The order type tells a venue what you want it to do; it does not tell you what actually filled. A market order prioritizes reaching available liquidity. A limit order puts a boundary on the price you will accept. A stop waits for a trigger, then activates another order. A time-weighted average price (TWAP) instruction schedules smaller attempts over time. Every choice leaves a different uncertainty: price, completion, timing, or some combination of them.
Start with the trade-off: price control or execution priority?
Suppose a fictional ask book offers 1 BTC at $100.00 and another BTC at $100.20. A market buy for 1.5 BTC that consumes both levels costs $150.10: (1 × $100.00) + (0.5 × $100.20). Its average fill is $100.067 per BTC before fees. The last trade or best ask alone would not show that average. This is arithmetic on a fixed, invented snapshot, not a quote or a prediction. Actual orders can fill at several prices or only partly, depending on available size and venue rules.
A limit order sets the worst price you will accept: a buy limit is a maximum and a sell limit is a minimum. It can execute immediately if it crosses available orders, rest on the book, fill partly, or never fill. On the same invented book, a buy limit of $100.10 can take the one BTC at $100.00 but cannot buy the next level at $100.20. Depending on the order’s time-in-force, the remaining BTC might rest or be canceled. A marketable limit can trade immediately while retaining its price ceiling; it is not the same instruction as an unrestricted market order. Coinbase’s Advanced Trade API guide lists its own market, limit, stop, bracket and attached TP/SL controls and time-in-force settings; use that venue’s page for its specific behavior.
Price control has a cost: a resting limit order can miss a move. Even if the market trades at its price, candles do not reveal whether enough size traded ahead of it to reach the order in the queue. “Limit” also does not automatically mean maker. A marketable limit can take liquidity, while post-only instructions on supported venues aim to prevent an immediate take and may reject an order that would cross. Fees depend on the actual fills and that venue’s current schedule. See the separate guide to crypto spreads, depth and slippage.
A stop has a trigger and a second instruction
The trigger price and the resulting order are separate settings. For a hypothetical long entered at $100, a sell stop-market could activate at $95. If the market gaps through that level, it sends a market sell into the bids then available; $95 is not a guaranteed fill price. A sell stop-limit could use the same $95 trigger and a $94.80 limit. That sets the lowest acceptable sale price, but if the next available bids are below $94.80, the triggered order can remain unfilled while the position is still open.
Never assume the trigger uses the same price series you see on a chart. Venues can trigger from last, mark, index, or another defined reference. Hyperliquid’s current perpetual TP/SL documentation says those triggers use mark price and lets users choose market or limit behavior. A market-style exit prioritizes execution with uncertain price; a limit-style exit preserves its bound with non-fill risk. Check the market type and venue documentation before applying those terms elsewhere.
Stop-market and stop-limit orders are not the only stop designs. A trailing stop moves its trigger as a defined favorable price reference changes; the formula and supported markets vary. A take-profit/stop-loss bracket can link two exits so one cancels when the other activates, but partial parent fills and cancellation rules matter. Coinbase’s API documentation says its attached TP/SL size follows the originating order and the other side is disabled after one side fills. Read the product’s own rules; the names alone do not establish how child orders are sized, triggered, or canceled.
Time-in-force and position flags are a separate layer
These labels modify an order’s life or permitted effect; they are not substitutes for market, limit, or stop instructions. Common examples include:
- Good till canceled (GTC): keep an eligible resting order open until it fills or is canceled; a venue may impose its own expiry rules.
- Immediate or cancel (IOC): execute whatever is immediately available and cancel the remainder.
- Post-only: ask the venue to add the order as resting liquidity instead of immediately matching it; a crossing order may be rejected or handled according to the venue’s rule.
- Reduce-only: on supported derivatives products, prevent the order from increasing or reversing the current position. It does not guarantee an exit fill.
These are common meanings, not a promise that every exchange offers them or implements them identically. Confirm whether a control applies to the parent order, a triggered child, or each TWAP slice.
TWAP spreads an attempt through time; it cannot promise completion
A TWAP divides an intended quantity across scheduled time intervals. In a deliberately simple example, 1.2 BTC over six one-minute intervals means a planning amount of 0.2 BTC per interval. If fills are 0.2, 0.2, 0, 0.2, 0.1 and 0.2 BTC, the order has filled 0.9 BTC and 0.3 BTC remains when the schedule ends. A venue’s live algorithm may redistribute missed size or change slice sizes, so the even schedule is only a teaching example.
TWAP controls when attempts are spread; it does not ensure the full quantity trades, cap every price, or guarantee an average near any benchmark. Hyperliquid documents a minimum 30-second interval, a per-suborder slippage constraint and the possibility of finishing short when slices fail to fill. Those settings describe Hyperliquid’s own implementation and may change. Do not reuse them as generic crypto rules. TWAP is also different from volume-weighted average price (VWAP): a time schedule weights intervals, while VWAP weights observed traded volume. Either can have market impact, timing risk, or incomplete fills.
Keep an order record that can be reconciled
For a real venue order or a hypothetical research fill, preserve the decision time, market and product, side, requested size, order type, trigger reference, trigger and limit prices, time-in-force, and any position flags. Then record acknowledgments and rejects, each partial fill with its price, quantity, fee and timestamp, and the final state of the unfilled remainder. Coinbase’s order-management view exposes fields including type, side, amount, percentage filled, fee, time and status; treat that as an example of one interface, not a universal API schema.
In a backtest, say exactly how a hypothetical order becomes a modeled fill. OHLC bars cannot reconstruct changing depth, queue priority, latency, or which price was reached first inside a bar. If one candle crosses both a stop and a target, the data may not tell you which happened first. SuperQuant does not connect to an order-placement venue or submit live orders; its sample trades are simulated and must not be described as exchange executions. Compare this guide with stop loss versus liquidation and the crypto trading fee guide before interpreting a modeled result.
The venue references below were checked on 25 September 2026. They support the product-specific examples above, not a recommendation to use a venue, order type, asset, or strategy. This guide is educational, and an order instruction cannot remove the risk of loss.