# Decision ## 1. Settings and purchase decision **Test assessment RR 1 at 3% risk, then funded RR 1 at 1.5%; buy zero accounts now.** The candidate deployment is **one $3,000 account, $38 fee, 70% split**, no add-on, against the $1,000 wallet. Release that purchase only after the economic and execution gates below pass. This is a concrete conditional strategy, not a claim that the supplied evidence establishes a profitable compliant deployment. Risk is a fraction of nominated initial balance: $90 assessment and $45 funded, before costs. The assessment uses 3%, rather than today's 4%, because the unresolved 35% rule is assumed to apply to both stages: its conservative single-day ceiling is 35% × 9% = 3.15%. The mapped RR-1 grid's largest compliant risk is 3%. Aggregate daily gains, including news legs and swap, must also satisfy that ceiling. [rules.json; paper §§2,14] Use EURUSD, fixed short, first clear entry in **07:00–10:00 UTC**, at most one logical bracket daily and one position at a time. Stop **max(0.75 × yesterday's daily VolatilityKF sigma, 30 pips)**; filter rho **0.0025**, seed **30 days**. Floor lots to **0.01**, never round risk upward; check **30:1** margin against available equity. Short is an execution convention with full price risk, not a forecast or a swap-harvesting objective. No live EA/HFT, hedging across accounts/firms, intentional losing trades, or loss-driven size increases. Execute manually. [paper §§2,12; vault] Clip assessment profit to **+9% net of costs** and both stages' stops to the remaining **9% static equity floor**, using actual lots and all opening fees. Also restrict each day's remaining loss to **5% of previous-day equity**, including floating P&L, swap and re-entry costs; an overnight profitable position can lose more from the daily equity reference than from entry. Do not rely on completed-trade balance checks. Close before weekend gaps; do not submit a Friday-close order that fills Sunday. These controls require broker verification and change the published rates. [paper §§13,16] Around all high-impact releases, flatten at a clear instant before every touched hour, earlier if the boundary is within **±10 minutes**; rejoin only when clear, shift the remaining bracket as in §13, and recheck costs, daily budget and account barriers. Never hold naked through news. Require **3 trading days** each stage; do not allow **30 idle days**, and do not assume an open position counts as activity. Withdraw eligible profits after **14 calendar days from funding/previous request**, only when the net share is at least **$100** and the largest day's profit is at most **35%** of requested gross profit. The nominal funded hurdle is **$142.86 gross, 4.7619%**; maintain the actual daily-profit ledger. [rules.json; paper §§2,13,14; vault] At the mean **34.8-pip** stop, base cost is **2.8736% of risk**. Exact chains give assessment pass **45.38%**, funded any-payout **52.36%**, **1.099** payouts/account, lifetime gross **6.659%**, and first payout median **18 trades conditional on payout**. Cash probability/purchase is **23.76%** and expected lifetime receipts **$63.45**, assuming payment. These omit extra news costs and are not deployment forecasts. [C, defined in rationale] ## 2. Re-applicable optimization rule Use a constrained, conservative selection: keep RR 1, the existing stop/window and one account unless a prospectively declared, calendar-clustered broker comparison demonstrates a compliant alternative's advantage; among the paper's RR-1 assessment risks choose the largest at or below 4% whose gross win is at most 3.15%, and among its funded risks choose the highest exact-chain probability of any payout, breaking ties by shorter conditional first-payout duration. Use the smallest tier whose fee fits the paper's approximately 6%-of-wallet screen, and prioritize absolute wallet drawdown over growth or concurrency. Require positive expected receipts net of fees under measured compliant costs, positive wallet log growth and at most 5% modeled ruin over the paper's 500-trade comparison horizon; report wallet drawdown separately, never substitute account risk for it. Treat cells with fewer than ten calendar blocks of twice account span as unresolved, neither disproved nor validated; use their chains provisionally and give them no empirical superiority credit. Require the broker to implement daily floating equity, news costs, calendar payouts and gap handling before purchasing, with payment explicitly conditional rather than inferred from the investor's identity. If any economic or execution gate remains unestablished, the selected live allocation is zero, not a larger bet to recover the shortfall. ## 3. What changes the choice - **Assessment cap:** if confirmed funded-only, raise assessment **3% → 4%**, pass **45.38% → 46.24%**; broker conditional days-to-pass **21.4 → 12.0** are descriptive, commission-only. Funded setting stays. [C; broker_map.json DE] - **News/calendar/cost:** §13 measured **1.33 extra round trips/trade**. Applying **2.33 × base cost** as a sensitivity gives receipts **$36.81**, below $38, and payout probability **41.81%**. That closes the purchase gate. It is not a measured cost for this new cell. A confirmed narrower calendar or lower actual episode count may reopen it; retain 0.5-pip spread, do not assume free re-entry. [paper §13; C] - **Payment/fees:** no recurring fee is assumed. Base-cost EV requires payment above **59.88%**; the news sensitivity requires **103.24%**, impossible. Recurring charges reduce receipts further. An actual refusal stops replacement purchases. The operator's investor prior supplies no numerical payment estimate; wallet breakeven is stricter than EV. [C; paper §11.8] - **Clock:** assume funding starts the 14-calendar-day wait. Assessment-opening eligibility can advance first eligibility by up to **14 days**, but does not establish an equal cash-time improvement. Recompute before changing risk; current chains use 14 trades. [rules.json; paper §14] - **Floor, reset, gaps, fee timing:** retain tick enforcement and nominated funded balance regardless of the older overshoot results. If execution cannot enforce the daily/floor guards or calendar, live count remains **0**. Opening commission is **$5/lot**, not half deferred; account and gap effects need replay, not an invented numerical haircut. [paper §§14–16] - **Swap, lots, instrument, discretion:** keep short without counting credit; if short becomes the more expensive side, compare the opposite fixed side with measured costs before switching. Keep the confirmed **0.01** lot grid; skip infeasible positions. No instrument or concurrency increase without its own pace/correlation test. A stricter gambling/swap interpretation or unresolved inactivity/calendar reading keeps allocation **0**; many small payouts do not prove acceptability. [vault; paper §§9,16] ## 4. Prospective falsification Next broker test: **funded RR 1, 1.5% versus 4%**, identical nominated $3,000 starts, deployed stop/window, commission-only and 14-trade clock to match the chains. This isolates the funded-risk mechanism; it cannot authorize deployment without the separate compliance/cost gates. Statistic Δ is the difference in proportions ever reaching an eligible payout before death. Null: **Δ = 0.24709**, the commission-only chain difference in broker_map.json's agreement rows. Retain right-censored accounts and use matched chain censoring; never score unfinished paying accounts as failures. Cluster all seeds sharing calendar history together, using blocks at least twice the longer account span. Predeclare a two-sided **5%** test: reject below **−1.96** or above **+1.96** for Z=(Δ−0.24709)/SE. Target **SE ≤ 0.07071**, equivalent to at least **100 independent Bernoulli accounts per arm** at worst-case variance; then rejection thresholds are **Δ < 0.10850 or Δ > 0.38568**, with approximately **94% normal-design power** if the advantage vanishes (Δ=0) or doubles (Δ=0.49418). These are design constants, not historical findings. A lower rejection withdraws the funded-risk recommendation and purchase authorization pending reranking; an upper rejection also rejects the quantitative model and requires refitting before scaling. The existing low-risk cell has only **4 blocks**, below the **10-block** minimum: current history cannot deliver the declared power, and additional seeds cannot repair that. No replay was run here.