ATCH VWAP Strategy

Honest caveats — read first

• Published today (1 Nov 2026, v1.0.0.0). No live track record, downloaded twice at time of capture.
• Seller's own backtest, on one instrument (MNQ) and one period (2025-10-01 → 2026-10-30). The seller says it wasn't hand-picked; we can't verify that.
• No slippage modelled in their report (tick-replay fills), and their commission is $1.30/contract round trip.
• No IS/OOS split: the defaults may have been tuned on the same year the results come from. Their report also uses settings that differ from the published defaults (see contradictions C1, C3–C6, C9).
• Farm-fee re-cost: $16,791.16 net (vs $17,431.10 published) at MNQ $2.28 round trip (NT Lifetime + 1 tick per side). This is re-costed from their own trade list, not a farm run.
• Hunter 7 has no VWAP logic yet (UseVWAPBias is an empty placeholder). The +61 change request (49 new switches, default off) is with the H7 Builder. No farm backtest has been run; nothing promoted.
• The PC4 cross-check proves their trades sit on real MNQ prices at the stated minutes. It does not prove the strategy's logic generates them, or that the results will repeat.

Model

Each day a VWAP (HLC3, 15m candles by default) anchors at 09:30 ET with ±1.28σ std1 bands. On each closed chart bar a long arms when close > rising VWAP and < upper std1 (short = mirror). The trigger is the first counter candle (Immediate) or the first close beyond the prior bar's extreme (Wait For Pullback). Before entry: % change vs the HTF close N back ≥ 0.1%, trigger bar doesn't touch a std1 level, closes on the trend side of the VWAP and is still inside the band. Stop: VWAP / SuperTrend / last swing (5) / Auto (nearer of swing and std1, VWAP fallback) + offset. TP: W/L ratio (2 default) or points / $. Optional %-breakeven and SuperTrend trailing. Sizing: $-risk adjusted ($100 target, $120 max, qty 1–6). Daily stops; flat 30 s before session close.

Session clock: VWAP anchor 09:30 ET = 23:30 AEST (US summer) / 00:30 AEST (US winter). Report entries all fell 09:50–11:29 ET (23:50–01:29 / 00:50–02:29 AEST). Report exits by 16:55 ET (06:55 / 07:55 AEST).

Check sequence: 1) in session → 2) daily stop not hit → 3) update HTF VWAP + bands → 4) arm → 5) trigger candle → 6) pre-entry checks (VWAP side, min % change, no std1 touch, inside band, SuperTrend filter); any fail = keep waiting → 7) stop, qty (skip if < Min Qty), TP → 8) breakeven / SuperTrend trail; flat before close.

Rules (55) — labelled against Hunter 7 +60

All VWAP rules are Missing: H7's UseVWAPBias is an empty placeholder (H7 Builder, 1 Nov 2026).

ATCH VWAP pullback locked rules template one — A+ checklist (draft for David)

1. In trading session (09:30–16:00 ET default; report evidence 09:30–11:30)
2. HTF VWAP rising (long) / falling (short), last vs previous point
3. Close above VWAP and below upper std1 (long) — mirror short
4. First counter candle (Immediate) or close beyond prior extreme (Wait For Pullback)
5. Price change vs HTF close N back ≥ Min % (0.1 default; 0.15 in examples)
6. Trigger candle doesn't touch a std1 level; closes on the trend side; still inside band
7. Stop by chosen mode + offset; qty from $ risk within max risk, Min ≤ qty ≤ Max
8. TP = W/L ratio × stop (2 default; 1.1 in examples)
9. Optional breakeven (Example 5: −90%) and SuperTrend trail
10. Flat 30 s before session close; daily stops respected

Contradictions and unknowns (switchable, unresolved)

• C1 W/L ratio 2 (default) vs 1.1 (Ex 1) vs ~1.26 (report: median winner 86.25 pts / median loser 68.25 pts)
• C2 Breakeven −90% in Ex 5 vs off by default
• C3 Max Qty 6 vs report quantities 7, 12 and 19
• C4 Target risk $100 default vs the "$400 risk" report
• C5 HTF 15m default vs 10m in Ex 1; earliest report entry 09:50 ET fits 10m
• C6 Min price change 0.1% vs 0.15% in Ex 1/4
• C7 Price-change lookback default not stated vs "two VWAPs before" (Ex 1)
• C8 Stop-mode default not stated (examples show std1, swing, swing)
• C9 Sessions 09:30–16:00 default vs all 226 report entries in 09:50–11:29 ET
• Unknowns U1–U10: intrabar VWAP, disarm rule, WFP needs counter candle, "touch std1", Auto "allowed risk", exec-TF band, BE offset, Adaptive SuperTrend, SuperTrend MA/smooth use, qty rounding.

A+ examples (seller's chart images)

Trade list (226) — seller's backtest, captured in full

Time zone: the list is in Europe/Zurich time (page script pfReportTimezone; all 226 match their chart API's UTC timestamps to the minute after conversion). Zurich − ET = 6 h, except 20 trades in the DST-gap weeks (late Oct and March), where it's 5 h. Every row has ET and UTC times, entry and exit price, qty, P&L and commission. Stops are filled for 89 full-size losers and targets for 119 full-size winners (their exit prices); the other rows carry the stop-distance range implied by $400 sizing.

As published: 226 trades · net $17,431.10 · PF 1.48 · win 54.4% · max DD $-2,262.20 · avg $77.13
Re-costed at farm fee $2.28 RT: 226 trades · net $16,791.16 · PF 1.46 · win 54.4% · max DD $-2,285.72 · avg $74.30 (from their list, not a farm run)
Long (farm fee): 101 trades · net $11,131.96 · PF 1.75 · win 59.4% · max DD $-1,656.70 · avg $110.22
Short (farm fee): 125 trades · net $5,659.20 · PF 1.26 · win 50.4% · max DD $-2,873.30 · avg $45.27

PC4 1-minute cross-check — 185 trades checked, 175 inside the bar on both entry and exit

Run on PC4 with a script (F:\trader_packs\ATCH_VWAP\scripts\atch_crosscheck.py) against PC4's NinjaTrader MNQ 1-minute export (UTC, bar-end stamps). Every 2026 trade was checked: 185. The 41 trades from Oct–Dec 2025 can't be checked because PC4 has no MNQ 12-25 data. Their prices are a back-adjusted "MNQ 12-26" series. One constant offset per contract fits every trade:

Result: entries 181/185, exits 179/185, both 175/185. The 10 misses are all within 0.45–0.55 pt (about 2 ticks) of the bar's range. That's consistent with PC4's export rounding prices to 0.1 and a ~0.75-pt basis between their feed and PC4's on the 12-26 contract. Caveat: 1-minute bars are often 20–60 pts wide, so landing inside a bar is a loose test on its own. The strong evidence is that one fixed offset per contract fits all 185 trades. A sample of 21 rows:

Hunter 7 status

The +61 change request (plugin AV_VWAP, 49 switches, all default off, plus the test plan: parity → reproduction gate on the 226 trades → same period at farm fees → IS/OOS/hold-out → equity curves → Stage One instruments) was sent to the All_Hunter_7 Builder by David on 1 Nov 2026. This lane does not edit H7 code. No farm results yet.

Claims log (claims, not facts)

• MNQ 1m $400 risk: 226 trades, net $17,431.10, PF 1.48, max DD −$2,262.20, 54.4% win. The totals recompute exactly from their list; the strategy itself isn't verified by a farm run.
• "Intraday tick-level backtesting ... 99% conformant to real-time".
• "The backtest period wasn't hand-picked."

Data-gap note

• PC4 MNQ 1m covers 1 Jan → 30 Oct 2026 (contracts 03-26, 06-26, 09-26, 12-26). Oct–Dec 2025 (41 trades) isn't covered — affects results: yes, those trades are unchecked.
• Their prices are back-adjusted to MNQ 12-26; offsets were derived per contract (above) — possibly (a constant ~0.75-pt basis remains).
• PC4's export rounds prices to 0.1 pt; tolerance ±0.30 pt — possibly (accounts for the 10 two-tick misses).
• Their fills are tick-replay; farm bars are 1-minute, so stop-vs-target order inside one bar isn't resolved — possibly, for later farm runs.
• Their chart API returned no OHLC bars, only trade times; prices were checked against PC4 data only.

Pack

PC4 F:\trader_packs\ATCH_VWAP\: master rules, rules CSV, 226-trade CSV (raw + converted), PC4 cross-check CSV and summary, summary JSON, chart-API JSON, H7 change request v61, source page, images, script. sha256 list in SHA256SUMS.txt.