# Prime Academy: the settings, the rule, and what has to be measured before the first fee

Synthesised from three arena candidates (base: candidate-2; grafts from candidates 1 and 3; see
SYNTHESIS.md). Every number recomputed from `modules/prime.py` at the deployed stop's mean of
34.8 pips. Cost shares of the risk: 1.44% commission only, 2.87% with the 0.5-pip spread, 3.35%
with the random side's swap, 7.18% adding the measured 1.33 step-asides a trade at that spread.

At 3.35% (the random side) the cell reads pass 44.6%, any payout 51.2%,
cash 1.56 times the fee, breakeven honour 64%; with the step-asides (7.18%) 0.90 times the fee.
The paper's Section 1 and notebook section 25 carry the full gate table.

## 1. The settings

| setting | value | why |
|---|---|---|
| instrument, side | EURUSD, a random side each trade (a fresh coin seed each run) | direction is free under the fair game, so the side changes only the swap. The random side pays the average swap: 0.74 swap nights a trade, 0.17 pip, +0.48% of risk on the cost share |
| entry | first clear hourly bar 07:00-10:00 UTC, one trade a day | tightest spread, same-day resolution |
| stop, target | max(0.75 x yesterday's sigma, 30 pips), target the same distance (RR 1) | walk-forward winner; RR 1 makes every trade an exact coin, the only ratio the broker confirms fair |
| assessment risk | **3.0%** of the bankroll (was 4%) | a 3% win is 2.91% of the bankroll, under the 35%-of-target ceiling of 3.15%; a 4% win is 3.89% and breaks it if the firm applies the rule in the assessment, which its documents do not settle. Cost: 0.86 points of pass rate (45.38% against 46.24%) and about 9 days |
| funded risk | **1.5%** | the ridge: chance of any payout 52.4% against 29.4% at 4%, 1.10 payouts an account, first payout at a median of 18 trades; 1% pays as often but waits 30 trades |
| account rules in the strategy | target clipped to land on 9%, stop placed at the floor, `set_account` on | the floor is enforced on the tick (operator's position); unclipped, 3% is a lattice trap at 40.2% |
| tier, accounts | $3,000, fee $38, 70% split; **one account** | 3.8% of the $1,000 wallet, ruin 0.1%; the $10,000 tier is 9.9% of the wallet with 12% ruin; a second account waits for the first honoured payout |
| news | step aside on every touched hour | the only policy with zero breaches; each aside costs one round trip |
| payouts | request after 14 calendar days and 3 trading days, equity at least $143 gross, largest day at most 35% of the request | rules.json |

**Fallback if the target clip is not wired into the live runner:** assess at 2.5% (42.0% unclipped), never 3% (40.2%).

## 2. The rule

Maximise the funded account's chance of ever paying, on the exact chain with the floor on the
tick, the target clipped, the tier's own minimum request and the cost share at the mean deployed
stop including the measured step-asides. Admit only cells that comply with every rule under its
strict reading: one win at most 35% of the target in the assessment and of the minimum request
when funded; one loss, plus any overnight floating gain given back, at most 5% of the previous
day's equity; one trade a day. The assessment cell is the highest pass rate among compliant
cells, ties within a point going to fewer trades. The tier is the largest whose fee is at most 6%
of the wallet with simulated ruin at most 5%, held once until a payout is honoured. A broker cell
with fewer than ten blocks of twice its account span takes the chain's value; a resolvable cell
more than two block standard errors below its model is dropped. A purchase is made only when the
wallet's expected log growth is positive at 80% of payouts honoured, at the measured cost.

**One sentence:** pass as fast as the rulebook allows, get paid as often as the rulebook allows,
and buy only what the wallet can lose one time in twenty and the firm need honour no more than
four requests in five.

## 3. The finding that gates the purchase

The map ran with the news rule off. Stepping aside costs one round trip, and the cohort paid 1.33
asides a trade at the deployed stop. At the assumed 0.5-pip spread that puts the cost share at
6.7% of the risk, and the cell reads: pass 39.3%, any payout 41.8%, expected cash $36.81 against
the $38 fee, wallet log growth -0.29 with every payout honoured. Charged commission only (the
broker's own accounting) it reads $59.40, 1.56 times the fee, growth +0.12 at full honour and
-0.04 at 80%.

| cost share | what it assumes | cash per fee | breakeven P(paid) | wallet growth at P(paid) 1.0 / 0.8 |
|---|---|---|---|---|
| 2.87% | no asides, 0.5 pip | 1.67 | 60% | +0.167 / +0.006 |
| 3.4% | 1.33 asides, commission only | 1.56 | 64% | +0.121 / -0.041 |
| 4.5% | the expected-value gate | 1.33 | 75% | 0.000 / -0.167 |
| 6.7% | 1.33 asides, 0.5 pip on each | 0.97 | 103% | -0.290 / -0.477 |

So the gate is closed at the measured aside rate and the assumed spread. It opens only if the
asides are fewer or the spread is smaller than assumed. Both are measurable without risking an
account:

1. **Aside rate on a EUR+USD-only calendar.** A broker replay with the news rule on and the
   calendar narrowed. 33 of the 46 cohort breaches were USD releases, so the narrowing may not
   halve the count; measure, do not assume. Whether the firm's list is narrower than all
   currencies is itself unverified.
2. **The live spread at the aside instants.** From the soak's quote feed, over a few trading days,
   at the hours the asides happen. The RAW account is quoted "from 0.0 pips".
3. **The strict daily-loss breach rate.** From the cached replays and the bars: for each trade
   held across 21:00 UTC, the floating gain at the boundary against the loss the next day. At 3%
   a breach needs a gain over 0.64 of the stop; at 4%, 0.22.

## 4. What would change the choice

| if | then |
|---|---|
| the 35% rule is confirmed funded-only | assessment to 4%: +0.86 points, 12 days to pass instead of 21 |
| the clip is not wired live | assessment 2.5% |
| spread 0.25 to 1.0 pip | funded stays 1.5%; at 0.0 pip, 1% (57.0% against 55.4%) |
| aside rate A at 0.5 pip | cash per fee 1.37 at A 0.5, 1.12 at A 1.0, 0.97 at A 1.33 |
| floor on review rather than on the tick | assessment 4% if the cap allows; funded stays 1.5% |
| a payout refused | stop buying |
| wallet above about $1,650 with payouts honoured | $10,000 tier, funded 1% (61.7% chance of any payout) |
| firm objects to the runner (EA ban) or to two accounts | place by hand; one account |

## 5. The declared test (before any number is looked at)

`HistoricalBarBroker`, vendor hourly EURUSD 2023-11-24 to 2026-09-10, `PrimeAcademyStrategy`
with `set_account` on and the news rule on (stepping aside), the $3,000 tier, RR 1 at 3% and at
4%, 20 coin seeds each plus the short path, accounts cycled.

- **Statistics:** P3, the pooled pass rate at 3% with its block-clustered error; D = P3 - P4; A,
  asides per closed trade, on the all-currency and on a EUR+USD calendar.
- **Null:** the chain at the broker's own cost, 1.44% x (1 + A). At A 1.33: P3 44.6%, D -1.0.
- **Reject downward** if P3 is more than 2 block SE below its null (below 41.2% at A 1.33) or D
  below -5.2 points: the 3% choice fails and the assessment moves to RR 0.75 at 4% (win 3%, inside
  the cap, chain 45.2%). **Reject upward** if P3 is more than 2 SE above its null: the aside cost
  model is wrong and the gate is recomputed.
- **Power:** block SE about 1.7 points at 3%: 50% against a 3.4-point shortfall, 83% against 5,
  94% against 6; a 2-point shortfall is invisible. No funded-stage test is declared: the 1.5% cell
  has 4 blocks in 2.8 years and cannot be resolved by this history; its test is the live payout
  record.

## What none of this sees

The spread and the swap in the broker; the equity between completed trades; whether a record of
$100 payouts every few weeks reads as trading to the firm; whether the firm pays at all. The
investor's identity is the operator's prior on the last, not data.