Playbooks

Reference playbook documents for high-conviction setups. Generated from gen_playbook.py + per-trade JSON. Best-trade selector picks weekly / monthly winners (pipeline staged).

● Live — Momentum State Engine (Alpha Trend Ride)

ATR Scanner · 15-ticker universe · 2026-05-15
5-state engine · sequential momentum machine · not OB/OS
3 OFFENSIVE — M5 Re-Expansion firing: XLE. M1 Expansion (hold/press): AAPL, QQQ. 4 DEFENSIVE — M2 Exhaustion (cooling but still extended >4.5% above 20 DMA; trim/tighten, no new longs): AMD, GOOGL, NVDA, TSLA. 3 WATCHFUL — M4 Stabilization at the rising 20 DMA (load-bearing state, prepare for M5): AMZN, IWM. M3 Reversion (price actively reverting to 20 DMA): SPY.

The framework is not RSI overbought/oversold. It models compression and expansion inside trend structure — the 20 DMA is Structural Equilibrium, not "support". The differentiator between M2/M3/M4 is price distance from the 20 DMA; state transitions are the trade triggers.
Open the scanner →

★ Today's Featured Setup — Backside Long (Mean-Reversion / Rubber Band)

MSFT + AMZN · 2026-04-30
Backside Long · +$127.51 combined
Both stocks extended below VWAP ≤ −2 SD, printed double-bottom failures, reclaimed −1 SD on volume, ran back to VWAP. AMZN held to target = +0.62R. MSFT cut on 4-EMA = +0.12R = 5× less reward on the same setup. The canonical management rule: exit ENTIRE position at VWAP, no partials, no trailing.
Open today's scorecard →

Backside Long — canonical rules (locked 2026-04-30)

The Rubber Band Effect. When price extends too far from VWAP, it snaps back toward the mean. Standard deviation bands (built around VWAP) formalize the entry, stop, and target.

Setup: Price extended to −2 SD or beyond below VWAP. Double-bottom failure pattern below VWAP. RVOL > 3 strongly preferred.
Entry: Reclaim of the −1 SD band on a green bar with volume. The reclaim IS the trigger.
Stop: Tightly below the −2 SD band. Band-based, NOT structural LL. A break of −2 SD invalidates the rubber-band thesis.
Target & management: Exit ENTIRE position at VWAP (the mean). No partials. No trailing stops.

Aliases (all resolve to backside-long): Bella Long mirror, Mean-Reversion Reclaim, −1 SD Reclaim, Rubber Band, Double-Bottom Reclaim.

Day-2 Continuation — canonical rules (drafted 2026-05-07 · d2c-001 · pending live-trade approval)

Meta-qualification, not a new trigger. Day-2 Continuation is the 5-filter qualification that sits on top of an executing branch (vwap-pull-in or fl2-fash-late). Tagging both slugs — meta + branch — lets us measure whether the qualification adds expectancy over the trigger alone. Filter lives in the shannon_gates table; entry trigger is inherited from the branch.

5-filter qualification (must all pass): (1) prior-day trend day, (2) RVOL ≥ 1.2×, (3) no overhang, (4) clean structure, (5) liquidity.
Entry: Inherited from the executing branch — vwap-pull-in or fl2-fash-late. Day-2 row does NOT add its own trigger.
Sizing: From the branch. Day-2 row does NOT add a size multiplier.
Hard invalidation: 5m close below prior-day low (or equivalent structural reference) ends the Day-2 thesis for the session. No re-entry.
Documented failure mode: averaging into the base. 5/4 NVDA — 12 long round-trips, all losses, −$59.98. The rule: once the thesis is invalidated, stop; do not redeem the base.

Canonical row: slug=day-2-continuation, id=d2c-001, category continuation, correlation_family vwap_pullback_continuation, direction long, is_approved=false until first live attempt is scored.
Anti-confusion (see today's NVDA scorecard): 09:45 alpha blocks and 09:46 clusters that follow a prior-day trend day are Day-2 Continuation + executing branch, NOT Backside Long. Backside Long requires −2 SD extension and a double-bottom reclaim — if the entry is on the FIRST pullback inside a trend day's continuation structure, it is Day-2, not Backside.

Reference playbooks

ONE viewer. Every session opens in the MTF Strategy Lens (house chart theme: white, hollow black candles; σ + AM ATR renderers; 1m → Daily; executions; inspector). Narrative pages are the write-up around that same viewer — they embed it, they do not carry their own chart.

MARA GAP-AND-GO · TYPE A
2026-09-18 · MIX · 87 broker fills · 200 sh max · gross +$72.03 · mara-2026-09-18.html
+4.1% gap above the PDH out of the tightest daily coil in 20 sessions (BW %ile 0) with the 5/10/20/50/200-DMA stacked, on a sector catalyst (SEC / tokenized stocks; IBIT +6.3%). 50 sh premarket at 12.03 (anticipatory), then the adds with reasons: 12.22 on the AM ATR break-confirm bar (09:32), 12.21 on the Back-Through-Open reclaim after the 09:36 dip to 12.05 held the PM low, 12.30 into the coil at the mean with σ CONT LONG printed — 200 sh avg 12.19 by 09:45; 75 sold at +2.4σ (12.38), 25 at +1.6σ (12.58). Morning +$73.50 in two sequences; then 15 afternoon re-entries (1,300 sh, one accidental 20-sh short flip) gave back $9 while MARA added 5% to close 13.23. Realized +$72.05, full-day fills reconstructed from the DB (v3 snapshot ended with 50 sh open). Family: compression → expansion / gap continuation, Back Through Open as the trigger, TYPE A confluence. LAB.
Open in Lens →Narrative →
TJX COMPRESSION → EXPANSION · TYPE A
2026-09-18 · LONG · 62 broker fills · 100 sh max · gross +$55.39 · tjx-2026-09-18.html
Flat open on a name 17% down in a month (0.9σ under the 20-DMA, 5-DMA just reclaimed), then a 2-minute coil at the 0.1st bandwidth percentile on the session VWAP with the 30m at the 7th and the daily at the 15th. σ engine LONG ARMED 09:34 → coil 09:40 → CONT LONG on the 09:46 bar; first fill 09:47:37 (100 sh at 126.72, 19¢ over VWAP); AM ATR break-confirm LONG 09:58, in play all day; trend day to 129.47 at 14:43. First hour +$46 on 21 fills (8 adds on price improvement, 11 partials up to 128.22); afternoon +$9 on 41 fills in 8 re-entries. Realized +$55.42, full-day fills reconstructed from the DB (v3 snapshot missed the afternoon). Family: compression → expansion / trend continuation with TYPE A engine confluence; adds classified on-book ORB continuation. LAB.
Open in Lens →Narrative →
INTU MEAN REVERSION · ENTRY REVIEW
2026-08-26 · LONG · 12 broker fills · 20 sh max · gross +$268.38 · intu-2026-08-26.html
Earnings gap-down of -9.5% (close 357.46 → open 323.49) whose after-hours low 301.50 tagged the daily −2σ band and the rising 50-DMA at once; premarket built a higher low (312.00) on 33.1× volume and opened at the daily −1σ. Bought the premarket last 15 s before the bell, doubled on the PM-high break, sold 15 of 20 within 99 s at the 20-DMA mean (339.80); the open bar ranged 322 → 347 on 11× volume. Realized +$268.38 = +0.90R on a MODEL stop (no stop was declared); hold MFE +1.63R at 351.39. Family: counter-trend mean reversion — the Backside Long / Rubber Band shape on the DAILY σ frame (LAB extension, needs review), Earnings Day 1 as context only; no engine had the long (σ short-armed, AM ATR SHORT at 345.71). The 11:13 one-share re-entry is flagged OFF-BOOK against the open ATR short. First page built on the new ENTRY REVIEW capability: per-fill state at the last completed bar, session levels, volume pace, rule-evidence classification over the strategy library, setup-vs-execution grading, and ✓ ~ ✕ review on every entry. LAB / N=1 — not a new strategy.
Open →
RESEARCH TOP TREND TRADES
2026-08-30 · 9 tickers · 20 trend windows · every number engine-computed · research/
The Google-Doc seed list (ALOY · BTDR · BBIO · IBIT · JNJ · EBAY · CRWD) plus the discovered NVDA 20-DMA pullback and the most recent AMZN session, run through the parity-verified engine framework. Headline: available R of 23.3 (IBIT 4/21–29), 14.5 (BTDR), 12.3 (ALOY), 11.2 (CRWD) against sub-10%% capture everywhere — the hold problem, quantified across the whole list. Exit-rule race per window (5/10/20/50-DMA vs AM ATR), stage snapshots (pre-setup → compression → trigger → expansion → pullback → failure), indicator verdicts (earliest signal / best hold / cleanest exit / noisiest), actual-vs-theoretical two-track for every traded case, and a side-by-side comparison matrix. Every ticker opens in the interactive multi-timeframe Strategy Lens with real executions. LAB.
Open →
LENS MTF · CANONICAL
2026-08-30 · Multi-Timeframe Strategy Lens — any ticker, any session, 1m → Daily · lens.html?ticker=&date=
One trade object per session: canonical minute bars → TF aggregation (1m/2m/3m/5m/10m/15m/30m/60m/2H/4H) + canonical Daily → per-frame σ + AM ATR engines → events tagged source_timeframe → reconciled fills at absolute timestamps. Chart TF and Strategy TF are separate controls (“what did the 30m engine see while I executed on the 1m?”), executions never resample (aggregated markers expand on click), and BW / BW %ile is a stored first-class feature (lifecycle stamps, min lead-in, comp→exp delta, full MTF context table). Auto-discovers every session you traded per ticker (MY TRADING DAYS chips, collapsed for heavy names) with a one-line SETUP OCCURRENCES row and a per-case CASE CHRONOLOGY (the selected setup’s development only). Candle/Event Inspector: hover any candle to scrub, click to pin — TIME/OHLCV, strategy events with the actual rule WHY, lifecycle state, lens values at that bar, market structure, and your executions. Lifecycle cards, chronology nodes and execution rows all jump the chart. Live: AMZN 8/28 · NVDA 6/1 (20-DMA pullback → 6/2 day-2 confirm) · VZ (13 sessions) · CRWD (6) · AAPL (2) + ALOY/BTDR/BBIO/IBIT/JNJ/EBAY (11 tickers). CANONICAL — every playbook inherits this viewer.
Open →
VZ LENS · TYPE B
2026-08-13 · LONG · 10 rt · 600 sh · gross +$221.35 · lens.html?ticker=VZ&date=2026-08-13
The Type B reference session in the canonical Strategy Lens: dual-engine (σ + AM ATR) view of the 8/13 long with full CASE CHRONOLOGY, Candle/Event Inspector and reconciled fills. 13 sessions traded (2026-08-07 → 2026-08-26, ticker gross +$233.77) — every one openable from the MY TRADING DAYS chips. Narrative playbook: vz-2026-08-13.html.
Open in Lens →
AAPL LENS · BW 0.0 %ILE
2026-08-28 · LONG · 50 rt · 505 sh · gross +$162.63 · lens.html?ticker=AAPL&date=2026-08-28
Compression-into-expansion flagship: daily BW hit 4.86 at the 0.0 %ile on 2026-08-27 — the tightest daily bands in the 20-day rank window — with 4 lead-in bars below the 20th %ile before the trade-day range expansion. 9 sessions traded (ticker gross +$230.65). Full 30m-engine chronology + inspector.
Open in Lens →
CRWD LENS · DUAL-ENGINE
2026-08-03 (most recent of 6) · LONG · 39 rt · 155 sh · gross −$11.09 · lens.html?ticker=CRWD&date=2026-08-03
The ticker behind the dual-engine study (σ bands vs AM ATR, bar-by-bar parity-audited), now viewable per-session in the canonical Lens. 6 sessions traded (2026-07-20 → 2026-08-03, ticker gross −$224.25). Note the 4:1 split (ex 2026-07-02) — raw feeds are unadjusted; lens data is split-normalized across the boundary.
Open in Lens →
VZ TYPE B · NEW
2026-08-13 → 08-17 · Adaptive Structure Without an ATR Entry · coil_continuation (long) · written if-then, followed
The deliberate counter-case to CRWD: leg up 8/6–8/11, then a COIL at the mean (8/10, BW %ile 15.8, 0.64σ from a rising 5-DMA basis) — and on 8/13 the day’s first PDH break (47.08) + 1σ expansion + a higher-low structure break (47.45). Plan written the night before (overnight scan: leg-up + consolidation); 600 sh avg 47.135 premarket/open, plan stop 46.81 (VRVP shelf, never traded, MAE −0.38R), scaled 47.25/47.48/47.80/48.37/48.17 = +$221.35 = +1.14R against an in-hold MFE of +4.53R — the hold, again. The lesson is the silence: AM ATR produced no entry, and the engine port names the exact gates — two valid reclaims (8/4 bowl 0.74 ATR, 8/7 bowl 1.30 ATR) failed their open-bar confirms, the break-confirm at 47.48 expired 8/12, one session before the breakout, and re-arming needs 2+ daily closes below the 5-DMA, which never came. ATR_NO_SIGNAL_REASON = BRK_EXPIRED → NO_RECLAIM_AVAILABLE. Ships the Strategy Lens component (same candles, σ / ATR / confluence views), a computed per-engine provenance table, the TYPE A–D signal-family taxonomy (no AND-gate — requiring both engines would have filtered this trade out), and research row playbook_data/vz_2026-08-13.json. LAB / unvalidated.
Open →
CRWD LAB · NEW
2026-07-31 → 08-04 · Day 1 Pullback to the 50 DMA · vwap_pullback_continuation (long)
−20% in nine sessions into the 50 DMA (7/29 closed 179.38 vs a 179.80 50-day), six sessions basing, BW %ile 13.5 on the 30-min, then the day’s first PDH break woke the VWAP play — pull-in held the line, box break filled 191.21, stop the hold low 187.98, target 197.68 (2R). Monday gapped it; 8/4 ran to 212.64 = +6.63R while 2R was banked. Full deck conversion: SPY/QQQ backdrop, daily BB+SMA+RSI, the 30-min trigger frame with 5-day basis + σ bands, 2-min annotated replay of 7/31 + 8/3, reading the tape, management + forward plan, NVDA→CRWD lineage. Reconciled to the cent (GATE PASS): 60 matched round trips — 7/31 day −$6.16 (21 RT, 23.8% win) · 8/3 −$11.09 (39 RT, 43.6%) = −$17.25 day-trading while the hold made the money. Two data traps caught: the daily feed is split-unadjusted (CRWD 4:1 on 7/2 → a fake 465 “50 DMA”), and an order-log FIFO mis-pairs the overnight carry (fake +$6,208) — P&L is sourced from the matched-trades witness instead. Also corrects the deck’s “26% win” bucket-edge stat (it is the target-hit rate, not the win rate). Upd 8/30: new Strategy Lenses slide — the same 30-min candles through σ / AM ATR / confluence views with a computed provenance table; the AM ATR side is now mechanically told (filtered reclaim → brk-confirm 185.46 → long 187.83, stop 174.12 → +2.07R runner — matching the TV labels), making CRWD the canonical TYPE A — CONFLUENCE case beside VZ’s Type B. LAB / unvalidated.
Open →
TOOL v5.2 → v6.0-SWING
2026-08-04 · Adaptive Mean + σ Bands — the technology behind CRWD · tool_playbook
The honest manual for the indicator that called the CRWD coil on 7/28 and the trigger on 7/31. Corrects the deck’s bucket-edge stat: “+.48 ExpR over 168 signals, 26% win” — the 26% is the target-hit rate, not the win rate (v5.2 scored a W only at the target and an L only at the stop; every trailed/covered/timed-out exit counted in N and in neither column). Proof from the two numbers alone: if the other 74% were −1R losses the average would be −0.22R, not +0.48R. Also flags that the 168 are the entry mechanic, not this setup — no coil, PDH or alignment filter. NVDA → CRWD lineage (pullback to the big MA → 5-DMA reclaim → ride; same unfixed habit — the hold). CRWD forward plan: projected move 209.15 tagged 8/4 — the reason-to-hold is spent; daily +2σ 211.87 arms the no-adds flag. Ships v6.0-swing: fades/CONT/execution-frame cut, RECLAIM entry added (the NVDA trade had no code), projected-move target, trail-after-paid, and alert() JSON capture. Two v5.2 bugs found: 26-bar auto-flat kills swings; the Win% column is what put the bad number on the slide. Build not yet backtested — verify against CRWD/NVDA first.
Open →
NVDA WATCH · TMT #2
2026-07-06 · Bottom Bounce at the 200-DMA · bottom_bounce_reversal · pre-registered, no position
Down from 236.54 (5/14) into the rising 200-DMA (191.29); pivot low 189.80 (6/29) held the level, close 195.55 coiling under every faster MA. Two mechanical triggers pre-registered: engulf through 200.06 (max of last-3 highs, close top-30%, stop = half the candle) or $200 reclaim + hold on RVOL ≥ 1.2 (≥~198.8M sh — gate is mandatory). Hard no-long below 195; lose 189.80 = different family, wait for the next reversal candle. Anticipation math: reclaim_threshold 196.70, ma_rolloff_date 7/16 — the 10-DMA headwind expires on a printed schedule (208.65 leaves the window 7/7). Fires → TMT shadow specimen #2. All values computed + independently re-derived. NOT A+ — counter-trend grades as its own family. LAB / watch only.
Open →
EBAY LAB · NEW
2026-07-01 → 07-02 · Level Breakout — Long Momo Burst · band_break archetype
One of the best trades: broke the 111.84 multi-day shelf on the open, held every pullback above a rising VWAP (MAs stacked 5>20>50), and rode the trend two sessions to 115.98. 2-day annotated replay (VWAP + bands + MA stack + 2-day AVWAP + reconciled fills). Realized +$152.53 on the 95 shares closed (reconciles to witness +$152.50; 5 still open — a live swing, not a flat book). Honest note: quiet tape (RVOL <1) so the volume-gated band_break didn’t fire; the engine’s validated long was the 09:48 vwap_pull_in (+1.21R). The A+ was alignment + the hold discipline. Upd 7/6: full library spine added — Daily BB/SMA/RSI + 30-min panels with the Alpha Trend env cloud (Daily GREEN ✓ / 30-min UNKNOWN at entry — honest data-window limit). Upd 7/6 PM: new Management Tracking section — Actual fills vs J3P 60-20-20 σ-track vs TMT (Tiered MA-Trail, shadow specimen #1: 33/34/33 on the 5/10/20 DMA, all tranches open, +0.56R). LAB / unvalidated.
Open →
MU LAB · NEW
2026-07-02 · VWAP Reject — Short Continuation · A+ “Broken Slot Machine”
Short-side mirror of the VWAP pull-in: broke below session VWAP, retraced UP into the 2-day anchored VWAP, rejected, then lost VWAP again on a fresh cross = trigger (vwap_reject_short, tier avwap, engine-confirmed at 10:10). Annotated 2-min replay with my reconciled fills (DATA-TRUST GATE PASS, +$110.86 penny-reconciled). The lesson: setup right every time — sizing was the variable (5-share probe stopped → 1-share churn → the 20-share A+). Upd 7/6: full library spine added — Technical—Daily (BB(20,2)+SMA 20/50/200+RSI stack) + 30-min panel, both with the Alpha Trend env cloud; environment traced to the indicator (Daily RED ✓ / 30-min MIXED at entry). LAB / unvalidated — discipline + narrative, not a validated $-edge.
Open →
DAY-2 CONTINUATION
d2c-001 · Meta-qualification · drafted 2026-05-07 · pending live-trade approval
The 5-filter qualification (prior-day trend day, RVOL ≥ 1.2×, no overhang, clean structure, liquidity) stacked on top of an executing branch (vwap-pull-in or fl2-fash-late). NOT a new entry trigger — the branch carries the trigger and sizing. Distinct from Backside Long, which requires −2 SD extension + double-bottom reclaim. Pending first scored live attempt.
Open →
BTC → IBIT AUTOPSY
2026-04-09 → 05-28 · MTF-VP swing · SIGNAL on BTC / EXECUTE on IBIT · chart-validated v2
The swing-scale management template. Entry $40.32 next-open on fresh alignment (MAE −0.32R day 1, stop never threatened). Signal-Event Timeline: T1 +1.65R → T2 +4.15R → 200-DMA wall test +5.22R (94% of the +5.55R MFE) → reduce-risk (1st close < 20DMA) +3.04R vs failure-confirmed exit +0.93R — acting on the first deterioration signal was worth +2.1R. T3 sat inside the 88–95k wall and never hit: don’t set targets inside known supply. Same event ladder now reparameterized for day-scale runners (TMT / EBAY specimen #1). Research record — read-only autopsy, no live changes.
Open →
NVDA
2026-05-14 · Alpha Trend Ride
Stage-2 markup trend ride — the $164 → $225.83 leg read, entered, and managed away. Multi-day hold playbook: qualification gates, fractal 5/10/20/50 MA-break exits, the 5/4 churn as the anti-case. Draft v1.
Open →
AMZN
2026-05-15 trigger · 5/18 follow-through pending · ★ BB-aware RSI panel
Wyckoff markup (Shannon 2-2). Brown/Cardwell bull-range RSI pullback — CMT L2. Phase A (counter-trend trim, RSI >70 cooldown) COMPLETE on 5/15 with RSI at 51 on the 50-pivot, 20 DMA broken intraday. Phase B trigger window OPEN. As of 5/18 close: chart RSI panel shows the full BB squeeze → BB↑ break → A peak → MA cross → BB↓ break → 50 test sequence auto-annotated. Canonical alpha-trend-ride doctrine doc (Hold Gates, Four-Layer Grade, exit hierarchy). Draft v2.
Open →
SOXL NEW
2026-05-18 · 20-DMA Reversion / Bull-Range RSI Pullback · 3x semi ETF
Leveraged expression of the same Brown/Cardwell template as AMZN — 3x Direxion semis. Phase A complete TODAY: a −10.99% session pulled RSI from 84 peak to 50.03 (Brown's pivot); intraday low $145.55 tagged the rising 20 DMA at $143.27. Phase B trigger window OPEN — primary close_above_rising_20_dma firing; awaiting 50-bounce + 5-DMA reclaim. 3x leverage = faster pullback, tighter void cushion. Draft v1 · first playbook with BB squeeze zone + BB-break auto-annotations on the RSI panel.
Open →
MSFT
2026-05-18 follow-through · Phase B CONFIRMED · ★ BB-aware RSI panel
Brown/Cardwell bull-range RSI pullback (CMT L2). The 5/14 "almost trend confirmation?" question is answered: 5/15 trigger (+3.85%, RSI 48 → 61, 20 DMA reclaimed) HELD; 5/18 closed $423.54 sustaining the move. Phase B continuation confirmed — this is the textbook bull-range continuation playing out. Chart RSI panel now shows the full phase + BB annotation set through 5/18. Draft v3.
Open →
XLE
2026-05-14 · Alpha Trend Ride — Case Study (1/12 → 4/8)
The case study of the framework working. XLE ran $46.10 (1/12) → $63.46 high (3/30), +37.7% peak — and didn't close below the 20 DMA the entire ~12-week trend. Tier-3 fractal gate completed 4/8 at $58.05. Brown/Cardwell Phase B continuation entries through every pullback; Phase A trim at 3/30 RSI > 70 cluster. Draft v1.
Open →
QQQ
2026-03-31 · VWAP Pull-In
Canonical reference playbook — the structural template the generator inherits. 17 sections from cover → data edge.
Open →
Bella Fade
Template · Backside Scalp (short)
Intern training template for the Bella Fade short setup. Walk-through of HH+HL above 9 EMA structure on extended VWAP-band moves. Long mirror = Backside Long (above).
Open template →

What lives here

Playbooks are per-trade dossiers for high-conviction or instructive setups. Each playbook is generated from playbook_data/<ticker>_<date>.json through gen_playbook.py against the canonical playbook_template.html. The 17-section spine covers Cover → Bigger Picture → Higher TF → RSI Edge → Fundamentals → Technical → Intermediate → Break of PDH → Strategy → Management → Executions → Review → Scorecard → Param Surface → Me vs Bands → SPY Correlation → Data Edge.

The select_best_trades.py step (still stubbed) will pick top-N trades for a week / month and call the generator in a loop — that's how the library grows from one reference playbook to a full archive.