Limit Orders on Crypto Exchanges vs Stocks
How limit orders on crypto exchanges differ from a US stock broker: post-only flags, venue-local pricing, maker taker fees, and a 24/7 book with no open.
Limit orders on crypto exchanges carry the same core instruction as a limit order at a US stock broker: fill at my price or better, never worse. What changes is everything around that instruction. Price discovery is local to one venue, the fee schedule charges the resting side less than the crossing side, the book never shuts for the night, and a post-only flag can cancel an order that a stock broker would have filled instantly.
How limit orders on crypto exchanges differ from a stock broker
Four mechanisms account for most of the surprises when someone moves from an equity broker to a crypto exchange.
- Best price is local to one venue. Nothing consolidates crypto books into a single national quote, and no best-execution duty points an order at another venue's better price.
- Post-only inverts the marketable limit order. A limit order that would cross the spread gets cancelled instead of filled.
- The fee tier, not only the price, favors resting. The maker side of a trade is charged less than the taker side, and at the largest tiers sometimes a negative fee.
- The book is continuous. There is no opening auction, no closing auction, and no queue that builds overnight, so "at the open" has nothing to point at.
The panels below measure the US equity side of each comparison, the baseline most readers already know. Every number has its SQL open underneath it.
Why best price is venue local on a crypto exchange
In US equities a limit order rests on one exchange, but the price it is judged against is consolidated. Each exchange publishes its best bid and offer to a shared tape, those quotes combine into the national best bid and offer, and a broker owes the customer an execution at that consolidated price or better. The individual venues still disagree moment to moment, which is exactly what the consolidation hides.
| venue | avg_spread_bps | tightest_spread_bps | quote_updates |
|---|---|---|---|
| Nasdaq | 0.75 | 0.3 | 1360 |
| NYSE Arca, Inc. | 0.89 | 0.3 | 338 |
| Investors Exchange | 1.06 | 0.3 | 74 |
The exact SQL behind every number
SELECT
coalesce(nullIf(ex.acronym, ''), ex.name) AS venue,
round(avg(10000 * (toFloat64(q.ask_price) - toFloat64(q.bid_price))
/ ((toFloat64(q.ask_price) + toFloat64(q.bid_price)) / 2)), 2) AS avg_spread_bps,
round(min(10000 * (toFloat64(q.ask_price) - toFloat64(q.bid_price))
/ ((toFloat64(q.ask_price) + toFloat64(q.bid_price)) / 2)), 2) AS tightest_spread_bps,
count() AS quote_updates
FROM global_markets.cache_stocks_quotes AS q
INNER JOIN global_markets.stocks_exchanges AS ex
ON toUInt32(ex.id) = toUInt32(q.bid_exchange)
WHERE q.ticker = 'AAPL'
AND q.sip_timestamp >= '2026-09-23 18:30:00'
AND q.sip_timestamp < '2026-09-23 18:35:00'
AND q.bid_exchange = q.ask_exchange
AND toFloat64(q.bid_price) > 0
AND toFloat64(q.ask_price) > toFloat64(q.bid_price)
GROUP BY venue
HAVING quote_updates > 50
ORDER BY avg_spread_bpsAcross the five minutes beginning 14:30 ET on September 23, 2026, 3 US exchanges posted two-sided quotes in AAPL. The tightest average spread measured 0.75 basis points (a basis point is one hundredth of a percent) at Nasdaq, over 1360 quote updates. The widest of the group averaged 1.06 basis points at Investors Exchange. A retail equity customer rarely has to think about that dispersion, since the consolidated quote papers over it.
A crypto exchange has no equivalent tape. Each venue publishes its own book, each book is quoted in its own pair (the same coin against dollars, against a stablecoin, or against another coin), and the gaps between venues persist. One practical effect on order placement: the same limit price can be immediately fillable on one exchange and resting untouched on another, in the same second. "Better price" is a statement about one book, not about the market.
Post-only orders, and how they invert a marketable limit order
A limit order priced at or through the other side of the book is a marketable limit order. It crosses, trades immediately against resting liquidity, and pays the taker fee. Our guide to marketable and non marketable limit orders walks through that distinction in the equity context, and market orders versus limit orders covers the choice one level up.
Crypto exchanges add a flag with no common retail equity equivalent. A post-only limit order is an instruction to join the book or do nothing: if the price would cross and take liquidity at the moment it arrives, the exchange cancels the order rather than filling it. Same price, same side, opposite outcome from the marketable limit order. Some venues also offer a slide variant that reprices the order to the near touch instead of cancelling it. Reading that cancel as a rejection or a glitch is a common mistake. It is the flag doing precisely what it says.
Maker taker tiers make a resting order cheaper, not only better priced
In equities the fee gap between adding and removing liquidity is real, and for a retail customer it is usually invisible: the broker absorbs it, and may be paid for routing the order. On a crypto exchange the maker and taker fees are normally itemized on each trade and charged to the account at a tier set by 30-day volume. As of October 2026, published retail schedules commonly price the taker side at several multiples of the maker side, and top tiers at some venues pay the maker a rebate. Treat any specific figure as typical rather than current, and read the venue's own fee page on the day: these schedules change without notice. How maker taker fees and rebates work covers the economics.
What shifts is the arithmetic of crossing now versus resting. A trader who crosses the spread gives up roughly half the quoted spread against the midpoint, and that cost is measurable.
| symbol | avg_spread_bps | half_spread_bps |
|---|---|---|
| SPY | 0.21 | 0.1 |
| NVDA | 0.64 | 0.32 |
| AAPL | 0.79 | 0.4 |
| KO | 1.25 | 0.63 |
| MSFT | 1.6 | 0.8 |
The exact SQL behind every number
SELECT
ticker AS symbol,
round(avg(10000 * (toFloat64(ask_price) - toFloat64(bid_price))
/ ((toFloat64(ask_price) + toFloat64(bid_price)) / 2)), 2) AS avg_spread_bps,
round(avg(10000 * (toFloat64(ask_price) - toFloat64(bid_price))
/ ((toFloat64(ask_price) + toFloat64(bid_price)) / 2)) / 2, 2) AS half_spread_bps
FROM global_markets.cache_stocks_quotes
WHERE ticker IN ('AAPL', 'MSFT', 'NVDA', 'SPY', 'KO')
AND sip_timestamp >= '2026-09-23 18:30:00'
AND sip_timestamp < '2026-09-23 18:35:00'
AND toFloat64(bid_price) > 0
AND toFloat64(ask_price) > toFloat64(bid_price)
GROUP BY symbol
ORDER BY avg_spread_bpsAcross 5 household names in that same five-minute window, the quoted spread averaged 0.21 basis points at the tightest name (SPY) and 1.6 basis points at the widest (MSFT). The half-spread column is the rough price of demanding immediacy: 0.8 basis points at the widest name here. Put an itemized taker fee on top of that half-spread and the resting order wins twice, on price and on the bill. On a thinly quoted pair, where spreads run far wider than any of these equity figures, the gap between the two sides of the fee schedule is the smaller half of the story.
A 24/7 book has no open and no overnight queue
US equity volume is not spread evenly across the day. It clusters at the session edges, and outside the regular session it nearly vanishes.
| et_time | volume_millions | share_of_month_pct |
|---|---|---|
| 04:00 | 2 | 0.29 |
| 05:00 | 0.7 | 0.1 |
| 06:00 | 0.9 | 0.13 |
| 07:00 | 2.2 | 0.32 |
| 08:00 | 3.6 | 0.52 |
| 09:00 | 118.3 | 17.06 |
| 10:00 | 118 | 17.01 |
| 11:00 | 88.2 | 12.72 |
| 12:00 | 65.9 | 9.5 |
| 13:00 | 62.9 | 9.07 |
| 14:00 | 70.7 | 10.19 |
| 15:00 | 129.1 | 18.61 |
| 16:00 | 27.4 | 3.94 |
| 17:00 | 2.3 | 0.33 |
| 18:00 | 0.8 | 0.11 |
| 19:00 | 0.7 | 0.1 |
The exact SQL behind every number
WITH month_total AS
(
SELECT sum(toFloat64(volume)) AS all_shares
FROM global_markets.delayed_stocks_minute_aggs
WHERE ticker = 'AAPL'
AND window_start >= '2026-09-01 04:00:00'
AND window_start < '2026-10-01 04:00:00'
)
SELECT
formatDateTime(toStartOfHour(toTimeZone(window_start, 'America/New_York')), '%H:%i') AS et_time,
round(sum(toFloat64(volume)) / 1e6, 1) AS volume_millions,
round(100 * sum(toFloat64(volume)) / (SELECT all_shares FROM month_total), 2) AS share_of_month_pct
FROM global_markets.delayed_stocks_minute_aggs
WHERE ticker = 'AAPL'
AND window_start >= '2026-09-01 04:00:00'
AND window_start < '2026-10-01 04:00:00'
GROUP BY et_time
ORDER BY et_timeSeptember 2026's AAPL tape fills 16 hourly buckets, the earliest labelled 04:00 ET and the latest 19:00 ET. That earliest bucket carried 0.29% of the month's shares. The shape of the chart is the lesson: volume piles into the regular session and thins at both ends, and between the last bucket of one day and the first of the next, nothing trades at all. A limit order entered at 9 p.m. ET does not trade at 9 p.m. ET. It waits, and at many brokers it joins a queue that releases into the opening auction, where an order priced through the market can fill well away from the last price its owner saw.
A crypto book has no such edge. Matching runs continuously through nights, weekends, and holidays. There is no opening or closing auction print, no official close, and "market on open" has no referent. Scheduled maintenance windows and, on derivatives venues, funding timestamps are the only recurring clock features.
The time in force flags read differently too. GTC, good till cancelled, is the usual default on a crypto book, while IOC (immediate or cancel) and FOK (fill or kill) behave as they do in equities. A day order is the awkward one: with no close, a venue that offers it has to pick an arbitrary UTC boundary, and the definition varies by exchange. One habit worth unlearning is the equity reflex of flattening orders before the bell. There is no bell, and an order left resting across a weekend is live for every hour of it. Cancelling a resting limit order works the same way mechanically here, and comes up more often.
Partial fills and fractional quantities are the normal case
Crypto books quote in fractions of a coin. A minimum order size and a step size, the smallest quantity increment the venue accepts, replace the equity round lot of 100 shares. An order for 0.37 of a coin is ordinary, and it can come back filled 0.11 now and 0.26 an hour later, or 0.11 and never the rest. The remainder of a partially filled limit order keeps its price and its place in line at that price.
Equities have drifted the same way, which makes the habit easier to carry over than most of this list.
| size_bucket | prints | share_of_prints_pct |
|---|---|---|
| 1 to 9 shares | 19943 | 42.52 |
| 10 to 49 shares | 20301 | 43.29 |
| 50 to 99 shares | 3280 | 6.99 |
| 100 to 499 shares | 3118 | 6.65 |
| 500 or more shares | 256 | 0.55 |
The exact SQL behind every number
WITH hour_total AS
(
SELECT count() AS all_prints
FROM global_markets.stocks_trades
WHERE ticker = 'AAPL'
AND sip_timestamp >= '2026-09-23 18:00:00'
AND sip_timestamp < '2026-09-23 19:00:00'
)
SELECT
multiIf(size < 10, '1 to 9 shares',
size < 50, '10 to 49 shares',
size < 100, '50 to 99 shares',
size < 500, '100 to 499 shares',
'500 or more shares') AS size_bucket,
count() AS prints,
round(100 * count() / (SELECT all_prints FROM hour_total), 2) AS share_of_prints_pct
FROM global_markets.stocks_trades
WHERE ticker = 'AAPL'
AND sip_timestamp >= '2026-09-23 18:00:00'
AND sip_timestamp < '2026-09-23 19:00:00'
GROUP BY size_bucket
ORDER BY min(size)In that single hour of AAPL prints, the smallest bucket, 1 to 9 shares, accounted for 42.52% of the hour's trades across 19943 prints, out of 5 size buckets in total. Prints below a round lot are routine on the US tape, so the fractional quantity itself is familiar. What does not carry over is the expectation that a fill arrives in one piece at one price.
Does the pattern day trader rule apply on a crypto exchange?
The pattern day trader rule is a FINRA margin rule for margin accounts at US broker dealers trading securities. Spot trading at a crypto exchange sits outside it: the count of four day trades in five business days does not restrict the account, and there is no $25,000 equity minimum attached to day trading activity. Account-level leverage limits, venue terms, and tax treatment are separate questions with their own answers. Whether the PDT rule applies to crypto and futures goes through the account types where it does bite.
FAQ
Do crypto exchanges have limit orders?
Yes. Nearly every spot and derivatives venue supports a limit order with a price and a quantity, plus flags that control whether it may cross the spread and how long it lives. The instruction matches an equity limit order; the surrounding rules on fees, sessions, and venue-local pricing differ.
What does post-only mean on a crypto exchange?
Post-only tells the exchange to add the order to the book or cancel it. If the limit price would cross the opposite side and take liquidity on arrival, the order is cancelled instead of filled, which keeps the trade on the cheaper maker side of the fee schedule.
Is a limit order cheaper than a market order on a crypto exchange?
A resting limit order that gets filled is charged the maker fee, which is normally lower than the taker fee a market order or a crossing limit order pays. The saving is not free: a resting order may never fill, and the price can move away while it waits.
What happens to a crypto limit order overnight?
Nothing special happens, which is the point. A 24/7 book keeps matching through the night and the weekend, so a good till cancelled order stays live and fillable the whole time rather than queueing for a morning auction.
Is there a best price rule across crypto exchanges?
No consolidated quote spans crypto venues, and no best-execution obligation routes an order from one exchange to another's better price. Each book stands on its own, so comparing prices before placing an order falls to the trader.
Data notes
Every panel on this page measures US equities, which serve as the comparison baseline; the crypto mechanics described in the prose are venue rules, not measurements. The quote and trade panels are pinned to fixed windows on September 23, 2026 (14:30 to 14:35 ET for quotes, 14:00 to 15:00 ET for prints), so those figures do not move on regeneration. The volume profile covers calendar September 2026 in Eastern Time and groups by ET clock hour rather than assuming a session window. Spreads are quoted spreads, measured from the two-sided quote, not realized execution costs.
Each panel above ships with the exact SQL beneath it, so any figure here can be recounted end to end. To measure the spread or the size distribution on a name of your own choosing, ask the question in plain English on the Strasmore terminal.