Final Report: Getting Funded on NQ — What the Data Actually Supports

Prepared for a trader with 5 losing years. Everything below was mined from the full Edgeful catalog (12 months, 2025-07 → 2026-07), cross-checked on NQ/ES/YM, and adversarially re-verified from independent API pulls. Where a number didn't reproduce, the corrected number is used. Where the honest answer is "unknown," it says unknown.

1. Straight talk

The audited base rates: in the two best datasets ever assembled (Taiwan, complete exchange records, 1992–2006; Brazil index futures, 2013–2015), fewer than 1% of day traders were persistently profitable net of costs, 97% of those who persisted 300+ sessions lost money, and performance did not improve with experience — five losing years is the norm, not an anomaly. Prop trailing drawdowns are calibrated so that a routine losing streak at "full" size kills the account: at a 60% win rate you have a ~92% chance of a 4-loss streak within 100 trades, and every strategy below survives that math only at 1–2 micros. The entire evidence base here is one bullish year of data — no edge below has ever seen a bear regime, so every number is a ceiling, not a promise. What separates the documented profitable minority is not entries: it's fixed-fractional risk, low trade frequency in filtered conditions only, obsessive cost control, and mechanical rules — the literal opposite of discretionary scalping. Only one strategy in this research passed verification. Three more are conditionally promising but gated behind bar-level backtests that have not been run. That is the honest inventory; anything padded beyond it would be another dream.

2. The strategies

Verified and deployable: 1 (Grade B). Conditional, do-not-fund-until-backtested: 3 (Grade C). The C-grades are included only because fewer than three passed — each carries an explicit go/no-go gate. Do not skip the gates.

STRATEGY 1 — NQ Friday 15-min ORB Continuation ★ GRADE B (the only verified pass)

Rules: - Trigger: Friday, 09:45 ET, 15-min opening range complete. First break of an ORB edge, only if it prints before 10:30 ET. - Orientation: arm the long side only on low-formed-first days (74.8% high-breaks-first, n=115 — verified). The short-side mirror did not verify (66%, not the claimed 74–76%) — skip short arming, or take shorts knowing they're the weaker, regime-untested leg. - Entry: stop order at the ORB edge. - Stop: ORB midpoint (0.5× ORB ≈ 55 NQ pts ≈ $110/MNQ). Never tighter — 31% of eventual winners retrace ≥50% (verified 9/29). - Target: 50% off at +0.5× ORB beyond the edge; runner held to 15:55 ET close. Note: the claimed 79%/69% partial-reach did not reproduce — the honest unconditional figure is 60.5%/51.2%. - Filters: no ORB-size filter (edge survives all size cells; Crabel narrow-range preference is refuted). Reduce size on NFP Fridays.

Verified numbers: Continuation to close 66.0% (33/50), identical at 15-min and 30-min. H1 65.4% / H2 66.7% — dead flat. Long 65.5% / short 66.7% — symmetric, which clears the bull-regime objection. Friday double-break only 38% vs ~50% base. Break-before-10:30 on 97.6% of single-break days (124/127).

Expectancy after costs: deployable win ~56% (66% core minus midpoint-stop conversions), avg win ~+1.15R, avg loss −0.9R → +0.175R ≈ +9.6 NQ pts/trade. Per MNQ: $19.30 gross − $2.80 costs = +$16.50/trade (costs = 2.5% of risk). At 2 MNQ × 2.2 trades/mo = ~$73/month. Break-even win rate ~46–48%. Yes, that's small. It's also real.

Prop fit: Topstep 50K — worst day is one stop-out: 2 MNQ ≈ $226 vs $1,000 DLL (trivially safe); a realistic 6-loss streak = $1,356 vs $2,000 trailing MLL (survives; 3 MNQ = $2,034 = blown, hence the hard 2-MNQ cap). Typical winner at 2 MNQ ≈ $260 — clears Topstep's $150 winning-day threshold. One trade/week cannot feed the 5-winning-day payout cycle alone: this is a sleeve, never a standalone. Lucid: fine everywhere; good cheap seasoning for a LucidFlex scaling buffer.

Manual vs algo: Manual is trivial (one decision at 09:45, one bracket). Attended semi-automation (ATM bracket) is Topstep-compliant. Fully automatable on Lucid via Tradovate/TradersPost webhooks.

Three biggest ways it fails: (1) Selection artifact — Friday is the best of 5 weekday cells over a 50.4% base (Thursday sits at 38%); roughly half the edge could be mining, disclosed but unresolvable at n=50. (2) Regime — never tested in a bear market; the flat H1/H2 split is the only comfort. (3) Target math — the partial-reach figures in the original spec were outcome-conditioned (circular); real reach is ~60/51%, already haircut into the +0.175R above, but further slippage there erodes a thin edge.

4–8 week validation plan: Weeks 1–4: sim on TopstepX/Tradovate, both the aligned and unfiltered variants, log realized R per trade. Go/no-go after 8 Fridays: realized fills within 1 tick of modeled, no rule ambiguity. Weeks 5–8: live in a Combine at 1 MNQ as the Friday sleeve of the broader book. Kill-switch: rolling 15-Friday hold-to-close continuation < 55% → back to paper.

STRATEGY 2 — NY Big-First-Candle Day-Hold ⚠ GRADE C (best conditional; gate: 5-min bar backtest)

Rules: - Trigger: 09:30–09:35 candle range ≥0.2% of price (~≥50 NQ pts). Direction = candle colour. Skip sub-0.2% (red sub-0.2% is a 51.4% coin flip — verified). - Do NOT skip Thursdays — the verifier proved that filter was cross-metric contamination (Thursday is actually 81.2% in this strategy's own metric; Tuesday is the weakest at 74.3%). Drop it. - Entry: longs full structure — half market at 09:35 close, half limit at the candle 25%-retrace level (the original midpoint-limit justification was mis-derived; move it to the 25% level or measure the true fill rate in the backtest). Shorts half size, fade-structure only, never extension bets (red extension 30–41% both halves). - Stop: opposite extreme of the signal candle (median ≈ 120 NQ pts ≈ $240/MNQ). - Target: time exit 15:55 ET. Longs: optional 30% scale at PDH. Shorts: 50% at candle low, no holds below it. - Displacement filter: 09:30 open ≥0.6% beyond midnight open in candle direction → retrace odds collapse to ~30–34% (verified, but n=56 thin) → full size; opposite displacement → skip.

Verified numbers (cleanest reproduction in the dataset — every claim matched to the decimal): ≥0.2% day-direction continuation NQ 83.4% (141/169; green 84.9/red 81.9), ES 82.3%, YM 81.3%; monotonic in size (79.6% → 90.9% at ≥0.9%); sub-0.2% control 58.6%. The tradeable margin: green ≥0.2% close-above-candle-HIGH 55.8% NQ (48/86), ES 55.6% — YM fails (36.8%). Close-back-above-open floor 84.9%. H1/H2: core 86.2/80.9; extension leg drifts 57.9 → 54.2 (toward coin flip).

Expectancy after costs: honest band on 1 MNQ is $0 to +$72/trade ($0–$865/mo). Central inferred case (60% × 1.2R): +$72/trade net. Candidate's own gate floor (55% × 1.0R): +$20/trade ≈ $240/mo. Bear case with stop-sweep contamination (days that sweep the candle-low stop then close above the high are counted wins in the 55.8% but are realized losses — unmeasurable from Edgeful, no MAE data): ~breakeven. Positive expectancy is plausible but not established. That is why this is a C.

Prop fit: Topstep — the archetypal attended manual trade: one decision/day, $240 max daily exposure vs $1,000 DLL; winners cluster $150–500, feeding the XFA 5×$150 payout engine without tripping the best-day cap. Max streak at 55–60% win ≈ 5–6 losses = $1,200–1,440 at 1 MNQ — survives the $2,000 trail with margin; 2 MNQ breaches it. Lucid: fine on all programs.

Three biggest ways it fails: (1) The 60%×1.2R geometry is inferred, not measured — stop-sweep leakage could drag realized win% to ~50–52% where costs decide the sign. (2) The extension leg decays toward 50% in H2 and fails outright on YM — the tradeable margin is thin and possibly regime-tinted. (3) The pullback half is adversely selected (fills more on losing days) and forfeits ~43% of clean runners.

Validation plan (gate is mandatory): Weeks 1–2: 5-min bar backtest over the same 12 months — half at 09:35 close, half at 25%-retrace limit, opposite-extreme stop, 15:55 exit, counting stop-sweep days as losses. Gate: ≥55% win at ≥1.0R net, plus a 2022–23 bear-period check on the green side. Fail → shelve. Pass → weeks 3–6 sim (12+ signals), realized R within 0.1 of backtest → weeks 7–8 live at 1 MNQ long-only (the short side adds little and can be cut without losing the verified core). Clears → this graduates to B and becomes the workhorse.

STRATEGY 3 — Prior-Day Magnet, LONG-ONLY reframe ⚠ GRADE C (gate: MAE bar test)

Rules (as corrected by verification — materially different from the original spec): - Trigger: open inside yesterday's RTH range AND yesterday closed green → long, target = PDH touch. Drop the short side entirely — 55.6% touch ceiling in a bull year minus any stop-out haircut lands at/below breakeven at 1:1; "half size" just halves the bleed. - Fix the Monday filter: skip only prev-RED Mondays (PDL touch 31.2% — toxic). Keep prev-green Mondays — Mon prevGreen→PDH is 75.0% NQ, the best cell in the table. The original blanket Monday skip discarded the strategy's best day. - Entry: limit on first pullback toward prior-day midpoint, or market at open if already driving toward PDH from above mid. - Stop: 0.5× the open-to-magnet distance below entry, hard cap $250 = 125 MNQ pts (the original spec's "150–175 pts" arithmetic was wrong by 20–40%). - Target: full exit at PDH touch. No extension holds (PDH-breaks extend ≥0.5× only 31%, ≥1× only 5.2% — verified). - Overlay: week opened above prior-week midpoint → +1 confidence (PWH broken 96.6% of such weeks, n=29 — verified stronger than claimed).

Verified numbers: prevGreen+insideOpen→PDH touched 72.2% NQ (n=126), ES 70.5%, YM 63.7% — every one of 12 half-cells across 3 tickers above base (NQ base 59.0%), symmetric both directions, the strongest anti-regime evidence in the daily families. But plan on the H2 figure: 66.7%, not 72.2% (H2 decayed 12–20 pts everywhere; ES to barely +4.7 over base).

Expectancy after costs: touch% is a ceiling, not a win rate — deployable win = touch% × (1 − P(stop hit before touch)), and that haircut is unmeasured. Breakeven at 1:1 needs the haircut < ~29%. At 58% realized: +$36.70/trade → ~$385/mo on 1 MNQ; pessimistic ~$225; claimed-but-unverified ~$910. Better framing per the verifier: entry near open with a 0.5×-distance stop is actually ~2:1 R:R (breakeven 33%) — if the bar test shows even 50% win at 2:1, that's +0.5R and this upgrades to B.

Prop fit: Topstep — one bracket at the open, $250 max/day vs $1,000 DLL; near-certain 4–5-loss streak = $1,000–1,250 vs the $2,000 trail — survivable at 1 MNQ, near-certain death at 2 MNQ from a cold start. Winners ≈ $250 clear the $150 day threshold. Semi-automated bracket-at-open is trivially attended-compliant. Lucid: fine; trade count suits Pro's 3-day cycles. NQ/ES only — YM too thin.

Three biggest ways it fails: (1) The stop-before-touch haircut is the whole game and is unknown — 29% breakeven margin is plausible in either direction. (2) H2 decay was material (ES 80.3 → 60.6) — the effect may be fading. (3) Part of the lift is mechanical proximity (green close parks the open near PDH) — real alpha over "always bracket toward the near side" is smaller than the headline.

Validation plan: Weeks 1–2: 1-min bar test measuring MAE from entry to magnet touch. Gate: mid-based stop preserves ≥58% at 1:1, or ≥45% at the 2:1 framing. Weeks 3–6: sim, 20+ trades. Weeks 7–8: 1 MNQ live, long-only. Kill-switch: rolling 30-day touch rate < 60%.

STRATEGY 4 — London Asian-Range Break Drive ⚠ GRADE C (gate: full bar backtest; the London slot, if you want one)

Rules (as corrected): - Trigger: first 1-min close beyond the Asian Range (19:00–03:00 ET high/low) between 03:00 and 09:00 ET. First signal only. Drop the Tuesday filter — it was a mined single-ticker cell (p=0.066 uncorrected across 15 searched cells; the claimed ES/YM "echoes" were just those tickers' base rates). Honest planning number: 67%, not 70.1%. - Entry: market at trigger close, break direction. - Stop: 0.5× AR from entry (~55–90 NQ pts). Warning: this distance is a proxy, not a measured stat — no AR-extension endpoint exists in Edgeful. - Target: time exit 11:25 ET. Optional 25% scale at +0.5× AR (only ~31% extend — no runner holds). - Caution: the 09:00 cutoff is an invented parameter (no break-time data exists) — verify it in the bar test. NQ/MNQ only (YM H2 fell to 58.2%; drop it from the evidence).

Verified numbers: pooled continuation 67.0% NQ (n=206), ES 65.4% (n=217), YM 63.3% — base edge real on all three at n≈200. NQ H1/H2 65.7/68.4 — stable. Down-side 70.1% > up-side 64.7% on NQ — anti-trend side stronger, so not a bull artifact (though ES reverses and YM is flat — the symmetry is noisy).

Expectancy after costs — read this carefully: the measured 67% is P(close beyond the level), not P(profit vs your entry), and stop-outs on counted winners aren't visible. At the inferred geometry (63% vs entry, +0.9R winners, R = $140/MNQ): +$24/trade net, ~+$316/mo at 13 trades. But at 63%/0.6R winners it's +$1/trade — negative after costs — and at 58%/0.75R it's breakeven. The sign of this strategy flips inside the plausible parameter range, and the two deciding parameters are exactly the ones no Edgeful endpoint measures. That is the definition of "backtest first."

Prop fit: Lucid is the natural home — the 42% median intraday retracement on winning days is real, and Lucid's EOD-only trail ignores it; fully automatable via Tradovate webhooks (VPS allowed); 11:25 ET exit is hours inside the 4:45 PM EST flat rule. Topstep: viable only as attended trading (no VPS) — fine since it's UK daytime — but note a 1-MNQ winner (~$126 avg) sits below Topstep's $150 winning-day threshold, so at minimum size this strategy cannot feed Topstep's payout engine. Run it on Lucid.

Three biggest ways it fails: (1) Entry/stop geometry unmeasured — expectancy sign unproven (the whole C-grade). (2) London's structurally low volatility means small winners in points; costs eat a bigger share on any decay. (3) The broader literature found zero credible persistent London-hours edge on US index futures (the one documented edge, the 2–3 a.m. drift, decayed to zero post-2021) — priors are against this session, and one bullish year of Edgeful data doesn't overturn them.

Validation plan: Weeks 1–2: 1-min bar backtest — fill vs level, MAE distribution, 0.5×-AR stop-out rate on counted winners, close-vs-entry P&L in R, and the 09:00 cutoff. Gate: ≥60% win vs entry with ≥0.75R average winner. Fail → abandon the London session entirely (the literature says you lose little). Pass → weeks 3–8 sim then 1 MNQ live on Lucid. Re-run the H1/H2 split quarterly.

3. What was rejected and why

One line each. Do not re-try these out of hope.

• London 04:00 IB "Failed-Retest" Rejection — REJECTED (F): the thesis was disproven by its own data — the zone field is relative to the last-formed extreme, so 45/48 "failed retest" days were actually red candles closing on their LOW (trivial momentum, not rejection); the short leg's trigger condition occurs ~3 days/year, stop math was wrong 3×, and the salvageable momentum leg prices to breakeven-at-best with unresolvable sequencing.
• NY IB Colour Continuation w/ Retest — REJECTED for live capital (C, paper only): the colour edge is real (78%/73%, cross-ticker), but day-level bounding puts trade expectancy at floor −0.13R / midpoint +0.09R / H2 +0.04R — below its own +0.15R gate; at the only survivable size (1 MNQ) that's ~$156/mo against near-certain 6–8-loss streaks — a real but untradeably thin edge.
• Tuesday-skip filter (AR-Break): mined single-ticker cell; the ES/YM "confirmation" was their base rates. Dead.
• Thursday-skip filter (Big-First-Candle): cross-metric contamination from a different report family; Thursday is actually strong (81.2%) in the relevant metric. Dead.
• Short side of the Prior-Day Magnet: 55.6% touch ceiling in a bull year minus any stop haircut = breakeven or worse. Dead as specified.
• Formed-first "mirror" orientation (Friday ORB shorts): 66%, not the claimed 74–76%. Only the long-side 74.8% verified.
• Overnight drift (long 2–3 a.m. ET): Fed-documented, and Fed-documented as decayed to ~zero since 2021. Do not build.
• Gap fade on NQ/MNQ: systematically falsified after costs (arXiv walk-forward study — failed at every entry time).
• VWAP-band mean reversion entries: no peer-reviewed evidence exists; the circulating win rates are vendor marketing.
• Naked NR7/Crabel contraction patterns: not tradeable after costs in modern re-tests; the Friday ORB data actively refuted narrow-range preference.
• Discretionary scalping generally: 95–99% loss rate in complete audit-trail datasets. This is the thing the last 5 years already tested.

4. Portfolio & execution plan

⚠ Correlation warning first: Strategies 2, 3, and (rejected) IB-colour are all the same underlying edge family — "NY open drives, trend day continues." On a clean trend day they all win together; on a chop day they all lose together. They are diversified in trigger, not in risk. Rules that follow from this:

1. Max 2 NY positions per day, combined NY risk ≤ $500 (micros).
2. One-strike rule: if the first NY trade stops out, the second may only be taken at half size; two NY stop-outs = done for the day (self-imposed, well inside any DLL).
3. London (Strategy 4) is the only genuinely uncorrelated sleeve — different session, different range logic.
4. Personal daily stop at $500 on a 50K — half the Topstep DLL, always self-enforced before the firm enforces it.

Firm assignment:

Buy one Topstep 50K Combine ($49/mo) and, only after Strategy 4 passes its backtest, one LucidFlex 50K (~$175 one-time). Do not stack accounts before a single strategy is proven live. On Lucid: withdraw early and often — the firm is ~1 year old and has rewritten rules four times; counterparty risk is real.

Realistic monthly expectations at minimum sizing (this is the part to read twice):

Against that: $49/mo Combine subscription, $149 XFA activation, ~$175 Lucid Flex, data/platform fees. Year one at minimum size is approximately break-even to modestly positive even if everything works. The product of year one is not income — it's a verified live track record and a funded account with the MLL locked at breakeven. Anyone selling you more than that from a $50K sim account with a $2,000 trailing drawdown is lying.

Scaling path (mechanical, no discretion): 1. Combine: trade the book at the sizes above; target ~$150–400 days; the <50%-of-target best-day rule is automatically satisfied at these sizes. Expect 2–4 months to pass, not 2 weeks. 2. XFA, phase 1: you start at 2-mini scaling cap anyway — stay at 1 MNQ per strategy until +$2,000 EOD closed profit locks the MLL at breakeven. This is the single highest-leverage move in the whole plan: after the lock, the account cannot die below start balance. 3. XFA, phase 2: after lock, step to 2 MNQ per strategy per +$1,000 banked. Run a DLL on the account — it doubles the payout caps ($2,000 → $4,000 per cycle on a 50K). 4. Payouts: request every time 5 × $150+ days accumulate; minimum $125, Aeropay free. On Lucid, pull every cycle. 5. Add contracts, never new strategies, on schedule — new strategies only via the full gate process above. Re-verify every edge quarterly (rolling H1/H2 split); kill-switches as specified per strategy. 6. LFA (Topstep live) is a lottery ticket — 0.71% of XFA traders were called up in 2025. Plan the P&L as if it never happens.

Automation notes: Topstep — attended only, own PC, no VPS ever (Terms of Use violation = account pulled regardless of P&L); TopstepX API $29/mo if you want programmatic brackets while present. Lucid — full automation legal including VPS/webhooks (Tradovate + TradersPost); keep order rates human-plausible, holds >5 seconds. Both: flat by 15:10 CT (Topstep) / 4:45 PM EST (Lucid) — all four strategies exit hours earlier by design.

5. Verdict

One strategy survived verification, and it makes about $73 a month at the only size that survives the drawdown math — the other three are unfunded hypotheses until their bar-level backtests clear, and the correct response to that sentence is relief, not disappointment, because it's the first time in five years the number in front of you is real. The edge, if you have one going forward, will not come from a better entry — it comes from 1–2 micros, a fixed stop on every order, at most three mechanical decisions a day, and the discipline to shelve anything that fails its gate. Run the backtests before the money; if only the Friday trade ever passes, trade only the Friday trade.

Full-Universe Verdict — 40 Symbols, 9 Templates, 75 Qualifying Hits, 12 Adversarially Verified

1. Direct answer

After sweeping all 40 viable symbols (25 futures, 11 forex, 4 crypto), the prior verdict is largely confirmed with one genuine upgrade and one diversifier. Exactly one finding beat the NQ playbooks head-to-head under adversarial verification: RTY 60-min IB-colour continuation (81.3%/78.6%, n=208, both sides both halves, better geometry than NQ's version, ~+$60–95/trade full RTY) — that's a new playbook. One more survives as an additive overlay in an uncorrelated asset class: FX midnight-open continuation (6B/GBPUSD) — 85.6% no-retrace at ≥0.2% displacement (n=111, stable, 5-instrument cluster), but it's a stop-survival rate, not a measured expectancy, so it gates on a target study. Everything else that out-scored NQ on paper — crypto PDR conditionals (+29pts!), ZW's 77.7%, SOL's 97.3%, MBT's 87.7% — reproduced exactly and then died to mechanical nulls, hindsight conditioning, or untradeable geometry. NQ/MNQ remains the best execution home for the index templates; the PD-magnet is now the most-validated finding in the book.

2. Cross-symbol scoreboard

Named as noise/killed after verification: MBT ORB, SOL London 97.3% (as an edge), BTCUSDT PDH→green, ETHUSDT PDL→red, ZS IB-colour (as a cross-symbol claim), ZW PD-magnet (as a standalone), ZN/ZB IB stats (as trades), NG Thursday EIA (+39.5 H1 / +2.6 H2 — the multiple-comparisons signature), CL Wednesday EIA (sub-threshold), AUD Tuesday 17/17 and EURGBP Monday 19/19 (in-sample artifacts), FX weekend-gap fade (90.8%→42.5% H2 collapse).

3. New strategies unlocked

3a. RTY IB-Colour Continuation — Grade B- (the only verified beats-NQ finding)

• Trigger: 60-min IB (09:30–10:30 ET) on RTY. IB candle colour known at 10:30 — verified NOT hindsight.
• Entry: break of the colour-side IB extreme (green→IB high, red→IB low), wick criteria.
• Stop: opposite IB side (~1.0x IB, est. 15–20 pts ≈ $900 full RTY / $90 M2K).
• Target: +0.5x IB extension (~$450 / $45). This is a direction edge, not a distance edge — median max extension 0.25–0.5x IB; do not run it as a runner strategy.
• Verified numbers: red 81.3% (n=91), green 78.6% (n=117) vs bases 47.6/52.4; honest incl.-double-break 78.3%/80.8% (n=242); H1/H2 all cells +23 to +35; only 13.5% of trigger days double-break, 5.7% run ≥0.5x adverse. Haircut forward hit rate to cluster mean ~76% (best-of-4 selection).
• EV & costs: ~+$60–95/trade full RTY after ~$9 costs under flat-scratch assumption; the ~37% scratch bucket is unresolved — if scratches average −0.1x IB, EV compresses toward +$60 full / ~breakeven M2K. ~192 triggers/yr (~16/mo, ~4/wk). Topstep/Lucid; RTY endpoint slow (cosmetic only).
• Gate: 30-trade sim log resolving scratch-day P&L before live sizing; drop to filter status if realized hit <70% or scratch drag >0.15x IB.

3b. FX Midnight-Open Continuation Overlay (6B primary; 6E/6A cluster) — Grade C+ (gated)

• Trigger: NY open (09:30 ET) sits ≥0.20% from the 00:00 ET midnight open (~26–27 ticks on 6B). Fires ~45% of sessions (n=111/244).
• Entry: with the overnight displacement; never fade it (the NQ midnight-magnet inverts in FX).
• Stop: just beyond the midnight open — that level held 85.6% (n=111; 0.2–0.39% bucket 81%, ≥0.4% 96%+, monotone dose-response). ~$175–200 full 6B / ~$18–20 M6B.
• Target: unquantified — the report family has no MFE/extension data; 85.6% is stop-survival, not win rate.
• Verified: H1 87.0/H2 84.2; symmetric (above 85.0/below 86.3); cluster 6E 86.1, 6A 81.8, 6C 90.5, spot GBPUSD 81% at same threshold; NQ/ES 46.9/50.4% on identical metric. Costs benign (1 tick + ~$2 RT).
• What it adds: uncorrelated asset class, defined micro-risk, works at Topstep (6B/M6B) or FTMO (spot).
• Gate: pair with a session-extension/MFE study before any standalone entries; until then use purely as stop-placement + do-not-fade rule.

4. Amendments to the existing playbooks

1. NQ PD-Magnet long: upgrade C→B, with a haircut and a mirror. The join reproduced to the decimal (72.0%, n=125) and ES/YM confirm (+14–16 lifts, all 12 half-cells positive). Amendments: (a) book forward expectation at the H2 figure ~67% (long leg is bull-regime-heavy; honest clean-win floor 60.8% counting both-break days as losses); (b) add the RTY short leg — 71.2% (n=104), H1 70.7/H2 71.7, the most regime-stable index expression, and it's the counter-trend side; (c) do NOT migrate to ags/FX despite bigger paper lifts — ZW's autopsy (proximity artifact, inverted R:R, overnight window) applies across that cluster. Trade it on NQ/MNQ + RTY where the cost profile is best in the screen.
2. Big-First-Candle day-hold: keep gated, expectations trimmed. It's a 36-symbol class effect that's mostly mechanical; strict after-entry NQ edge is thin. The backtest gate stands; if it fails, the spot-FX-London strict version (GBPUSD/AUDUSD, +10.6/+11.1, >50% abs) is the only follow-up lead, FX-prop only.
3. London AR-Break (Lucid): keep on NQ/ES only. The full sweep says its home is NOT FX or gold (dead-to-weak); do not extend it there. The yen (~70–73%, with-trend) and ZC corn (73%, flat halves) cells are unverified leads, not trades. Remember the ~52% mechanical floor when reading any AR number.
4. Kill-list additions (all verified or cluster-confirmed): grain ORB/hold-to-close logic (break-and-fade); FX London 15-min ORB (chop 72–78%); anchoring any crypto template to the UTC-midnight open (both-sides-broken 40–45% vs NY 72–74% single-break); FX weekend-gap fades; crypto PDR close-colour "edges"; NG Thursday/CL Wednesday EIA plays; fading FX away from its midnight open (that's now the opposite of the 6B rule); early-London metals first-break trust (GC double-break 61.4%, n=254 — regime filter only); all micros except MBT as data sources (duplicates).
5. Free filter upgrades (no new trades): ZB/ZN behave like index for IB-direction templates (single-break 73–75%) but with noise-level $ geometry — never trade rates IB; opening-candle colour is a legitimate secondary confluence filter on any existing index/commodity entry (69–75% everywhere) but never standalone.

5. Final portfolio

Bookable today: still only ~+$36/mo (strategy 1). The sweep's real yield is one genuinely better instrument for an existing effect family (RTY), one validated uncorrelated overlay (6B), a hard upgrade of the PD-magnet's evidence quality, and a much longer kill-list that prevents bleeding edge-chasing into crypto/grains/FX-London.

Unswept — exactly: - Parameter grids: ORB sizes other than 15-min; IB periods other than 60-min; opening-candle sizes other than the 60-min API default (5/15/30-min exist and differ); breakout_criteria=close vs wick; alternate session windows beyond those tested. - Stocks/ETFs: market_type=stocks never censused — entirely unswept. - Unsupported on the feed: full-size BTC (no data), BNB (404 all formats), MET (87 days, discarded). No SOL/XRP futures route exists at CME micro level. - Known data holes: USDCHF ORB-family intraday starts 2026-03-02 (other families full); ZF/ZN opening-stats-by-weekday persistent 500; XRPUSDT intraday H2 gaps (339/361); PDR by-levels coverage decays after 2026-04; crypto by-levels ignores intraday windows. - Structural limits: the API caps history at 12 months — everything above is one regime, H1/H2 is the only out-of-sample check performed; no news-day (CPI/FOMC) conditioning, no month-of-year seasonality, no Asia-session futures windows, and no report anywhere exposes open-to-level distances or MFE distributions (the single biggest gap between these hit-rates and real expectancies).

The DanTrades IB25/IB75 setup — 60-min IB (09:30–10:30 ET); trade when the IB closes in its top quartile (75–100) or bottom quartile (0–25) of its range. Investigated 17 Jul 2026: ending-zone semantics resolved with per-day cross-report joins, stats computed on NQ/ES/YM/RTY, then adversarially re-verified and priced. Bottom line: the direction signal is real (88–91% continuation), the fade version is a dead circular artifact, and the standalone trade prices to ≈ nothing after costs — use it as a 10:30 bias filter, not an entry system.

ADVERSARIAL VERIFICATION + PRICING: DanTrades IB25/IB75

1. INDEPENDENT REPRODUCTION — PASSES

Fresh API pulls (not the saved files) + independently written classification code (independent re-verification script + raw JSON pulls, archived locally):

• NQ IB75: n=90, cont-first 87.8% (H1 86.0 / H2 89.4); NQ IB25: n=62, 88.7% (88.9/88.6) — exact match.
• RTY IB75: n=93, 88.2% (90.9/85.7); RTY IB25: n=78, 88.5% (87.2/89.7) — exact match.
• Join integrity: by-rejection firstBreak vs by-color first break side: 0 mismatches, both tickers, fresh pulls.
• Colour composition: NQ IB25 = 59R/3G, IB75 = 86G/4R (RTY 74R/4G, 90G/3R) — as absolute-position semantics requires.
• Geometry and fade tables reproduce exactly (note: maxBearishLevel is NEGATIVE in the API; the stats agent handled this correctly — my first pass didn't, verified after fixing).

Circularity residue: none found. Both predictors are 09:30–10:30 objects; no-break days exist INSIDE the quartile classes (NQ IB25 has 2, RTY IB75 has 2, IB25 has 4), so the class is not conditioned on a break occurring; the edge is continuation (colour-edge direction, sharpened), not the previously-killed fade shape. Residual 4%: without raw IB OHLC we cannot literally verify Edgeful stamps ibEndZone at 10:30 — the 95-96% colour agreement is strong indirect evidence, not proof.

2. MARGINAL VALUE TEST — THE DIRECTION GAIN DOES NOT REACH THE P&L

Signal-level: quartile beats colour by +12-19 pts of first-break-side accuracy (NQ 88.7 vs red 69.2; RTY 88.5 vs 74.3), costing ~40% of triggers. Real.

But the trade construction already harvests direction. Entry is a stop order at the IB extreme: wrong-side-first single-break days are simply NO FILLS (free) under either filter. So the quartile filter mostly converts would-be no-fills into skipped days. Fill rate: NQ green long 112/138 (81%) vs IB75 80/90 (89%). Identical EV model applied to both filters (full-IB stop, +0.5×IB target, EOD scratch exits priced via by-close; EV in IB-units = R; base counts target&double-break days as wins, adversarial as losses):

Per-fill EV is statistically indistinguishable from the colour filter's (differences of ±0.01–0.02×IB at n≈60–110 are noise). Quartile = same edge, sharper lens, fewer trades. Its only per-fill improvements are a lower double-break drag (NQ 13.9% vs green 19.0%) and skipped dead days — worth ~+0.01×IB.

Midpoint-stop variant (new data: initial-balance-breakout-by-retracement has per-day post-break retracement depth — stop at 50% retracement is DIRECTLY testable except for ordering vs target). Bounds (pess = ambiguous target-AND-50%-retrace days counted as losses; opt = wins):

• NQ IB75: +0.03R to +0.13R (4 ambiguous days); NQ IB25: −0.18R to +0.22R (11 ambiguous)
• RTY IB75: −0.10R to +0.14R; IB25: −0.12R to +0.17R
• Colour equivalents: NQ RED −0.11R to +0.21R; RTY RED −0.05R to +0.21R — again indistinguishable.

Every cell's bounds straddle or graze zero; the 4-13 ambiguous days per cell are decisive and Edgeful cannot order them. +0.25×IB target: unpriceable (by-levels floor is 0.5×).

3. PRICE OF THE TRADE (assumptions: NQ ≈ 25,000, RTY ≈ 2,450 — scale linearly; median quartile-day IB: NQ 0.76–0.82% ≈ 190–205 pts, RTY 0.93% ≈ 23 pts)

1 MNQ ($2/pt), full-IB stop (+0.5×IB target): risk ≈ $380–410/trade; gross EV +0.02×IB ≈ 3.9 pts = $7.80; costs (≈$1.30 comm + 2–3 ticks slippage ≈ $1.0–1.5) ≈ $2.50 → net ≈ +$5/fill base, ≈ −$3 adversarial. ~11.3 fills/mo (both sides) → ≈ +$55/mo base, −$35/mo adversarial, per MNQ. Midpoint stop: risk ≈ $195; net per fill spans −$10 to +$18 (central ≈ +$5–7). Same monthly picture. 1 M2K: IB ≈ 23 pts = $115 risk; gross $3.65/fill, costs ≈ $2.80 → ≈ +$0.85/fill — pointless. Full RTY ($50/pt): ≈ +$18/fill base, negative adversarial, $1,150 risk. Full NQ ($20/pt): full-IB stop risks $3,900/trade — instantly incompatible with prop limits.

Topstep 50K fit ($1k DLL / $2k trailing): 1 MNQ passes risk limits under either stop (2 full-IB losers = $780 < DLL). But expected P&L of ~$50/mo base (bounds negative) means the $3k profit target takes 4+ years of base-case luck. Not a prop vehicle.

4. VERDICT

Signal: A−. Trade: C. Overall grade C+ (NQ), C (RTY); not an upgrade over the colour playbook.

• Does IB25/IB75 work as a stand-alone edge? As a direction call, yes — genuinely: 88–91% continuation, symmetric, all 4 tickers, both halves, non-circular, reproduced from fresh pulls. As a stand-alone TRADE, no: per-fill EV is +0.01 to +0.03×IB before the ambiguity haircut, bounds straddle zero, same as the colour edge that already priced at +0.09R/untradeable on NQ.
• Better/worse/same as IB-colour? Same edge, fewer trades. The +12–19 pt accuracy gain is absorbed by the stop-entry construction (wrong-side days were already free no-fills). Nothing in the quartile version changes the prior tradeability verdict on any ticker.
• Which ticker? NQ IB25 short is the best single cell (58%/36% reach 0.5×/1.0×IB, 92% resolve before noon, +0.022×IB) — but "best" here means least thin, not tradeable. RTY's colour version remains as good as its quartile version.
• Should Ashley trade this? No — not as a mechanical standalone; the honest per-trade expectancy on 1 MNQ is roughly a fiver with error bars wider than the edge. Its real value is as the best available 10:30 direction filter to overlay on a setup that has actual follow-through geometry (e.g. trade selection / bias / holding runners), not as an entry system.
• Fade version: dead, explicitly. Opposite side breaks first 3–12%, breaks ever 17–34%. The old 87–91% fade numbers are confirmed circular artifacts; nothing in the resolved data resurrects them.

What Edgeful cannot measure (bounds confidence): target-vs-stop ordering on days both levels trade (4–13 days/cell — decisive for the midpoint variant); MFE/MAE finer than 0.5×IB buckets (no +0.25× target pricing); retest/entry fill quality (assumed stop-market at the extreme); intraday path on EOD-scratch days (priced by close bucket only); and anything beyond one bull-year (12-month API cap — regime risk untested).