VWAP Pull-In — Forward-Test Tracker

Static live-use capture · entries are saved in this browser only (localStorage) — export to CSV/JSON to keep them · Schema: vwap_pull_in_forward_test_schema.json · no server, no database.
Forward-test whether VWAP Pull-In improves when James separates calm pull-ins from hot tape and cuts failed pull-ins within 3–5 minutes.

Status

Strategy Library promotion HOLD
Scorecard production integration HOLD
VCET production alerting HOLD
Forward-test capture START

Why: the walk-forward / OOS validation showed the entry gate alone flips negative out of sample — the durable edge is post-entry management. Collect 20–50 live attempts under the frozen rules before promoting anything.

Frozen rules being tested

RuleWhat we're testingOut-of-sample status
Green / Yellow / Red gatedoes it separate expectancy live?entry-gate alone flipped OOS
Gap RED condition (|gap|>1.5%)big gaps keep losingheld directionally
Instrument-type warningmegacap-tech vs crypto-proxy / small-cap / high-betarelative filter only
3-minute fade exitfade at 3 min = failed pull-inHELD — strongest
5-minute rotation vs expansionclassify the mechanism at 5 minheld
10–15 minute trend confirmationexpansion identifies Pull-In → Trendheld directionally
Chase / Extended reclassificationextended/FOMO not logged as a pull-insentinel, not an edge
Plain Pull-In managed as rotation unless trend confirmsdon't hold a pull-in like a trend tradecore finding

RVOL is captured for context only — the cross-tab showed it doesn't separate winners from losers once instrument type is controlled. Log it; don't gate on it.

Capture a trade

Fill what you have (only Date, Ticker, Side, Setup, Gate, Instrument type, Rule-followed, Mistake-type are required) and click Add entry. Saved locally; nothing leaves your machine until you export.

Progress toward minimum sample

Targets: 20 attempts = initial review · 50 = consider rule changes · at least 10 each of Green / Yellow / Red-or-pass · 10 post-entry fades · 10 expansion confirmations · keep ETF/crypto separate from single-stock · never promote from one symbol or one regime.

Captured entries (0)

No entries yet. Add your first VWAP Pull-In attempt above.

Forward-test grading rules

Grade the process, not the P&L. Use these to fill rule_followed + mistake_type honestly.

Grades WELL — even at a loss

  • Gate was valid
  • Classification was correct
  • Size matched the gate
  • Invalidation was defined before entry
  • The 3–5 minute checkpoint was respected

Grades POORLY — even at a profit

  • It was a RED-gate trade
  • It was actually Chase / Extended but tagged as Pull-In
  • Size was too large for a Yellow or Red condition
  • No invalidation was defined
  • The 3–5 minute failure was ignored
  • A plain Pull-In was held like a trend trade without confirmation

Minimum sample requirements before promotion

Daily review

After each trading day, answer:

  1. Did I call the setup correctly?
  2. Was this calm tape or hot tape?
  3. Did I size according to the gate?
  4. Did I respect the 3–5 minute checkpoint?
  5. Did I hold only if 10–15 minute expansion confirmed?
  6. Did I tag the mistake in plain English?
  7. What is one rule I will follow better tomorrow?

Promotion checklist

The rule can move toward the Strategy Library only if all hold:

How this eventually plugs into the system

LayerRoleStatus
Strategy Library /strategies/Official setup definitions (the WHAT)HOLD
VCET / live classifierReal-time gate + reason codes at the ticket (the WHEN)HOLD
Execution Scorecard /scorecards/Process grading + mistake tagging (the HOW)HOLD
This trackerForward-test evidence layer BEFORE promotion (the PROOF)START

The tracker feeds the other three: once 50 attempts confirm the frozen rules across symbol types and regimes, the rules graduate into the Strategy Library definitions, the VCET live gate, and the Scorecard grade logic — in that order. Until then, everything upstream stays HOLD.

Related

Static forward-test capture · localStorage only, no server / no database · no edits to /strategies/, /scorecards/, VCET production, gen_scorecard.py, _scientific_overlay.js, taxonomy, or DB · nothing promoted to production.