Trend-day continuation

M5.11SetupsPractitionerLoreOne trader's study≈ 16 min

Learning objectives

Write the trend-day playbook as a checkable rule: with-trend pullback entries only, a per-trade stop, a trailing exit, and a written condition that withdraws the day context.
Explain why a trend day is an end-of-day label, and what the NQ tests say about calling it by late morning.
Read the late-day momentum evidence correctly: measured on SPY 1993–2013, absent in our lab on QQQ/SPY since late 2023, not replicated on NQ.
Size and manage a pullback long and its short mirror, and say what a trailing exit costs on the days that do not trend.

Setup: Trend-day continuation (tiếp diễn trong ngày xu hướng).

It is 12:00 ET (00:00 ICT, winter). NQ opened at 18,040, drove to 18,100 and has spent the late morning above VWAP, every dip shallow. Every reversal setup from M5.01 to M5.06 has fired at least once, and every one would have lost. If today looks like a trend day, what is the plan, and how much of “looks like” can you trust? The plan: with-trend pullbacks only, a stop on every trade, a trailing exit, and a written condition that ends the playbook. The trust: the tests say you cannot know it is a trend day until the close.

Key terms

Trend day (ngày xu hướng, owned by M4.03): a session that opens near one extreme and closes near the other, with value migrating one way all day and few rotations back through VWAP Definition. Note what the definition needs: the close. A trend day is a label you can only apply after 16:00 ET. Before that you have a with-trend context: a morning that has behaved like the start of one.

Term you need: accepted rotation. In this lesson an accepted rotation is price trading back through VWAP and staying there: a 30-minute bar closes on the other side and the next bar does not reclaim it. A wick through VWAP that closes back on the trend side is a probe, not a rotation.

Chart first: one session, read left to right

ANIMATED FIGURE

A synthetic NQ session that ended as a trend day

Lore

WHAT TO SEETrend day up: long only, on pullbacks toward VWAP; the trail sits under the last pullback low.

Synthetic data

How it works

The playbook, written as a rule

This is a context rule, not a pattern: it switches some setups off and one on. The switch is a house rule House rule, a choice about which trades you allow, not a measured edge.

Context: the morning has driven one way and held: the opening move has not been retraced, price has traded beyond the initial balance and stayed there, pullbacks have stopped short of VWAP or at it. Write the criteria down before the open (M4.03). If you are unsure whether the criteria are met, they are not.
Allowed trades: with-trend pullbacks only: to VWAP (M5.08), to the developing value-area edge, to the prior balance. The reversal setups of M5.01–M5.06 are switched off for the day. The level is still there; the context vetoes the trade (M5.13).
Invalidation, per trade: each pullback long has its own stop below VWAP or below the pullback low, set before entry. One stopped pullback is a trade loss, not the end of the day context.
Exit: trail. Move the stop under each new higher pullback low (longs). The trail is a house rule House rule, not a measured improvement: it keeps more on days that do end as trend days and gives more back on days that rotate.
The day context ends on the first accepted rotation through VWAP, or when the other side reclaims the initial balance. After that, take the trail’s exit and stop taking with-trend entries. A stop-out is not a signal to flip short.

The catch: a trend day is known at the close

Myth / hypothesis: “By late morning you can tell it is a trend day: open drive, narrow IB, broken IB, price holding above VWAP. From there, just ride it.” Lore Each piece has been tested on NQ in a public practitioner study, and none of them forecasts the day type. Calling the day type by 11:00 ET from the open, the IB and the open location did no better than without them. Open-drive follow-through was inconclusive. A narrow IB went with fewer trend days, not more, the opposite of how it is usually taught. Cumulative delta “confirming” the trend on trend days was inconclusive too.

One finding looks like support and is not: on days that ended as trend days, price held one side of VWAP after 10:30 ET more often than on other days (+11.6 pp [+7.9, +15.3], n = 1,818, same study) Definition. That describes trend days with the end-of-day label already known; it cannot tell you at noon that today is one. So the playbook is a bet on a context that may end at any bar, which is why every trade carries its own stop.

The afternoon: what is measured and what is not

An earlier version of this lesson called the afternoon pullback the best entry of a trend day. That claim is untested Lore. Here is what is known about the afternoon:

Intraday momentum (động lượng trong phiên): on SPY from 1993 to 2013, the first half-hour’s return predicted the last half-hour’s return (Gao, Han, Li and Zhou, JFE 2018) MeasuredOther market. That is a historical result on another market.
Recently it has not paid. Our lab re-ran the idea on QQQ and SPY hourly bars from December 2023 to November 2026: the morning return did not predict the last 30 minutes (slope t-statistics −0.8 to +1.6, sign hit rate 44–52%) Measured: no edgeNasdaq index.
On NQ it has not been replicated. The public practitioner study found the first half-hour told you nothing about the last half-hour on NQ: 0.0 pp [−2.3, +2.3], n = 1,811, with a different definition from the paper’s. Applied to NQ, the effect is plausible at best, extrapolated from SPY PlausibleOther market.
Dealer gamma. On the S&P 500 from 1996 to 2020, late-day momentum was stronger on days when dealers were short gamma (Baltussen, Da, Lammers and Martens, JFE 2021) MeasuredOther market. That is a condition on gamma, not a trend-day setup, and on NQ the same idea is inconclusive with too few days (M8.06).
Closing flows. On NQ, leveraged-ETF rebalancing amplifying the last hour was not supported (−4.6 pp [−10.2, +1.0], n = 1,811), and “the last hour trends with the day” was inconclusive (+2.8 pp [+0.5, +5.1], n = 1,812) Lore.

So the afternoon is no reason to switch off at noon and no reason to expect more: each afternoon pullback gets the same rule and stop as a morning one.

Why it is still worth knowing

The playbook earns its place without an edge claim. It stops you fading a market that keeps making new highs, the usual way a trend day hurts a reversal trader. It puts stops where they are cheap: below a pullback that held VWAP. And it turns “the trend is over” into a written condition you can check in a journal.

What the data saysLoreOne trader's study

You can recognise a trend day by late morning (open drive, narrow or broken IB, price holding one side of VWAP) and trade with it for the rest of the session.

Tested in pieces on NQ: none of the pieces forecasts the day type. The playbook as a whole is not yet measured.

Result: day type called by 11:00 ET from open, IB and location: +0.6 pp; open type predicting direction: +4.8 pp; narrow IB leading to a trend day: −11.9 pp (fewer trend days, not more)
Sample size: 625; 1,290; 1,809
Interval: [−3.3, +4.5]; [−0.2, +9.8]; [−14.8, −9.0]
Sample: NQ, every trade 2018–2026; effect = hit rate with the signal minus without (claims AP-08, OI-02, OI-04)

Three caveats: one practitioner's pre-registered study; intervals not clustered by session; a study like this caps at Plausible even when it supports a claim.

Price holding one side of VWAP after 10:30 ET on trend days (+11.6 pp [+7.9, +15.3], n = 1,818) is descriptive: it is measured with the end-of-day label already known.

ES was not tested; the grade is set on NQ results.

What the data says about the afternoonPlausibleOther market

The morning's direction predicts the last half-hour, so the late session of a trend day pays.

Measured historically on SPY; absent in our lab on QQQ and SPY since late 2023; not replicated on NQ. Applied to NQ: Plausible, extrapolated.

Result: SPY 1993–2013: the first half-hour return predicts the last half-hour (peer-reviewed). QQQ/SPY 2023-12 to 2026-11: slope t-statistics −0.8 to +1.6, sign hit rate 44–52%. NQ 2018–2026: 0.0 pp.
Sample size: 721 sessions (our lab); 1,811 (NQ study)
Interval: NQ: [−2.3, +2.3]
Sample: SPY (Gao et al.); QQQ and SPY hourly bars, prior close or open to 10:30 ET predicting 15:30–16:00 (our lab); NQ first to last half-hour (practitioner study)

Our lab's test used hourly ETF bars, so the first half-hour was proxied by the move to 10:30 ET; the result agrees with the NQ practitioner test for the recent period.

The NQ test used its own window and outcome, not the paper's regression, so it is not a like-for-like replication. Running the paper's regression on NQ is on the lab's backlog.

The dealer-gamma version (Baltussen et al., JFE 2021, S&P 500 1996–2020) is a separate, gamma-conditioned result; on NQ it is inconclusive (n too small), direction as predicted.

Worked example: a 100-point day, long side

A synthetic NQ session (numbers invented for teaching; no real date), the one in the figure. Open 09:30 ET (21:30 ICT) at 18,040; drive to 18,068, pullback to 18,060 above VWAP, then 18,100 by 12:00 ET with VWAP at 18,075 and no pullback over 12 points. Your written context criteria are met, so the reversal setups are off: two “double tops” and an exhaustion wick printed, and you passed on all three.

14:00 ET (02:00 ICT): price pulls back to 18,082, where VWAP (18,081) meets the developing value-area high. The pullback bar’s lower wick holds 18,080 and the next bar trades back above 18,085.

Context: the with-trend context is in force (criteria met, no accepted rotation yet).
Trigger: pullback to VWAP plus the value-area edge, and a bar that holds it.
Entry: long 18,082.
Invalidation: 18,074, below VWAP and the pullback low. 8 points: $160 on 1 NQ ($20 a point) or $16 on 1 MNQ ($2 a point), before costs.
Exit: trail. Each new pullback low moves the stop up.

By 15:00 ET price reaches 18,140, fails to hold it, and the 15:30 bar closes back under 18,130: the trail, by then at 18,128 under the last pullback low, fills. Result: +46 points, $920 on 1 NQ, about 5.75R before costs.

The shorts at 18,090, 18,095 and 18,100 that the reversal setups offered were all stopped out; the context veto kept you out of them. But this is the day that ended as a trend day, and the 12:00 chart could not have told you so.

The mirror and the failure

Short side. Flip the figure. NQ opens at 18,140 and drives down; price stays below VWAP all morning and bounces stall at or under it. At 14:00 ET a bounce reaches VWAP and the developing value-area low at 18,098, and the bar that tags it closes back down. Short 18,098, stop 18,106 above VWAP and the bounce high (8 points), trail above each lower bounce high. The day context ends on the first accepted rotation back above VWAP.

The failure. Another synthetic morning, built to look the same at noon: drive from 18,040 to 18,100, price above VWAP, shallow pullbacks. You buy the 13:00 ET pullback at 18,084 with the stop at 18,076. At 13:40 price trades through VWAP, the stop fills (−8 points, $160 on 1 NQ), and the 14:00 bar closes at 18,071 with the next bar failing to reclaim VWAP: an accepted rotation, so the day context ends and you take no second long. The session closes at 18,055, mid-range: a normal day with a strong morning. The noon chart could not tell the two days apart, which the tests above say is the usual case; the stop and the end condition kept the cost to one loss.

Prop-firm note. A trailing stop is order handling. In an evaluation account your stop and limit orders fill by the simulator’s rules, not a real queue, so a sim fill can look kinder than a live one, and firms forbid exploiting simulated fills (PROP1.04, in the Prop Firm course). Size from the first stop distance: that is the loss your daily limit must absorb. Read your firm’s help-center article on prohibited conduct before you trade a firm account.

Common mistakes

Common mistake: Treating the end-of-day label as a forecast: “trend days hold one side of VWAP after 10:30, so today, holding above VWAP, is a trend day.” The first half is measured on days already known to be trend days. On an open day, holding above VWAP is a context to trade with a stop, not a prediction.

Common mistake: Fading because “it’s extended.” On a with-trend day the reversal setups are switched off by your own rule. If you cannot sit out the reversals, trade the day as a normal day with normal size, but do not run both playbooks at once.

Common mistake: Turning a stop-out into a flip. One stopped pullback long is a trade loss. An accepted rotation ends the with-trend context. Neither one is, by itself, a short signal; a new idea needs its own written setup.

Drill: score your own trend-day calls

Over 20 sessions of NQ replay, stop the replay at 12:00 ET each day and write one line: “with-trend context: yes or no”, using the criteria from M4.03. Then play the session to the close.

Label each session at the close: trend day (opened near one extreme, closed near the other) or not.
Count how often your noon “yes” ended as a trend day, and how often a noon “no” did.
For each noon “yes”, log the first pullback entry, its stop and the trail exit, and total the R.

Solution

Expect your noon “yes” to be right less often than it feels. With 20 sessions, a rate near 50% has a 95% interval of about ±22 percentage points, so the drill cannot prove a skill. The R column is the useful part: what the playbook cost on the “yes” days that did not trend, weighed against the trend days it caught.

What we don’t know yet

Whether the playbook as a whole (with-trend pullback entries, trailing exit, written end condition) has an edge on NQ after costs. Not yet measured.
Whether trailing beats a fixed target on NQ days that start like trend days. Not yet measured; the same practitioner study found wider fixed targets do not rescue a near coin-flip signal after costs.
Whether intraday momentum holds on NQ with the paper’s own regression. Not yet measured; it is on the lab’s backlog.

Recap

A trend day is an end-of-day label. During the session you have a with-trend context, written as criteria before the open.
The playbook is a house rule: with-trend pullback entries only, a stop on every trade, a trailing exit, and the first accepted rotation ends the context.
Calling the day type by late morning was tested on NQ and did not work; a narrow IB went with fewer trend days, not more.
Late-day momentum was measured on SPY 1993–2013, has been absent on QQQ/SPY since late 2023 and is not replicated on NQ: no reason to expect the afternoon to pay more than the morning.