SMCI 2026-06-11 — VCET Visual Replay (post-trade execution & management overlay)

Analyst replay 2026-06-11 · Symbol SMCI · Account d1c277 · Upgrades the text post-mortem into a chart-based replay · Framework: VCET v1.1 Chart Overlay
Corrected 2026-06-12 against the FULL Gr8Trade order log (SMCI_20260611_orderlog_full.json): the 12:24 event entry was 100 sh wavg 29.570 (20+20 @ 29.585 + 60 @ 29.56), not the 200 sh @ 29.5725 the summarized log implied; the exact 12:24→12:36 campaign P&L is −$16.79 (replaces the ≈+$7.20 residual-flatten estimate); a previously unseen 12:16 starter (B 100 @ 29.425 → S 12:21 @ 29.45, +$2.50) was added. Campaign peak position: 120 sh.
Data provenance — what is real vs inferred
REAL · COMPUTED Candles, VWAP, ±0.5/1/2/4 SD bands, 2-day VWAP, CBO 20-bar channel, volume. Built from actual SMCI 1-minute bars pulled from the karb price server (the firm intraday feed — separate from the blocked Liberator/CloudQuant entitlement), downsampled to 2-minute. VWAP & bands use the OHLC/4 volume-weighting convention. These are math on real bars, not estimates.
REAL · ORDER LOG Trade markers (every buy/sell/short) are your actual Gr8Trade fills from the FULL order log (SMCI_20260611_orderlog_full.json; TransactTimeUTC−4h = ET; partial fills included — never filtered to exec-type "fill"). Key fills carry exact time/price/size; two morning chop clusters (10:33, 11:36) are plotted as single representative markers. The 12:24-campaign P&L is computed exactly from this log — no flatten assumptions (2026-06-12 correction).
INFERRED VCET regime states, zone labels, CBO break direction, and the +3/+10/+15/+25 checkpoint calls. The VCET v1.1 engine was NOT run against these bars (that is Stage B, currently out of scope). State labels are an analyst read of the real price action against the v1.1 rule definitions — they are what the overlay would have said, not computed engine output.
MODEL · WHAT-IF The automation simulation (section 2) is hypothetical hotkey output — NOT actual fills. Its arithmetic is computed from the real bars and real band levels, but every marker on the simulation chart is a model execution. The actual and simulated views are kept on separate charts so real fills are never confused with the model.
No DB writes · no classifier run · no taxonomy change · no scorecard refresh. The trade plan levels (stop/T1/T2/runner) are structural choices anchored to the real bands, labeled as such.
Parameters — exactly what every computed series on this page uses
ParameterValue on this pageNotes
VWAP anchor09:30 ET session open, RTH onlypremarket / after-hours bars are EXCLUDED from VWAP, bands and the chart
VWAP price inputOHLC/4 per bar, volume-weightedmatches the TraderDale_VWAP.pine convention
SD methodvolume-weighted variance of OHLC/4 about the running VWAPσ = √(Σp²v/Σv − VWAP²), cumulative from the anchor
Band levels±0.5σ · ±1σ · ±2σ · ±4σT1/T2 use the +2σ/+4σ values PROJECTED at the entry bar, not the live (expanding) band
2-day VWAPseparate accumulator: cumulative OHLC/4·vol over 6/10 RTH then 6/11 RTHcomputed independently of the session VWAP — not an average of the two session VWAPs
CBO lookback20 barshighest high / lowest low of the PRIOR 20 bars
CBO current-bar exclusionYESwindow = bars[i−20 … i−1]; matches the Pine overlay spec (per Pardo)
CBO timeframe2-minute bars (this chart's timeframe)parameter-dependent: the live Pine overlay computes on whatever chart TF is loaded (often 1-min) — break times and levels can legitimately differ between timeframes
CBO warm-upfirst 20 bars have no channelon a 2-min chart the channel only exists from 10:10 ET — no CBO is drawable 09:30–10:10
CBO break strengthnot shown on this pagethe Pine overlay normalizes strength by ATR or channel width; the replay generator uses (close−channel)/σ — same name, different denominators; do not compare across tools
Gate model (RED / pass)NOT computed — inferred analyst readno gate engine ran; zone colors are narrative judgments
RVOLNOT computed on this page"peak RVOL" in the morning narrative is a qualitative read, not a calculated series
Audit note (2026-06-11): on these computed series the 12:24 bar closed 29.684 vs a prior-20-bar channel high of 29.52 — the CBO LONG break at 12:24 is verified on computed values (the break itself; the surrounding regime/state labels remain inferred). The model entry 29.57 is the trader's actual 12:24 fill average (100 sh, FULL order log) adopted by the simulation — no automated parameter generated it. A close-confirmed CBO rule would have signaled at the 12:24 bar close (29.684), with a realistic automated first fill ≈ the next bar open (29.69) — about 12¢ worse than the discretionary fill the sim inherits.

Headline — you found the trade, then scalped it. The overlay would have held it.

SMCI gapped, ran to 30.14, then washed out to the 28.61 day-low at 10:16 (you stopped at 29.10 — into the hole). It put in a higher low (28.77) at 11:01 and based for ~80 minutes. You probed it with a 12:16 starter (100 @ 29.425 — scalped 12:21 @ 29.45 for +$2.50). At 12:24 it broke out above the compression shelf and reclaimed/expanded through +1 SD — your best entry, 100 sh wavg 29.570 (20+20 @ 29.585 + 60 @ 29.56, full order log). You sold 4–5 minutes later for +2.5¢/share avg, then churned the re-add / fat-finger / chase sequence to an exact campaign P&L of −$16.79. It then ran to 32.05 (close 31.96), tagging the +2 SD (29.84), the +4 SD projected target (30.33), and the 2-day VWAP (≈31.2). Managed to plan, that 100-share entry was worth ~1R at +2 SD, ~2.8R at +4 SD, ~6–9R on the runner. Morning chop = RED gate (pass). 12:24 = the conversion (hold to the checkpoints). One reframe from the trader: the 12:32 short was a fat-finger — an execution error, not a strategy decision — and section 4 turns it into a set of execution-protection rules the hotkey should enforce.

1. Actual replay — what you actually did (real bars · real fills)

ACTUAL · REAL FILLS
SMCI SMCI · 2-min · 2026-06-11 · 09:30–16:00 ET
Black hollow candles · real karb bars · VWAP & σ-bands by OHLC/4 · zones & checkpoints are inferred VCET reads · markers = real Gr8Trade fills · PLAN price-lines & shaded boxes = the never-placed plan, not working orders
VWAP +2 SD (T1) +4 SD (T2) ±1 SD ±0.5 SD 2-day VWAP CBO 20-bar channel buy sell short
Drag to pan · scroll to zoom · hover for OHLC. Zone bands, checkpoint lines, callouts and the risk/reward boxes are HTML overlays positioned on the real time/price axes.

Morning (09:30–12:20) — RED gate

Opening spike → washout to 28.61 → higher-low base. ~10 entries, all RED-tape: extended gap, peak RVOL, bad location, no compression-breakout signature. The overlay says pass or tiny-starter only. Your 09:53–11:42 fills netted −$96.25 in SMCI alone (exact, full order log).

12:24 (the conversion) — HOLD candidate

Compression → reclaim → expansion, CBO breaks long over the 20-bar channel. The +3 min checkpoint passes ("no early failure"); +10/+15 say "trend conversion candidate / holds." This is the trade to hold to the checkpoints, scale at the bands, and trail.

2. Automation simulation — what if the hotkey managed it?

Same real SMCI bars, same real VWAP/σ levels — but every marker below is model output: the planned VCET hotkey executing the 12:24 conversion (entry → stop → T1 → T2 → runner) and blocking the trades the gate forbids. Not an actual trade — automation what-if.

SIMULATION · MODEL OUTPUT
MODEL HOTKEY PLAN — not actual fills
SMCI SMCI · 2-min · 2026-06-11 · hypothetical hotkey execution
Real bars & bands · teal = model entry · purple = model T1/T2 exits · gold = model runner exit · red dashed = model stop · red circles = orders the guards BLOCK · CBO 20-bar channel NOT drawn on this chart — prototype display omission, not a different computation (identical computed channel as section 1; see the Parameters box)
model entry model T1/T2 exit model runner exit model stop 29.30 plan path blocked order
Every event on this chart is hypothetical/model output. The actual fills live on the chart in section 1 only.

2.1 Simulation assumptions

2.2 What-if table — every scenario priced from the real bars

ScenarioEntryStopRisk/shSh T1 pxT1 shT1 P&L T2 pxT2 shT2 P&L Runner exitRun shRun P&L TotalRvs actual

Scenario A is computed exactly from the FULL Gr8Trade order log (12:24:52→12:36:22 flat-to-flat campaign): +$2.50 (entry → 12:29 sells) − $14.20 (12:30 re-add 120 sh → 12:32 flatten) − $0.59 (12:32 fat-finger round-trip) − $4.50 (12:36 chase → stop-out) = −$16.79, no assumptions. All other rows are model arithmetic on real band levels. The C2 trail result is an honest negative: on this day a tight +2SD-close trail cut the runner at 13:12 for less than holding to the 2-day VWAP — trail rules trade some upside for protection. Per-leg P&L is rounded to the cent for display; totals are computed before rounding, so legs may differ from a total by $0.01.

2.3 Actual vs automated

ACTUAL (what happened)

  • Took a 12:16 starter (100 @ 29.425) — scalped it 12:21 for +$2.50.
  • Bought a good setup (12:24, 100 sh wavg 29.570).
  • Sold before the +10 confirmation (12:29, +$0.02–0.03/sh).
  • Re-entered higher and bigger (12:30, 120 sh @ 29.705/29.725).
  • Fat-fingered a 120-sh short into the conversion (12:32).
  • Chased again 12:36 @ 29.86, stopped out 29.815 (−$4.50).
  • Missed the planned +2 SD / +4 SD / 2-day VWAP move.
  • Net on the campaign: −$16.79 (exact, full order log).

AUTOMATED (model)

  • Entry accepted only after a valid gate / expansion context.
  • Stop placed immediately at 29.30 (max loss defined: −$27.00 on the actual 100 sh; −$54.00 on the 200-sh what-if).
  • T1/T2 orders placed automatically (29.84 / 30.33) — both filled.
  • Runner left open, trailed toward the 2-day VWAP (31.16).
  • Opposite-side short blocked by the direction lock.
  • Morning RED-gate full-size attempts blocked or reduced.
  • Net at actual size (100 sh): +$68.10 (2.52R) — ≈ $85 missed vs the −$16.79 booked. (200-sh what-if: +$136.20.)
ComparisonActualAutomated (100 sh model — actual event size)
P&L on the 12:24 campaign−$16.79*+$68.10
Planned riskundefined (no stop placed)$27.00 (100 × $0.27)
Max loss if stoppedunknown−$27.00 (−1.0R, bounded)
R multiple−0.62R (imputed vs the model's $27.00 risk)2.52R
Missed opportunity≈ $84.89—
Rule violations prevented4 committed (early exit · chase re-add · fat-finger short · 12:36 chase)all blocked/warned + ~10 morning RED-gate full-size blocks

*exact, computed from the FULL Gr8Trade order log — no flatten assumptions (see 2.2 note). The 200-sh what-if model makes +$136.20.

3. VCET decision table — what you did vs what the overlay would advise

TimeChart condition (real)VCET state (inferred)What you didWhat VCET would adviseCorrect action
09:53Opening drive faded; price chopping under the 9:31 spike, gap still extended, peak RVOLRED GATE ROTATION / volatileB 100 @ ~29.55, then S 100 @ 29.35 (−$21)Pre-trade gate RED → no plain Pull-In. +3 min fade fires by 09:57.Pass, or 1 tiny starter, hard stop, no re-chase.
10:46Chasing the prior 10:30 high after price had already washed to 28.61 & bouncedCHASE / EXTENDED UNKNOWN setupB 300 @ 29.41–29.50, S 300 @ 29.34 (−$37.50, biggest morning loss)Buying a prior high into a RED tape = the textbook chase the dashboard screams at.No setup. Stand down. Not a Pull-In.
11:36Base building — higher low 28.77 (11:01), BB tightening, grinding upCOMPRESSION formingScalp churn ~29.33–29.43, in and out fast (−noise)Compression = wait. Don't scalp the coil; mark the shelf and the trigger.Wait. Set alert at shelf high (~29.5).
12:16Reclaim out of the base through VWAP/+0.5 SD — newly visible in the full order logRECLAIM STARTER base intactB 100 @ 29.425 — scalped 12:21 @ 29.45 (+$2.50)Valid starter location ahead of the shelf break. The starter's job is to be held into the confirmation, not scalped.Hold the starter into the 12:24 trigger.
12:24Breaks the compression shelf; the 2-min bar closes 29.68, reclaiming/expanding above +1 SD (29.59); CBO breaks long over the 20-bar channelEARLY EXPANSION CBO LONG · Pull-In→TrendB 100 wavg 29.570 (20+20 @ 29.585 + 60 @ 29.56 — your best entry of the day)Real conversion. Manage as HOLD CANDIDATE: stop below shelf/VWAP, scale at +2/+4 SD, trail.Enter, set bracket, hold to +10 min.
12:29Normal post-breakout drift; price 29.59–29.60, still above entry & VWAP rising+3 min: NO EARLY FAILURES 50 @ 29.59 + 50 @ 29.60 (+2.5¢/sh avg) — scalped outBetween the +5 and +10 checkpoints. The +3 check passed. This is not a fade signal.Do NOT scratch. Hold; first scale is +2 SD (29.84), not here.
12:32Shallow breakout-retest holding ~29.59; trend intactHOLD / trend up (active context = LONG)Re-add 120 sh @ 29.705/29.725 (chase), dump @ 29.59, then FAT-FINGER short 120 @ 29.6101 (execution error, not a strategy decision), cover 29.615This was never a valid VCET short — it fired while the active trade context was LONG / CBO LONG / expansion candidate. Classify as an execution-protection failure, not a strategy idea. A direction lock blocks or warns on the opposite-side order (section 4).System blocks the short. Hold the long / trail; a retest is not a reversal.
13:43Established uptrend; price 30.37 above all bands, trending to 4 SD areaTREND UP continuationB 20 @ 30.37, S 20 @ 30.37 (3 min later)This is trend-continuation (not a fresh Pull-In): smaller R, tighter stop, the runner is the thesis.Hold the runner, trail under higher lows / 2-day VWAP.

Chart conditions are read from the real bars; VCET states are inferred (engine not run). The +2 SD / +4 SD / 2-day VWAP prices are computed from the real bands.

4. Fat-finger / execution protection

The 12:32 short was an execution error — and the system should have caught it

Proposed protection rules

RuleNameTrigger → behaviorApplied to 6/11
AActive Direction LockIf trade direction = LONG and VCET state ∈ {expansion, CBO LONG, hold candidate} → block the short hotkey unless the user confirms an explicit override.Blocks the 12:32 fat-finger short outright.
BCooldown After ExitIf the user exits a long during a valid conversion candidate → block an immediate opposite-side short for 30–60 s unless explicit override.12:29 exit → 12:32 short is inside the cooldown window; second line of defense if A is overridden.
CWrong-Side CBO GuardCBO direction LONG → block/warn on short entry. CBO direction SHORT → block/warn on long entry.CBO broke LONG at 12:24; any short before a fresh SHORT break is wrong-side.
DRed-Gate Size GuardGate = RED (morning chop) → hotkey forces tiny starter size or blocks full-size entry.Would have cut all ~10 morning entries to starters — including the 300-sh 10:46 chase (−$45).
EPlanned Trade ModeOnce the hotkey creates a structured trade (stop/T1/T2/runner), the system knows the intended direction and rejects accidental opposite-side orders while the bracket is active.The 12:24 bracket carries direction=LONG until stop/targets resolve — the 12:32 sell-side order over position size is auto-rejected.

Rules A–C and E are state checks the existing hotkey engine (Z:\trading\dev\hotkey-engine\) can host in its ConfirmationGate; D is already spec'd as the grade/gate sizing path. Design notes only — no engine edits made.

5. The 12:24 entry — risk, targets, and what management would have captured

EntryStopRisk/shTarget 1 (+2 SD)T1 RTarget 2 (+4 SD)T2 RRunner trailCorrect action
29.57
zone 29.56–29.585
29.30
est. structural*
$0.27 29.84
tagged 12:36
~1.0R 30.33
tagged 13:04
~2.8R +2 SD → 2-day VWAP
≈31.2 (tag 15:30) → 31.96 close
~6–9R

*Estimated structural stop = just under the 12:00–12:23 breakout shelf and back below session VWAP (29.34) = "reclaim failure." Computed from real bars, but it is a structural/discretionary level — a tighter 2-min momentum stop (entry-bar low 29.44, ≈$0.13 risk) is the aggressive alternative. T1/T2 are the +2 SD / +4 SD band values projected at the 12:24 entry (29.84 / 30.33); both filled as price trended. The live (expanding) +4 SD band was never tagged — price rode under it the whole trend, which is exactly why +4 SD is a scale-out, not an exit.

Scalp vs managed — same 100-share entry

What you booked: −$16.79 exact on the whole 12:24–12:36 campaign (full order log). What the plan captured (50/30/20 model, section 2): +$68.10 — 50 sh at +2 SD (+$0.27/sh), 30 sh at +4 SD (+$0.76/sh), 20 sh runner to the 2-day VWAP (+$1.59/sh); +$84.10 if the runner rides to the close. The entry was never the problem. The hold was.

6. Before / after behavior

Old behavior (what happened)

  • Over-traded the morning RED-gate chop (~10 entries, into the wash-out)
  • Chased the prior high at 10:46 with 300 sh → biggest morning loss
  • Found the 12:24 best entry (100 sh) — then scalped it in 5 min for +2.5¢/sh
  • Re-entered 13–15¢ higher with MORE size (120 sh — chasing what you sold)
  • Fat-fingered a short into the conversion at 12:32 (execution error — section 4)
  • Missed the +2 SD, +4 SD and 2-day VWAP targets the trade ran to

VCET behavior (what the overlay drives)

  • RED gate → pass / tiny-starter only; no repeated Pull-In attempts
  • Compression → wait for the shelf break
  • Enter the 12:24 conversion (CBO long + reclaim/expansion)
  • Hold through +3 / +10 / +15 confirmation (don't scratch a passing trade)
  • Scale 50% at +2 SD, 30% at +4 SD
  • Leave a 20% runner, trail +2 SD / 2-day VWAP / higher lows

7. Where the hotkey fits

Your note: "the black square is where I want the hotkey — get my fills off VWAP/−0.5 SD, risk whatever, sell % at 2 & 4 SD, and leave runners at 2 SD / 2-day VWAP." That is the canonical Pull-In execution, and the chart shows exactly where it does and does not belong.

Hotkey Plan: what should have happened (12:24)

  1. Enter long in the VWAP / +0.5 SD / +1 SD reclaim zone (29.47–29.59).
  2. Auto-place stop below structural invalidation (≈29.30, shelf/VWAP-reclaim).
  3. Auto-place Target 1 at +2 SD (29.84).
  4. Auto-place Target 2 at +4 SD (30.33).
  5. Leave a 10–20% runner.
  6. Trail the runner with +2 SD / 2-day VWAP / higher-low structure.

Scale model: sell 40–50% at T1 · 30–40% at T2 · leave 10–20% runner. Modeled in section 2 as 50/30/20 → on the actual 100-sh event size = 50 / 30 / 20 (200-sh what-if = 100 / 60 / 40).

Morning chop — hotkey should be blocked

  • Gate = RED → hotkey disabled, or force a tiny starter only.
  • No full-size Pull-In allowed.
  • No 2 SD / 4 SD target plan unless conversion confirms.
  • 3-min fade = reduce / exit.
  • Chase into a prior high = invalid setup (10:46).

The hotkey is an execution tool, not a permission tool. It should only arm when the gate is not RED.

Hotkey rule, stated three ways

8. Rule cards

SMCI 6/11 — trader rule card

SMCI 12:24 Rule

If I enter compression → CBO LONG → VWAP-band expansion, I cannot scalp out before the +10 minute checkpoint unless the stop is hit.

Morning Rule

If the gate is RED, the hotkey blocks full size or forces a tiny starter only. No 2 SD / 4 SD plan until a conversion confirms.

VCET identifies when a trade deserves automation.
The hotkey enforces the plan: entry · stop · T1 · T2 · runner · and wrong-side protection.

9. Limitations & how to make this exact

Exact next step — to replace every inferred label with computed engine output: run VCET v1.1 against these same SMCI 6/11 bars (already staged at Z:\trading\dev\smci-vcet-replay\bars_2day.json). That yields real state per bar, the real CBO break flag/strength, and the real +3/+10/+15/+25 checkpoint evaluations — at which point this page's yellow "inferred" tags flip to green "computed." That run is the only blocked piece, and it needs no new data.

Related: SMCI text post-mortem · VCET v1.1 Chart Overlay Usage Guide · Mechanism Validation Plan · VWAP Pull-In Quality Gate