# VWAP Pull-In Quality Gate

**Source:** [VWAP_PULL_IN_SPLIT_DIAGNOSTIC.md](./VWAP_PULL_IN_SPLIT_DIAGNOSTIC.md)
**Sample:** 127 plain Pull-In trades (-$364) + 15 Pull-In → Trend trades (+$367) on `main.trade_context` through 2026-06-03.
**Use:** pre-trade or live-review filter. Print this. Refer to it before reaching for a VWAP Pull-In setup.

> **Mechanism caveat (added 2026-06-04):** The slug `vwap-pull-in -> trend` is currently outcome-based, not mechanism-confirmed. It should be read as "profitable Pull-In with wide exit," not automatically as "true trend conversion." First 5-minute post-entry behavior determines management: expansion + slope acceleration = possible trend conversion; flat band + flat slope = rotation scalp; opposite state/slope = wrong-side escape. See [PULL_IN_TO_TREND_MECHANISM_AUDIT.md](./PULL_IN_TO_TREND_MECHANISM_AUDIT.md) for the data behind this caveat.

---

## 1. Executive summary

- **Plain VWAP Pull-In and Pull-In → Trend are two different setups that share the same entry zone.** The first one bleeds money in hot tape; the second one is the strongest edge in your dataset. Same playbook name, opposite outcomes.
- The discriminator is **RVOL and gap, not the entry technique**. Plain PI losers: RVOL 4.0, gap +2.1%. Plain PI winners: RVOL 1.4, gap +0.4%. PI → Trend winners: RVOL 1.14, gap **-0.9%**. Calm/controlled tape wins; hyped/elevated tape loses.
- **The first hour (09-10 ET) holds 91 of 127 plain PI trades and -$401 of the loss.** You're not overtrading per-symbol-per-day — every plain PI is 1-per-symbol-per-day. The leak is **selecting hot-tape symbols in the open hour across many days**, not repeat-entries on one name.
- **IBIT alone is -$185 of the -$364 total loss** (51%). Plus AMD, ETHU, RGTI = -$360 combined. Four symbols carry the bulk of the damage. Pattern: leveraged-product / high-RVOL crypto-adjacent names don't behave like a plain VWAP Pull-In.
- **The "Pull-In → Trend" outcome is a management decision, not a separate setup.** It's what happens when a Pull-In entered in calm tape gets held through the +2 SD break instead of being scratched. Today: 15 conversions in 142 PI-family attempts. The edge lives in the conversion.

---

## 2. VWAP Pull-In Quality Gate

| Gate | Condition | Action | Notes |
|---|---|---|---|
| 🟢 **GREEN** | RVOL 1.0–1.5 **AND** gap < 0.5% **AND** clean MTF (grand > 25) **AND** not emotionally extended | **Normal Pull-In allowed.** Still requires execution trigger (LB1 hold + EMA cross / volume confirmation). **Hold for trend conversion** if price closes above +2 SD — do not auto-exit at +2 SD. | This is where PI → Trend lives. Sample of 14 winners + 1 loser. |
| 🟡 **YELLOW** | RVOL 1.5–2.5 **OR** gap 0.5%–1.5% | **Reduced size only.** Take ONLY if structure is clean and you can articulate the trend-conversion thesis BEFORE entry. If you can't, pass. | Mixed-outcome zone. Filter aggressively. |
| 🔴 **RED** | RVOL > 2.5 **OR** gap > 1.5% **OR** symbol in {IBIT, AMD, ETHU, RGTI} **OR** entry < 10:00 ET in hot tape | **Do NOT treat as plain Pull-In.** Reclassify (see §3) or pass. | This is the leak. Plain PI losers cluster here. |

---

## 3. Setup reclassification rules

When RVOL/gap is hot (🟡/🔴 zone), do NOT call it a plain Pull-In by default. The setup is something else — name it correctly before entering, or pass.

| If the tape shows... | The real setup is... | Different rules apply |
|---|---|---|
| Stock making HH+HL after +2 SD close, you're entering on a pullback to +1 SD with VWAP slope rising | **Pull-In → Trend / Trend Trade** | Wider stop (VWAP), bigger target (+4 SD), hold the core |
| Stock extended >2 SD from VWAP avg with no setup support, momentum chase | **Chase / Extended** | This is a behavioral chase tag, NOT an edge. Pass. |
| Stock at intraday/HTF range edge, volume building toward the edge | **Range Break / Hitchhiker** | Wait for the break, stop outside the consolidation |
| Stock extended ≥2.5 SD BELOW VWAP with double-bottom failure forming | **Rubberband / Backside Long** | Band-anchored stop at −2 SD, full exit at VWAP, no trailing |
| Cannot honestly name what it is in calm-tape Pull-In terms | **Pass.** | The setup may not exist on this name today |

**Rule:** if the only honest label is "I want to be in this stock right now," that's not a Pull-In — that's a chase. The taxonomy has a slug for it (`chase-extended`) and its net P&L over the last 8 weeks is +$200/56 trades = $3.57/trade (basically flat). It is NOT an edge.

---

## 4. Practical tomorrow checklist

Before clicking buy on any VWAP Pull-In, answer **all seven** out loud:

1. **What is RVOL right now?** (target: 1.0–1.5; >2.5 = stop)
2. **What is the gap %?** (target: <0.5%; >1.5% = stop)
3. **Is this first-hour emotional tape?** (09-10 ET with hot RVOL = double-strict filters)
4. **Is price controlled around VWAP, or is it hot expansion?** (controlled = OK; expansion = either Pull-In → Trend management or pass)
5. **Is there evidence this can convert to trend?** (HH+HL structure, MTF grand > 25, +2 SD reach achievable)
6. **Is this really a plain Pull-In, or am I chasing an extended move?** (if you hesitate on the answer, you're chasing)
7. **If it fails to convert, where do I exit?** (LB1 / VWAP / time stop — pick one BEFORE entry)

If you can't answer 1–7 quickly, the trade isn't ready. Pass.

---

## 5. Post-entry management rule (mechanism-aware)

Once you're in a Pull-In, **the FIRST 5 MINUTES tell you which mechanism you're actually in.** Manage accordingly:

- **If band width EXPANDS AND slope ACCELERATES in trade direction within 5 minutes:** manage as **possible trend conversion** — trail wide, hold the core, don't auto-scratch at +1 SD.
- **If band width stays FLAT AND slope stays near zero:** manage as **VWAP rotation scalp** — target the opposite band, take the rotation, don't extend stops hoping for a breakout that isn't coming.
- **If slope or state moves AGAINST trade direction:** **exit or reduce quickly** — you are in a wrong-side trade; the outcome will be lucky or it will be ugly.
- **Do not blindly hold** because the classifier later labels the trade `vwap-pull-in -> trend`. That label is outcome-based, post-hoc. It does not validate the trend-conversion management decision you made at minute 5.

This is mechanism-aware management. The slug name describes the OUTCOME after the trade closes; the **trajectory in the first 5 minutes** is what tells you in real time how to manage it.

---

## 6. Risk rule candidates (research, not final hard rules)

These are research candidates from the diagnostic data. Validate against the next 2 weeks of live trades before locking in.

| Candidate | Estimated effect |
|---|---|
| **R1.** Skip plain Pull-In if RVOL > 2.5 | Excludes ~50–60 trades, recovers ~+$300 |
| **R2.** Skip plain Pull-In if gap > 1.5% | Excludes the +2% gap-up hyped names |
| **R3.** Combined hot-tape filter: skip if (RVOL > 2 AND gap > 1%) OR RVOL > 3 OR gap > 2% | Strongest combined — recovers ~+$400 |
| **R4.** Flag IBIT, AMD, ETHU, RGTI for sample review before any more plain Pull-Ins (consider blacklist OR require stricter R1 = RVOL < 1.5 on these names) | Recovers ~+$100 incrementally |
| **R5.** First-hour strictness: 09:00–10:00 ET requires R1+R2 in stricter form (RVOL < 2, gap < 1%) | Tightens the worst-performing hour |
| **R6.** In calm tape (R1 GREEN), default to managing for Pull-In → Trend conversion — do not scratch at +1.5 SD; hold for +2 SD close or VWAP-based stop | Captures the conversion edge currently under-fired (15/142) |

Combined R1+R2+R4 estimated P&L delta on the historical sample: **+$500 swing in `vwap_pullback_continuation` family** (from +$3 net to ~+$500 net).

---

## 7. What VCET should test next

These are the hypotheses VCET-at-entry validation (Phase 4) should answer empirically using `vcet_state_at_entry` joined to `canonical_strategy_slug`:

| Hypothesis | What we'd expect to see |
|---|---|
| **PI → Trend winners occur during EARLY_EXPANSION → EXPANSION_CONFIRMED → TREND_UP/DOWN** | ≥70% of PI→Trend wins should land in those 3 states at entry; <30% in ROTATION |
| **Plain PI losers occur during ROTATION, FAILED_EXPANSION_TENTATIVE/CONFIRMED, weak compression, or MEAN_REVERSION against the trade** | Concentration in regimes where the engine says "no directional commitment yet" |
| **VWAP-band expansion + slope + range-break confirmation separate converters from non-converters** | `vwap_band_width` and `vwap_slope` at entry should be higher (band fan + slope rising) on PI→Trend wins than on plain PI losses |
| **CBO range-break (v1.1) refines the conversion call** | Once CBO lands, EXPANSION_CONFIRMED + CBO confirmed should produce near-100% PI→Trend hit rate (currently 14/15 = 93.3% without CBO) |

If these hypotheses hold, VCET state becomes the **state filter** the gate above approximates with RVOL+gap. Until then, RVOL+gap is the best available proxy.

---

## 8. The one-sentence rule

**"Plain VWAP Pull-In is only a clean setup in calm, controlled tape; in hot tape, it must either be managed for trend conversion or be passed."**

---

## Glossary

- **Plain VWAP Pull-In** = `vwap-pull-in` canonical slug. Entry near VWAP, target +2 SD, stop −1 SD. Net loser in the current sample.
- **Pull-In → Trend** = `vwap-pull-in -> trend` canonical slug. Same entry, but managed as a Trend Trade once price closes above +2 SD. Net winner.
- **Hot / hyped tape** = RVOL > 2.5 OR gap > 1.5% — the empirical signature of the losing plain PI trades.
- **Calm / controlled tape** = RVOL 1.0–1.5 AND gap < 0.5% — the empirical signature of both Pull-In winners and PI→Trend winners.
