# Rationale: alternatives considered and rejected (candidate 2)

Sources as in decision.md: bracketed file names for read numbers, "chain" and "wallet_mc" for numbers I computed from `modules/prime.py` at the 34.8-pip mean deployed stop.

## Alternatives for the assessment

**RR 1 at 4% (the notebook's pick, section 23).** Best pass rate at RR 1, 46.2% against 45.4% at 3% (chain, 2.87% cost), 5.9 trades against 9.7, 12 days to pass through the broker. Rejected because a 4% win is 44% of the 9% target and rules.json lists as unverified whether the 35% cap applies in the assessment. The paper's Section 14.9 hatches this cell for that reason. The two errors are not symmetric: assuming strict when the rule is loose costs 0.8 points and about nine days, assuming loose when the rule is strict costs every pass. The operator's own words put "not flagged" ahead of speed.

**RR 0.75 at 4%.** Same compliance (win 3%), pass 45.2%, 7.5 trades, 15 days to pass [broker_map.json]. Nearly the equal of RR 1 at 3% and faster. Rejected as the primary because a 4% loss against a 3% win makes the funded and assessment brackets differ, adds a second ratio to freeze, and the broker reads this cell 1.9 SE under its model (43.8% against 47.6%) where RR 1 at 3% reads on it (z −0.03). Kept as the named fallback if the declared test rejects 3%.

**RR 1.25 at 2.5%.** Win 3.125%, 34.7% of target, on the edge of the cap. Pass 45.6%, 10.9 trades. Rejected for the thinner margin and the longer record for no gain.

**RR 3 at 3.5%, the map's best pass rate (48.4%, median 3 trades).** Rejected: one win is more than 35% of the target, and a three-trade pass is the record the operator wants not to show [paper §14.9].

**2.5% at RR 1.** Pass 44.6%, 13.6 trades, 28 days to pass. It is the right cell only if the target clip is not wired into the live runner, where it reads 42.0% unclipped against 40.2% for 3% (the lattice trap: three wins net of cost fall short of 9%) and 40.7% for 4%. Kept as the conditional fallback, not the choice.

**Lower assessment risk for a "trading" record (2%).** Pass 43.4%, 21 trades, 50 days. Rejected: the pass rate falls with the trade count at any positive cost [paper Corollary 3], the wallet waits twice as long for the same coin, and the funded stage is where the record is judged at payout time.

## Alternatives for the funded stage

**4% funded (the paper's published cell).** Gross withdrawn 7.2% against 6.7% at 1.5%, but P(any payout) 29.4% against 52.4% and 0.42 payouts per account against 1.10 (chain). Wallet log growth +0.357 against +0.203 at 4%/1.5% (wallet_mc), with ruin 5.3% and a $581 median drawdown against $304. Rejected: the operator's objective is payout frequency and shallow drawdown, and 4% loses on both while sitting on the ruin cap.

**1% funded.** P(any) 52.2%, the same as 1.5%, but the $143 minimum binds, the first payout waits a median 30 trades against 18, and mean life is 106 trades: the account pays the cost on twice the round trips for the same chance. Rejected on time to first cash.

**1.25% and 1.75% funded.** 52.1% with first payout at 23 trades, and 49.6% with 16 trades. 1.5% is the ridge point where the 35% rule and the $100 minimum meet [paper §14.4], and both neighbours lose on one axis.

**RR 3 at 0.5%, the grid's best P(any) (55.7%).** First payout at 37 trades, trades 3.5 days each through the broker, life 150 trades. Rejected: 3 points of payout chance for twice the wait, at a ratio the broker reads about a point under fair on this history [paper Table 28].

**RR 0.75 at 2%.** 51.5%, 1.06 payouts, first payout 22 trades. Dominated by RR 1 at 1.5%.

## Alternatives for tier and count

**$10,000 tier.** Cash per fee 2.19 at P(paid)=1 against 1.67 at $3,000 (chain), breakeven P(paid) 46% against 60%. Rejected by the wallet: fee 9.9% of $1,000, ruin 12.1%, median drawdown $693 (wallet_mc). The expected-value ranking inverts at a finite wallet [paper §7, vault 2026-09-09].

**Two accounts from day one (paper Corollary 9).** Ruin 2.7% and log growth +1.21 against +0.79 for one, measured at the paper's published cell (4% both stages, floor at review) [paper Table 20]. Rejected as the opening move: both fees are bets on an unmeasured P(paid), the second account in one name at payout time is a risk the paper says it does not price, and under the news-rule cost both accounts bleed. Adopted as the second step once a payout lands.

**Five independent accounts.** Ruin 5.0%, on the cap, and the independence is not available: same side, same pair, at most two well-separated entry hours measured [paper §11.6].

**90% add-on.** Rejected by Proposition 11: the fee rises 1.35 times and the take 1.29 times [paper §6, rules.json].

## Alternatives for the trade itself

**60-pip fixed stop.** Lower cost share (1.67%) and fewer asides per trade at a much slower pace, 3.2 to 4.7 days a trade. The walk-forward picks 0.75 sigma on both halves and both policies [paper Table 23]. Not re-opened.

**Going dark through news windows.** 33 breaches of the daily cap in 535 episodes, 26% of accounts lost [paper Table 24]. Rejected. The stepping-aside cost is the price of the play here, and it is why the purchase is gated.

**Random or long side.** Long pays $6.24 a night, random averages a debit. Short is credited [paper §9.2]. The side is free under the fair-game assumption [Proposition 17].

**Entry 16:00 to 19:00 UTC for the single account.** Collects more swap nights but holds through the US afternoon and the Asian releases, with the aside count unmeasured. Kept for the second account only, where the stagger is what buys the correlation of 0.26 [paper §11.6].

## Alternatives for the rule

**Maximise wallet log growth (the paper's objective).** Picks 4%/4% at this wallet (+0.357). Rejected because it is what the operator asked not to optimise for, and because its level is a comparison between cells at P(paid)=1, not a forecast.

**Maximise expected cash per fee (Table 27's 4%/4%, 1.70).** Rejected: expected value is blind to cash frequency, which is what ruins a wallet [paper §10.2].

**Treat unresolvable broker cells as failures.** Would drop every funded cell below 3% risk, including the ridge, and leave the assessment as the only stage with a checkable number. Rejected: the chains agree with the simulations to 1.9 SE on 16 cells [paper Table 26] and with the broker in 86% of the 73 resolvable cells, so the chain carries the unresolvable cells with the margin stated in the rule.

## What I did not consider

Any cell that assumes a forecast. Any instrument other than EURUSD, since Section 9's leader has had none of the checks. Any reading of the rules more generous than the documents support.