Does Modifying an Order Lose Queue Priority?

Order modification and queue priority pull against each other. On US equity books and on most futures venues, changing the price of a resting limit order sends it to the back of the line at its new price, and increasing the displayed size does the same. Reducing the size keeps your place. Cancelling an order to send a fresh one is a new arrival with a new timestamp, whatever the button on your platform is called.

Does modifying an order lose queue priority?

Four rules cover almost every venue a self directed trader will touch.

• Changing the price loses priority. The order arrives at its new price as a newcomer, behind everything already resting there.
• Increasing the displayed size loses priority. Some venues re-queue only the added quantity and leave the original slice in place, others re-time the whole order. Neither hands the extra shares your old spot.
• Decreasing the displayed size keeps priority. The venue reads it as a partial cancel, and handing quantity back costs nothing in line position.
• Cancelling and re-sending loses priority every time, with a gap in between where nothing of yours is working.

The asymmetry has a logic. Time priority is what a book pays you for having stood at a price while everyone else waited. Adding size or moving the price commits new risk, and new risk starts at the back. Taking size away withdraws risk, and there is nothing to charge for. On a pro rata book the arithmetic changes, since allocation follows the size you show rather than the moment you arrived. Price time priority versus pro rata walks through both.

Modify in place or cancel and replace?

At the venue these are different instructions. An amendment, which your platform may call a modify or a replace, is one message naming the existing order and the change. It applies atomically: the change lands, or the order sits untouched. A cancel followed by a new order is two events with a gap between them, and during that gap you hold no place at all. If the market trades through your price inside the gap, you were not there for it.

Some venues implement an amendment internally as a cancel and a new order, and some brokers translate an amendment into a cancel and replace on the way out even where the venue supports a true in place change. The outcome is identical: the priority is gone. A rejected amendment leaves the original order resting, while a cancel that succeeds ahead of a rejected new order leaves you with nothing working. Even a priority preserving amendment usually comes back with a new order identifier, so your own records show a new order where the book still counts the old one.

What an amendment costs

Queue position is worth only what the tape can pay it. The shares ahead of you have to trade before you do, and the rate they trade at swings widely across a session.

SELECT
    formatDateTime(toStartOfFifteenMinutes(et), '%H:%i')                    AS et_time,
    round(toFloat64(avg(volume)) / 1000, 1)                                 AS avg_k_shares_per_min,
    round(quantileDeterministic(0.5)(toFloat64(volume),
                                     toUInt64(toUnixTimestamp(et))) / 1000, 1) AS median_k_shares_per_min
FROM
(
    SELECT
        toTimeZone(window_start, 'America/New_York') AS et,
        volume
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker = 'AAPL'
      AND window_start >= '2026-06-01'
      AND window_start <  '2026-07-01'
      AND (toHour(toTimeZone(window_start, 'America/New_York')) * 60
           + toMinute(toTimeZone(window_start, 'America/New_York'))) >= 570
      AND (toHour(toTimeZone(window_start, 'America/New_York')) * 60
           + toMinute(toTimeZone(window_start, 'America/New_York'))) < 960
)
GROUP BY et_time
ORDER BY et_time

Across June 2026 the quarter hour beginning 09:30 ET ran a median of 182.8 thousand AAPL shares a minute, against 62 thousand in the bucket beginning 12:00. Same order, same price. The wait in front of it is a different animal at the open than in the middle of the day.

Fill probability before and after an amendment

Amendments are priced in probability. The panel below counts, over a month of regular session minutes, how often a minute of tape carried at least a given number of shares. Read it as the chance that a queue of that size in front of you clears inside one minute.

WITH minute_volume AS
(
    SELECT
        ticker,
        toStartOfMinute(toTimeZone(window_start, 'America/New_York')) AS et_minute,
        sum(volume)                                                   AS shares
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker IN ('AAPL', 'KO')
      AND window_start >= '2026-06-01'
      AND window_start <  '2026-07-01'
      AND (toHour(toTimeZone(window_start, 'America/New_York')) * 60
           + toMinute(toTimeZone(window_start, 'America/New_York'))) >= 570
      AND (toHour(toTimeZone(window_start, 'America/New_York')) * 60
           + toMinute(toTimeZone(window_start, 'America/New_York'))) < 960
    GROUP BY ticker, et_minute
)
SELECT
    q.ahead                                                                                           AS shares_ahead,
    round(100 * countIf(mv.ticker = 'AAPL' AND mv.shares >= q.ahead) / countIf(mv.ticker = 'AAPL'), 1) AS aapl_pct_of_minutes,
    round(100 * countIf(mv.ticker = 'KO'   AND mv.shares >= q.ahead) / countIf(mv.ticker = 'KO'),   1) AS ko_pct_of_minutes
FROM minute_volume AS mv
CROSS JOIN
(
    SELECT arrayJoin([500, 1000, 2000, 4000, 8000, 16000, 32000, 64000, 128000]) AS ahead
) AS q
GROUP BY q.ahead
ORDER BY q.ahead

With 500 shares ahead, 100 percent of AAPL minutes carried enough volume, against 100 percent of KO minutes. At 8000 shares the AAPL figure is 100 percent, and out at 128000 shares it falls to 22.2 percent. The curve only slopes one way, and losing priority is nothing more than sliding to the right along it.

Treat the line as an optimistic ceiling. A minute of volume prints across many prices and both sides of the market, and only a slice of it lands at your price level, so one real queue clears more slowly than this. Minutes with no prints at all are not counted. Estimating your queue position tightens the picture with depth and cancel rates.

The break even re-price distance

Now put a price on the amendment. Take a hypothetical buy order resting with 8,000 shares ahead of it and a 60 percent chance of clearing inside the next minute. Improving the bid by one cent leaves it alone at a new best price with nothing in front: call the odds there 99 percent. That improvement costs one cent per share on every fill.

Write V for what a fill is worth per share, your edge plus any rebate, and t for the re-price distance. The amendment breaks even when the improved odds times the reduced payoff match the old odds times the full payoff, which rearranges to a single line:

t = V x (1 - p0 / p1)

With a fill worth 3 cents a share and odds moving from 60 to 99 percent, the break even distance lands near 1.2 cents. A one cent improvement pays for itself. A two cent chase does not.

Run the same line for a bot whose only edge is the maker rebate. Top tier maker rebates at the large US equity exchanges sit near 0.20 to 0.30 cents a share on published fee schedules as of mid 2026, so call it a quarter of a cent. The formula puts the break even distance near 0.1 cents, and the smallest price change any US venue accepts is a full cent, about ten times too far. A bot re-pricing to protect a rebate pays a tick to chase a quarter of one.

The rebate is paid on fills, never on intentions. An order that holds a beautiful queue position and never trades earns nothing, and a rebate you never collect is not a saving. Maker taker fees and rebates sets out how those schedules pay.

When there is no room to improve

Improving a price is only an option when the bid and the offer have room between them. In the largest names, most of the time, they do not.

SELECT
    ticker                                                                                  AS symbol,
    round(avg(toFloat64(ask_price) - toFloat64(bid_price)) * 100, 2)                        AS avg_spread_cents,
    round(100 * countIf(toFloat64(ask_price) - toFloat64(bid_price) <= 0.0105) / count(), 1) AS one_tick_wide_pct
FROM global_markets.cache_stocks_quotes
WHERE ticker IN ('AAPL', 'KO', 'SPY', 'NVDA', 'COST', 'NFLX')
  AND sip_timestamp >= '2026-06-10 13:30:00'
  AND sip_timestamp <  '2026-06-10 20:00:00'
  AND bid_price > 0
  AND ask_price > bid_price
  AND toFloat64(ask_price) - toFloat64(bid_price) < 5
GROUP BY ticker
ORDER BY one_tick_wide_pct DESC

NFLX quoted a one cent spread across 81.4 percent of its quote updates on 10 June 2026, averaging 1.3 cents wide. At the other end of the panel COST averaged 77.53 cents. In a one tick market there is no price between the bid and the offer to move to, and the only route in front of the queue is to cross the spread. That turns you into the taker: you pay the taker fee instead of earning the rebate, and you give up the spread rather than a tick of it.

How often a re-pricing bot faces the choice

A follow the market bot does not make this trade once. It makes it every time the top of the book moves away from its order.

SELECT
    formatDateTime(toStartOfFifteenMinutes(et), '%H:%i') AS et_time,
    uniqExactIf(bid_price, ticker = 'AAPL')              AS aapl_distinct_bids,
    uniqExactIf(bid_price, ticker = 'KO')                AS ko_distinct_bids
FROM
(
    SELECT
        toTimeZone(sip_timestamp, 'America/New_York') AS et,
        ticker,
        bid_price
    FROM global_markets.cache_stocks_quotes
    WHERE ticker IN ('AAPL', 'KO')
      AND sip_timestamp >= '2026-06-10 13:30:00'
      AND sip_timestamp <  '2026-06-10 20:00:00'
      AND bid_price > 0
)
GROUP BY et_time
ORDER BY et_time

The best bid in AAPL visited 359 distinct prices inside the quarter hour beginning 09:30 ET on 10 June 2026, and 152 inside the quarter hour beginning 15:45. The KO best bid visited 78 prices in that opening bucket. Every one of those moves is an amendment a re-pricing bot might send, and every amendment restarts its wait. Grid trading bots take the other side of the choice by parking a ladder at fixed prices: the prices go stale, the priority survives.

Icebergs and self trade prevention

Two second order effects catch people out.

Reserve orders, usually called iceberg orders, display a slice and keep the rest hidden. On most books each refresh of the displayed slice joins the back of the queue at that price, so a large iceberg behaves as a run of small orders that keep surrendering their place to fully displayed size. Amending the display size restarts the same cycle.

Self trade prevention is the other one. Venues will not let one participant's order execute against its own resting order, and the usual remedy is to cancel one of the two, or on some venues both. A bot re-pricing across a narrow book can knock out its own quote on the far side, which turns one amendment into two lost places in line. Amendments normally inherit the original time in force while a cancel and replace resets it, a detail covered in order time in force.

FAQ

Does changing the price of a limit order lose my place in the queue?

Yes. On US equity books and most futures venues the order counts as new at its new price, behind every share already resting there. The one change that keeps priority is a size reduction at the same price.

Is cancel and replace the same as modifying an order?

They differ at the venue. An amendment is a single message and can keep priority when only the size falls. A cancel and replace is two events with a gap where nothing of yours is resting, and the replacement always joins the back of the queue.

Does reducing the size of an order keep queue priority?

Yes, at the same price. Venues treat a size decrease as a partial cancellation, and the remaining shares hold their original timestamp. Increasing the size does not work the same way.

What is queue position actually worth?

The fill probability it buys, multiplied by what a fill is worth to you. Both halves can be measured: the panels above show how much volume prints per minute, and your own edge plus any rebate sets the value of a fill.

Every panel here carries the SQL that produced it. Change the ticker or the date range and ask the same question of any name on the Strasmore terminal.