Grading Standards

Every public claim Helios makes is graded by machine against rules committed in advance. Calls are timestamped at issue, graded nightly, and published with wins and misses alike. Nothing is edited, nothing is deleted. These standards were written before the results existed; changes to them are dated and disclosed.

What gets graded, and how

• The morning signal — issued premarket with an entry-time price anchor; graded at the close against that anchor. A close beyond the anchor in the called direction is a win; within ±0.05% is a push, excluded from win rates. If issuance is ever delayed, the signal is graded from its actual capture price — anchors are never backdated.
• The Day Card — one call per symbol, fixed at 10:00 ET, unable to change intraday; graded at the close against its 10:00 price under the same push rule.
• Re-entry slots (10:30, 12:00, 13:30, 15:00 ET) — each graded from its own slot price. Anchor time is part of the signal.
• The traded mirror — published signals are additionally executed in a broker-connected paper account under fixed rules with broker timestamps. Sizing: ten contracts per signal since September 3, 2026; one contract before — sizing changes are dated here. Execution basis: since September 9, 2026 entries follow the committed day call and are held on day-scale exits (call flip, thesis stop, same-day flatten — no overnight positions); from August 25 to September 9, 2026 the account mirrored the engine’s own trade lifecycle. Basis changes are dated here. Amended September 15, 2026: one entry per symbol per day; deep in-the-money contracts (delta ≈ 0.9) so results track the graded call rather than amplify it; size reduces mechanically from ten contracts to three while the Day Card’s trailing-10 record is below 50%. The September 8–14 drawdown traced primarily to executor design (repeated re-entries, at-the-money convexity), not signal failure, and is corrected as of this date. One-day discretionary override, September 15, 2026: by founder decision, entries and exits mirrored the engine lifecycle at fifteen contracts for this session only. Dated here like every basis change. P&L is published as recorded. Paper results are not real-money returns and are labeled as such.
• Retired: the dwelled-wave cell (September 14, 2026) — the backtested 30-minute wave-continuation cell (87% in the June–August study) failed its pre-registered forward court: 29.9% across 335 forward-graded emissions against a 75% bar. Per the rules above, it is retired from measured claims; the wave lights remain as witness displays with no percentage. Backtests are hypotheses. The forward court is the record.
• Platform probation (September 15 – October 13, 2026) — following the September drawdown, the platform itself now stands a pre-registered trial: the morning signal must hold at or above 58%; the Day Card receives a regime gate and must stop its losses in reversal tape or retire at its published 55% bar; the simulated account under the amended execution basis must finish the window at break-even or better; and the business must add one paying member or a concrete institutional engagement. If the platform fails its own court, it winds down with this record published to the last day. If it passes, September was the origin story. Judged here, like everything else we publish.
• Display change: live structure state, executed-signal strip, exact exits (September 23, 2026) — on days with no committed call, the card headline now reports the measured structure state (for example, trading below the zero-gamma flip line), labeled explicitly as a state and not a graded direction call; every card carries the mirror’s executed entries and exits for the day exactly as filled; and playbook lines print entry, target, stop, and time stop. No new accuracy claims accompany these changes.
• Retired: the 1-Hour Engine (September 24, 2026) — the intraday hour-scale engine failed its trial: 50.3% across 1,088 graded signals, with no conditioning (day-call alignment, conviction band, symbol, hour of day, flip filtering) clearing the pre-registration bar. It is removed from the card. Its capture continues silently, and the hour horizon remains unstaffed until a successor passes a pre-registered forward court.
• Published and benched: the Qualified Signal gate (September 24, 2026) — the intraday entry gate (backtested 57–62% in August) recorded 47.2% across 161 forward-scored fires and is benched. Its Telegram entry alerts have been paused since August 18, 2026. The capture continues, and any return requires a new pre-registered forward court.
• Products in trial — graded on the same discipline but held off member surfaces until their forward record clears pre-set gates.

Statistical standards

Win rates are always published with sample size. Exact binomial p-values against the 50% coin are computed in our periodic audits. Young records are called young. No subset of results is ever presented as the record.

Pre-registered retirement rules

For each headline product we wrote down, in advance, the conditions under which we retire it: the morning signal is demoted if, at 30 day-level calls, its p-value is at or above 0.05 and its win rate is below 55%. The Day Card faces the same test at 60 graded calls. If a product fails its test, the retirement is published in the same channels its wins were.

Voids and incidents

When events outside the market break a grading window — an outage, a data failure — the affected entries are voided and disclosed, never re-graded against a different window. On the record: the August 31, 2026 provider outage voided two overnight entries rather than grade a four-night hold as a one-night call.

Corrections

Grading code changes apply forward. Any retroactive recomputation is disclosed. The record page states its current scoring basis.

What is not claimed

No performance promises. Educational only; not financial advice. The construction of the signals is proprietary and is not disclosed — grading is fully transparent, methodology is not. That trade is deliberate, and it is the whole point: you do not have to trust how we decide, because you can verify how we score.

Verification

The public record is served from the same database that grades it. Receipts post automatically at 16:10 ET to public channels. Historical entries remain retrievable. The live record: helios.markets/track-record.