheet
Walk a synthetic three-level book under a declared execution-price band. It models available quantity under one IOC-style interpretation, not the complete rules of any exchange. No account or network connection.
Scenario method and limits
The initial inputs and results are an original, synthetic DN example. They are provided to make the method reproducible, not as live prices, measured performance or a prediction for 2027. Change the inputs and download your own scenario above.
- Only one declared per-level band, three price levels and base-quantity instructions.
- Not a model of Coinbase/Bybit/Kraken/OKX complete execution rules.
- No live liquidity, fees, spending guard, whole-order cancellation rule or timing simulation.
- A partial result requires actual order and position reconciliation before any real replacement.
“Market” describes urgency, not unlimited execution
A trader asks to buy 100 units. The venue buys 75 and cancels the remainder. The average execution price looks reasonable, but a second system still expects 100 units to be available. If that system sells, transfers or hedges against the original request instead of the fills, the price protection has exposed a separate control weakness: the workflow confused an instruction with its completed result.
The same issue appears when closing a position. An exit request for 100 units that executes only 75 leaves 25 units exposed. The operator should not infer that a canceled order means nothing happened, nor that a market order means everything happened. Price protection and completion are different outcomes that need separate fields.
This guide examines execution boundaries and their effect on available depth. It complements DN’s stop-trigger protection guide, which asks whether a stop was eligible to activate, and its canceled-order reconciliation guide, which rebuilds exposure after an order’s lifecycle. Here the decision is what a particular execution guard permits the order to consume.
Four protections that are easy to confuse
| Mechanism | What it checks | Possible result |
|---|---|---|
| Stop-trigger protection | Whether a conditional instruction may activate under its reference-price rules | The stop remains untriggered even though a different chart price touched the level |
| Per-level execution band | Whether an execution price is inside the permitted boundary | Allowed depth fills; the rest may be canceled according to time-in-force |
| Spending limit | Whether a buy can complete within its allowed monetary budget | The order stops after consuming the permitted spend |
| Whole-order precheck | Whether the proposed execution estimate passes a rule before matching | The entire instruction may be canceled or rejected rather than partially executed |
A feature called “price protection” may implement any of these at a particular venue. It can also use a different reference: best ask, best bid, midpoint, mark price, index or a dynamic calculation. A 2% boundary from one reference is not equivalent to 2% from another. The order’s product and route must accompany the percentage.
What the venue documentation says
Bybit: market orders can become protected IOC orders
The Bybit order API describes market instructions being converted to immediate-or-cancel limit orders for matching under its protection mechanism. It also documents user-specified tolerance fields based on ticks or a percentage of best ask/bid, with product and order-type restrictions. The applicable instrument details still matter.
Bybit’s slippage-tolerance help page explains that only available quantity within the selected range may execute, with the remainder canceled. Its description of disabling the user-facing tolerance should not be interpreted as proof that every venue-level safeguard disappears: the API separately documents the engine’s market-order treatment. Verify the actual interface and product rather than assuming one toggle describes the entire path.
Coinbase: a midpoint-based protection boundary
Coinbase’s trading rules describe a Price Protection Point for specified Exchange order types, measured from the order-book midpoint at placement. Under that rule, matching proceeds to the protection boundary and the remaining portion is canceled. The rules refer readers to the market information for asset-specific parameters.
That reference matters even when the percentage looks familiar. A midpoint-based buy boundary may differ from a best-ask-based boundary when the spread is wide. Do not transplant a retail help example or another Coinbase product’s setting into the Exchange rule without confirming scope. The comparison here records the mechanism, not a promise that one percentage governs every Coinbase interface.
Kraken: spread protection and a buy-spend constraint
Kraken’s Market Price Protection page, last updated 30 October 2026, describes cancellation when the spread becomes unusually wide under pair-dependent limits. It separately states that market buys have a 5% spending limit and can partially fill before the remainder is canceled. These are the page’s documented settings at review, not universal defaults for this worksheet.
A spend guard concerns money committed, whereas a per-level band concerns which prices may trade. Those dimensions can bind at different times. A trader should record whether the request was entered in base quantity or quote value before interpreting a shortfall. “Buy 100 units” and “spend 10,000” are different instructions.
OKX: do not assume every guard permits partial execution
OKX’s price-limit rules, updated 27 August 2026, distinguish dynamic product price limits from a spot/margin execution check. The latter describes full cancellation when the estimated filled price exceeds the stated best-ask/bid range. It should not be substituted with a generic partial-fill model. Its API guide also describes futures market orders using the most aggressive price allowed by the relevant mechanism.
The difference is operationally meaningful. A book containing some acceptable liquidity does not guarantee that a particular venue will execute that portion if its applicable rule first checks the entire instruction. The DN calculator deliberately models one transparent per-level band, then states where that model is too narrow.
The DN protected-depth worksheet
The worksheet accepts a side, a reference price, a tolerance percentage, a requested base quantity and three price/size levels. For a buy, the upper boundary is reference × (1 + tolerance/100). For a sell, the lower boundary is reference × (1 − tolerance/100). It walks the entered levels in price priority and counts only sizes inside that boundary.
The levels must be ordered from best to worse: ascending prices for a buy and descending prices for a sell. Quantities must be nonnegative. Prices and requested quantity must be positive. A zero size is allowed to represent an empty level, but a missing size is an incomplete record rather than zero liquidity.
The model reports fillable quantity, unexecuted quantity, gross notional, average price, worst consumed price and completion percentage. These are arithmetic outputs from a static book. They are not orders, quotes, reserved liquidity or observed execution. The model assumes all entered in-band depth is available to this instruction with no competing activity.
A synthetic buy: 100 requested, 75 within the boundary
Assume a reference price of 100 comparison-currency units, or CU, and a 2% tolerance. The maximum permitted execution price is 102. The fictional ask book contains 40 base units at 100, 35 at 101 and 50 at 103. The trader requests 100 units.
| Ask level | Available units | Inside 102 boundary? | Modeled quantity consumed |
|---|---|---|---|
| 100 CU | 40 | Yes | 40 |
| 101 CU | 35 | Yes | 35 |
| 103 CU | 50 | No | 0 |
Modeled completion is 75/100 = 75%. Gross spend is 40 × 100 + 35 × 101 = 7,535 CU. The average price on those 75 units is 100.466667 CU. The unexecuted 25 units are not included in that average. Under an IOC interpretation, that remainder would not sit on the order book waiting for a better price.
The tempting conclusion is that the order cost only 0.4667% more than the reference. That is true for the modeled filled portion, before fees. It says nothing about the cost of completing the remaining 25 units later. A report that ranks execution solely by the first average can reward incomplete orders and hide a later expensive catch-up trade.
Widening the boundary changes both completion and cost
Increase the synthetic tolerance to 3%, leaving the book unchanged. The boundary becomes 103. The model can now consume 25 units at that third level and complete the requested 100. Gross spend becomes 4,000 + 3,535 + 2,575 = 10,110 CU; the average becomes 101.10 CU.
The extra completion is 25 units at 103. It is not a free improvement to the earlier average. Whether that trade-off is acceptable depends on the strategy’s authorized limit and cost of remaining unexecuted. The worksheet makes the arithmetic visible but does not choose a tolerance on the reader’s behalf.
The mirrored sell example
For a sell, use bids of 40 units at 100, 35 at 99 and 50 at 97, with the same 100-unit request and 2% tolerance. The floor is 98. Only the first 75 units fit. Gross modeled proceeds are 4,000 + 3,465 = 7,465 CU, for an average of 99.533333 CU. Another 25 units remain unsold.
This direction check catches a common implementation mistake: applying a buy ceiling to a sell. The sell boundary moves downward from its reference. A more permissive sell floor authorizes lower prices, so a higher completion percentage must be evaluated alongside reduced proceeds, not described as an unconditional execution improvement.
Reference prices can matter more than the headline percentage
Consider a fictional book with best bid 98 and best ask 102. Its midpoint is 100. A 2% buy boundary from the midpoint is 102, while a 2% buy boundary from the best ask is 104.04. The same percentage produces materially different permitted prices. Neither calculation proves which rule an exchange uses.
Record the actual reference and its observation time. A chart’s last trade, an index and the executable offer can all differ. A last trade from several seconds earlier is particularly weak evidence of the current price available for a large instruction. The model therefore asks for the reference explicitly instead of labeling every tolerance as slippage from “the market.”
A fixed tick allowance is another distinct unit. Ten ticks on an instrument with a 0.01 tick size means a 0.10 price allowance. Ten ticks on an instrument with a 1.00 tick size means 10.00. A wrapper that forwards the number but loses the tolerance type can authorize a very different boundary.
Why a static book is an illustration, not an execution forecast
The displayed levels can change between observation and matching. Other traders may consume the same liquidity; existing orders may be canceled; additional orders may arrive; and only some displayed liquidity may be accessible to the instruction’s route. The worksheet assumes away those interactions so readers can isolate the arithmetic effect of one price band.
It also does not model a venue’s complete pre-trade checks, quote-currency reservation, minimum order size, rounding, price amendments or market state. Fees are excluded from gross spend and proceeds. For a contract whose quantity is expressed in contracts rather than underlying units, normalize it correctly before using this base-quantity model.
Increasing the requested quantity beyond all three entered levels cannot reveal deeper liquidity. With a 150-unit request and a 3% buy boundary in the original example, only 125 units exist in the supplied book. The residual 25 is an unknown opportunity beyond the modeled snapshot, not proof that the real market lacks buyers or sellers.
The order response is not the fill ledger
The Bybit API FAQ explains that an order response can expose an internal protective limit while the actual execution price arrives in execution data. It also describes partial matching followed by cancellation where in-range liquidity is insufficient. That distinction is a reason to store the acknowledgement and execution records separately.
A useful record contains requested quantity, acknowledged instruction, cumulative filled quantity, average fill price, remaining quantity, final status and observed position. Preserve native identifiers and timestamps. If the process loses a response, reconcile state before resubmitting; a new request can otherwise duplicate the already filled portion.
Suppose the first request bought 75 units and the desired final position is 100. A replacement request for the original 100 could produce 175 units if it fills completely. The residual target is 25, subject to confirmed fills and the strategy’s current decision. That arithmetic is simple, but it is exactly where an execution guard becomes a portfolio risk if systems do not share reconciled state.
How to evaluate a protection setting before an important exit
Start with the decision the exit is meant to satisfy: maximum remaining exposure, acceptable price range and urgency. Then identify the applicable product rule and reference. The OKX API documentation provides a real-time price-limit endpoint and product-specific order semantics; the presence of an endpoint is useful, but calling it does not reserve that boundary or future liquidity.
Use several synthetic depth scenarios, including a shallow book, a wide spread, no in-range depth and a boundary-level fill. Check whether the strategy has authority to stop after a partial result, wait, use a permitted limit order, or request human review. A fallback that automatically widens tolerance without a defined cap can defeat the original risk decision.
Review linked protection orders against the confirmed remaining position. Closing only part of a position can change the quantity that a later stop or take-profit should address. Do not assume every platform automatically resizes every linked instruction. The intended post-execution state needs its own check, separate from whether the initial price guard behaved correctly.
Compare complete outcomes, including the residual
For an execution comparison, report the same target quantity and horizon. Include completion percentage, gross and net cost, reference convention, fees actually charged and the treatment of unexecuted quantity. If the remainder was deliberately left open, say so. If it was completed later, retain the second instruction and its price rather than hiding it inside an unexplained average.
In a synthetic two-step buy, the first 75 units cost 7,535 CU and the later 25 cost 2,650 CU at 106. The complete 100-unit cost is 10,185 CU, or 101.85 per unit, before fees. That can be compared with the earlier 101.10 static full-completion scenario only as a stated hypothetical. It does not prove that widening the first order would have achieved that execution in a live market.
The practical result of a 2027 review is a concise execution policy: which price references and protection settings are allowed, how much residual exposure is tolerable, what evidence confirms completion, and which fallback actions are authorized. Price protection is useful when its consequence for incomplete execution is understood and managed.
Frequently asked questions
Does a market order always fill completely?
No. A venue may apply execution-price limits, spending controls, market conditions or other checks. Depending on the product and rule, the order can partially execute, be canceled or be rejected. Confirm actual executions and the remaining position.
Is execution price protection the same as stop-loss trigger protection?
No. Trigger protection governs whether a conditional instruction activates. Execution protection governs what happens when an instruction reaches matching or another execution check. A stop can trigger successfully and still leave part of its resulting order unfilled.
Will switching off my custom slippage tolerance remove every price guard?
Do not assume that. A user-facing tolerance and the venue’s underlying engine protections can be different layers. Confirm the current rule for the exact product, interface and order type before interpreting a disabled setting.
Should I resubmit the original quantity after a partial fill?
First reconcile cumulative fills and the current position. If the desired final exposure has not changed, only the confirmed residual may be relevant. Resubmitting the original request without reconciliation can duplicate exposure.
Does the worksheet predict the next market order’s fill?
No. It walks three user-entered price levels under a single per-level band. It excludes changing liquidity, fees, spending limits, whole-order checks and other venue rules. The output is a transparent scenario, not a tradable quote or execution forecast.
Sources and verification
Primary documentation checked on 22 November 2026. Exchange rules, regional availability, fees and API behavior can change; recheck the exact product and account before using this 2027 planning guide.
- Bybit — Place Order API — Protected IOC handling, user tolerance inputs and product/order constraints
- Bybit — Market Order with Slippage Tolerance — Best-ask/bid tolerance formulas and in-range partial execution
- Coinbase — Trading Rules, Order Protections — Midpoint-based Price Protection Point and canceled remainder for covered order types
- Kraken — Market Price Protection — Pair-specific spread protection and market-buy spending limit
- OKX — Price limit rules — Dynamic product rules and spot/margin whole-order estimated-fill check
- Bybit — API FAQ — Asynchronous acknowledgement versus actual execution price and partial IOC outcomes
- OKX — API guide — Real-time public price-limit endpoint and futures market-order semantics




