vwap_pullback_continuation (long) · 30-min trigger, daily contextTrending since the March YTD low: reclaimed the 20- and 50-DMA in April, then held the 20-DMA for 41 straight sessions into the June high. First character change came on the loss of the 20-DMA and the break of the 5/19 low (~732); the bounce was rejected at prior highs and SPY fell to a new low into the YTD VWAP in early July.
Since then: a month of consolidation between 756–729 back above the 20- and 50-DMA, a tighter range 7/23–7/30, an upside break on Friday 7/31, and follow-through Monday (+1.5%) pressing through the top of the larger range.
Set its YTD low in late March and trended two-plus months after reclaiming the 20-DMA, 50-DMA and YTD VWAP in early April. First crack: broke the 5/26 gap lows in early June, then lower highs through June and July, new lows under the 50-DMA on 7/17, and a decline into a probe of the YTD VWAP late in the month.
Monday it bounced +1.9% off that probe — back above the 20-DMA (~700) but still below a falling 50-DMA (~715), no higher high yet.
QQQ/SPY broke below 0.9498 on 7/7 while QQQ was still above 706.92 — relative strength led the weakness, and QQQ confirmed by breaking 706.92 on 7/16. That is a rotation tape: money leaving semis, looking for a home. CRWD is software/cyber, not semis. The long side of a rotation is where you want a range break, and this one came with SPY breaking out the same day.
Backdrop into 7/31: pre-open 2026 Threat Hunting Report (AI across adversary operations); midday Fal.Con 2026 sold out at record pace; Street PT hikes still in the tape (MS $227, Argus $230, GS $208). CFO Form 144 (~7.6k) = noise against that.
Honest read on volume: 0.93× RVOL — the break did not have volume confirmation. It was tagged, not gated. 8/3’s 10.3M-share, +6.12% session is what supplied the confirmation after the fact.
The early dip-buy stopped out. An entry at 191.15 before the base formed took a full −1R. That was the knife. Same stock, same idea, four sessions early — and it is the single best argument for waiting for the quiet.
The stop came from structure. 187.98 was the hold low on the 30-min — the level that defines the setup. Not an ATR multiple, not a round number. If that low breaks, the thing you thought was happening is not happening.
All three frames agreed before size went on: daily quiet 7/28 → 30-min trigger 7/31 → daily confirm 8/3. The alignment is the grade.
BB(20,2) · SMA 20 (red) / 50 (blue) · 5-DMA (gold) · stacked RSI(10) + 10-MA + RSI-BB(10,2). All bars split-adjusted for the CRWD 4:1 on 7/2 — the raw feed is not, and an unadjusted 50 DMA reads 465 against a 211 close.
Into 7/31 the daily was still reading COMPRESSION — closes walking back up the bands, price re-approaching a flat 20-day from below, the flag saying “coil: break PDH or reclaim VWAP.” The PDH break printed on the daily as Day 1 of the range break and closed 190.86, right at the 20-day.
But there was no daily entry on 7/31. The 30-min owned it. The daily’s job was to say this is a place to be interested; the trigger frame’s job was to say now, and here is your stop.
8/3 closed +6.12% at 202.54, broke the coil and flipped the regime to EXPANSION UP. That flip was a fresh signal — a daily coil-break long at 202.54 — and it was not taken.
8/4 ran to 212.64 and closed 211.22. The daily confirmed the trade it never had to trigger.
True 5-day basis (65 × 30-min bars) with ±1σ and ±2σ, plus session VWAP. This is where the swing lives — the frame that gave the entry is the frame that gives the stop.
EXPANSION UP printed on the 1σ break and the +2.65% thrust carried price through 189.62 prior resistance. The retest held into the close — which is the tell that the break was real even without volume behind it.
Then the frame did the one thing that lets you hold: it never printed “overextended” while the move was still owed. Price only went stretched near the measured move — exactly on schedule.
| Prior-day high / low | 191.65 / 184.16 |
| Prior close | 190.86 |
| Prior resistance cleared | 189.62 |
| Entry (box break) | 191.21 |
| Stop (hold low) | 187.98 |
| Target (2R) | 197.68 |
Because the structure that makes the trade is the structure that kills it. The hold low is the price at which “price pulled in to the line and held” stops being true. An ATR stop would have been a number with no relationship to the idea; a round number would have been someone else’s idea.
Risk 3.23 on a 191.21 entry = 1.7% of price. At $100 risk per trade that is 30 shares. Size is an output of the stop, never an input.
2-min bars, session VWAP + ±1σ/±2σ, with every reconciled round trip marked. Lower panel = cumulative realized from those trades. All times ET.
The plan lines sit still. The fills do not. Across both sessions there are — reconciled round trips on — shares, and the day-session total is —. The swing made money. The trading around it did not.
Banked 2R at 197.68 while the trend paid 6.63R to 212.64. Left 4.6R on the table.
And the 8/3 re-add fired — a fresh daily coil-break signal at 202.54 — and was skipped. Not a missed setup: a signalled setup, not taken.
Note the asymmetry: the rule for getting in was written down in advance. The rule for staying in was not. That is the whole difference.
| Rule | On CRWD | Tested? |
|---|---|---|
| Take the full 2R and be done | +2.00R — what happened | baseline |
| Trail the fast line past 2R | would still be in as of 8/4 | no |
| Hold to the projected move, then manage | objective 209.15, tagged 8/4 | no |
Until those run over every historical firing, “hold longer” is a feeling with one good anecdote behind it. One trade that would have paid 6R does not make a rule — the same rule has to be checked against every time it gives back a 2R winner.
The lesson in the pairing: the break was quiet, the confirmation was loud, and the quiet one was the one that had to be traded. Waiting for the loud day means paying 202 instead of 191.
It picks the right average for each chart — 20-day on the daily, 5-day intraday, VWAP under 30 minutes — and draws bands showing how far price normally strays from it.
Why it mattered here: called the quiet 7/28 · woke the VWAP play on the 7/31 PDH break · never warned “too stretched” while the move was owed · daily confirmed 8/3.
The tool tallies VWAP L: +0.48R average over 168 fills, 26% “win”. That line needs two corrections before it can be said out loud.
1. The 26% is the target-hit rate, not the win rate. It scores a W only at the full target and an L only at the stop; everything trailed, covered or timed out counts in the total and in neither column. Proof from those two numbers alone:
2. Those 168 are not “this exact setup.” They are every VWAP pull-in long on the chart — no coil filter, no PDH filter, no alignment filter. This trade had all three. So 168 is the sample for the entry mechanic, not for this setup. The filtered number does not exist yet.
Also: one symbol, one chart (in-sample); the engine holds one position at a time, so signals arriving while it was busy were skipped — 168 is “signals that got a slot.”
The engine exits at its target. Every one of those 168 was closed at target, stop, trail or timeout — none was ever allowed to run past the objective. So the scoreboard measures getting in. It has never measured staying in.
Do not put +0.48R next to “I should have held for 6R.” They answer different questions, and side by side they imply the stat backs the hold rule. It does not.
| v | What went wrong | What got built |
|---|---|---|
| 4.2 | On a trend day price never returns to VWAP, so re-entries dried up | the 50-minute line as the pull-in rail |
| 4.5 | The NVDA mistake — “stretched” fired mid-move and scared out the hold | stretched must be earned |
| 5.2 | The plan only lived in text — a label and a table | entry/stop/target drawn on the chart, frozen at exit |
| 6.0 | No code for the NVDA reclaim; a 26-bar timer force-flattened swings; the Win% column produced the bad stat | swing build — RECLAIM entry, projected-move target, honest W-L-S table, JSON alert capture |
→ adaptive_mean_swing_v6.pine · → full tool playbook — the v6 build compiles but is not yet backtested; the RECLAIM entry has never fired.
The same 30-minute candles under every lens — only the overlays, markers and explanation change. Signals here are computed bar-by-bar by the engine port (σ v5.10 + AM ATR v4.21-CL, completed bars, confirmed prior-day ATR, split-adjusted). CRWD is the TYPE A — CONFLUENCE case: both engines fired on the same morning. The companion case where only one fires is VZ 8/13 — Type B.
The canonical MTF Strategy Lens mounted in this page (house chart theme: white, hollow black candles; σ + AM ATR renderers; chart TF 1m → Daily; reconciled executions; Candle / Event Inspector). Full page: lens.html?ticker=CRWD&date=2026-07-31.
Same skeleton, three months apart: pullback to a big MA → reclaim the 5 DMA → ride.
| NVDA · May 2026 | CRWD · Jul–Aug 2026 | |
|---|---|---|
| Pullback into | rising 20 DMA ~199–200 | 50 DMA — 7/29 closed 179.38 vs 179.80 |
| Trigger | 5-DMA reclaim 5/6–5/7 | 5-DMA reclaim 7/30 (185.22 vs 181.96) |
| Result | ran to 225.83, ~+38% off the base | ran to 212.64 by 8/4, +18.7% in four sessions |
| The failure | managed a daily swing on the 1–2 min; 12+ round trips on 5/4, −$59.98 | banked 2R while the trend paid 6.63R; skipped the 8/3 re-add |
NVDA’s mistake was managing a multi-day hold too tight and giving it back in round trips. CRWD is the same disease at a smaller dose.
And the reconciled fills say it a third time. Both CRWD day sessions lost money (— combined) while the position made it. The hold made the money. The churn gave it back.
Three months, two names, one unfixed habit. That is not a technique problem — the read was right both times. It is a rule problem.
Every reconciled round trip in CRWD across 7/31 and 8/3 lost money in aggregate, while the overnight hold and the swing piece made it. The deck’s “53% win” on 7/31 comes from mixing the overnight into the day count — the day session alone was — winners on — round trips.
This is not a scolding, it is the thesis of the whole playbook: the edge in this name was positional, and every attempt to also trade it intraday taxed the position.
| Day | In | Out | Sh | Entry | Exit | Held | P&L |
|---|
kti price minute / daily (karb). Daily bars are split-adjusted for the
CRWD 4:1 on 2026-07-02 — the raw feed is unadjusted and returns a 50 DMA of 465
against a 211 close if you don’t correct it.
P&L: kti trades date matched trades (Gr8Trade), account 14859 — not a local
FIFO over the order log, which mis-pairs because CRWD was carried overnight and produces a nonsense
+$6,208. Our per-trade sum reconciles to the raw witness sum to the cent on both days
(— difference). Timestamps are epoch-ms UTC converted to ET (UTC−4, EDT),
verified against the 15:50 ET tail.
Not reconciled here: the +$68.36 overnight scale-out and the options P&L are carried
from the Gr8Trade export in the source deck and were not independently re-derived.
Engine tallies (+0.48R / 168) are the indicator’s own in-sample counts on one chart — a
hypothesis, not a backtest.