What amending an order actually does
You nudged a resting order’s price by one increment to be more competitive, and it stopped filling while orders posted after yours went through.
Amending an order is rarely an edit. On most venues it is a cancellation followed by a new order, and the new order goes to the back of the line. Which changes trigger that, and which do not, is the whole subject.
Queue position is a thing you own
At each price level in the book there is a queue, and under price-time priority the orders at the front of it fill first. Queue position is therefore an asset: it determines whether you trade when that level is partially consumed, or watch the level get hit without being reached.
It was earned by arriving early, and it is not attached to you or your account. It is attached to that specific order at that specific price level, and anything that makes the engine treat your instruction as a new order destroys it.
This is why the innocuous-sounding operations matter. The question for every amend is not “did the order change” but “does the engine now consider this a different order”.
What loses position and what doesn’t
The general principle: a change that could take priority from someone else loses your priority. A change that cannot, may keep it.
Changing the price — always loses position. A different price level is a different queue. There is no sense in which you could be early in a queue you have just joined, so the order goes to the back of the new level. This is true even when the new price is worse than the old one.
Increasing the quantity — normally loses position. Letting an order grow while keeping its place would let anyone buy priority retroactively by posting small and expanding. Most venues treat the increase as a new order, and many treat the entire amended order as new rather than only the added part.
Decreasing the quantity — normally keeps position. Reducing size takes nothing from anyone behind you; it gives them more chance of being reached. Venues commonly allow this in place, retaining the original timestamp.
Changing time in force or flags varies. Some are treated as metadata edits, some as replacements.
None of this is universal. Some venues implement no amend at all and require you to cancel and resend explicitly, which at least makes the consequence visible. Some implement an amend that is internally a cancel-replace but reports as a modification, which hides it.
Cancel-replace and the gap between the two
If an amend is a cancel followed by a new order, there is an interval between the two, and things can happen in it.
The order is not in the book during the gap. If the market reaches your price in that window, you do not trade. On a continuous engine the gap is small; it is never zero.
The cancel can fail. If the original filled before the cancel was processed, the replacement is being applied to something that no longer exists in the form you assumed.
The replacement can be rejected. A price that has become invalid under a price band, or a size below a minimum after your reduction, fails validation — and now you have neither the old order nor a new one.
That last case is the one worth internalising: the failure modes of an amend include ending with no order at all. Venues that support an atomic amend — where the change either applies or the original survives — remove that specific risk. Whether yours does is a venue question.
The mechanism
THE MECHANISM — amending a resting order
· Change the price
→ new queue at the new level.
Position lost, always, in both
directions.
· Increase the quantity
→ generally treated as a new
order. Position lost.
· Decrease the quantity
→ generally kept in place with the
original timestamp.
· The interval between cancel and replace
→ NOT IN THE BOOK. A move through
your price during the gap does
not fill you.
· Replacement fails validation
→ you hold no order at all. The
original is already gone.
· Amend on an order that just filled
→ cancel leg fails. You have the
position you were editing away
from.
· Atomicity, which fields preserve
priority, and whether amend exists
→ VENUE-SPECIFIC. Some have no
amend; some hide cancel-replace
behind one.Worked example
Illustrative figures, synthetic throughout. Suppose the bid side has 40,000 with 6.0 units resting, and your buy order for 1.0 sits third in that queue with 2.0 units ahead of it.
A seller crosses with 2.5 units. The first 2.0 fills the orders ahead of you; the remaining 0.5 fills half of yours. You get a partial fill because you were early.
Now suppose that instead, a moment earlier, you had amended the order to 40,001 to be more competitive. You are now alone at a better level — genuinely first in a new queue — and the same seller crossing 2.5 units hits your 1.0 in full before reaching 40,000. Moving the price up helped.
Third case: you amended from 40,000 to 39,999 because you wanted a better entry. You are now at the back of the 39,999 queue, behind everything already there. The 2.5-unit seller never reaches you at all. You moved one increment and went from a partial fill to nothing — not because 39,999 is a bad price but because you rejoined a queue.
Fourth case: you amended the size from 1.0 up to 2.0 at the same price. On a venue that treats increases as new orders, you are now last in the 40,000 queue with 6.0 ahead of you, and the same seller does not reach you either. You did not change your price and you still lost the fill.
Why this is not a technique
It would be easy to read the above as a suggestion about how to manage orders. It is not, and the reason is worth stating.
Improving queue position improves the chance of being filled, and being filled is not a good outcome or a bad one in itself — it depends entirely on what happens next, which this site does not comment on. The order that did not fill was sometimes the fortunate one. What is mechanically true is only this: priority is real, amending usually forfeits it, and forfeiting it is invisible in the interface.
The failure mode
Amending an order presents as a small adjustment and executes as a withdrawal plus a resubmission. The interface typically shows one row that changed, which is a faithful description of your intent and a poor description of what the engine did.
The specific way it bites: an order that has been amended repeatedly in a moving market may have spent much of its life at the back of successive queues, and none at the front of any. It reads as an actively managed order and behaves as a series of freshly posted ones. If it is not filling, the history of your own amendments is the first place the explanation lives — not the book’s depth and not the venue.