← Accumulator -mech v7 v7

Graduation report

Walk-forward evidence for whether this version is ready to trade real money — computed from the stored ledger, equity snapshots and trades on every open. Computed, never enforced: a graduated version keeps trading on paper exactly as before.

Checked 2026-09-25 10:47 UTC · confirmations: 0 of 2 consecutive B-windows

Insufficient data: 3 of 13 ledger days populated

Walk-forward windows

No slot has been evaluated yet.

Evaluation paused — slot #0 is missing 7 of 10 ledger days (2026-09-26, 2026-09-27, 2026-09-28, 2026-09-29, 2026-09-30, 2026-10-01, 2026-10-02). Waiting for ledger day 2026-09-26.

Minimums

Evidence from the most recent evaluable slot.

GateRequirementResultEvidence
Fitness above cohortledger fitness > 0 and ranked on every day of both windows—no evaluable slot yet
Trade count≥ 10 executed trades across both windows—no evaluable slot yet
Drawdown capmax peak-to-trough drawdown < 20% within the slot—no evaluable slot yet
No single-trade dependencebest trade's |PnL| ≤ 50% of the B window's total realized PnL, total > 0—no evaluable slot yet

Decision sample

Up to 12 most recent judged turn moves — the human-auditable sample.

WhenAssetVerdictReasonJudge
9/25/2026, 6:28:04 AMETHholdpolicy: no triggerDimvest Judge
9/25/2026, 6:28:04 AMBTCholdpolicy: no triggerDimvest Judge
9/25/2026, 12:28:04 AMETHholdpolicy: no triggerDimvest Judge
9/25/2026, 12:28:04 AMBTCholdpolicy: no triggerDimvest Judge
9/24/2026, 6:27:56 PMETHholdpolicy: no triggerDimvest Judge
9/24/2026, 6:27:56 PMBTCholdpolicy: no triggerDimvest Judge
9/24/2026, 1:20:52 PMETHholdpolicy: no triggerDimvest Judge
9/24/2026, 7:19:47 AMETHholdpolicy: no triggerDimvest Judge
9/24/2026, 1:19:17 AMETHholdpolicy: no triggerDimvest Judge
9/24/2026, 1:19:17 AMBTCbuypolicy: ema_death_cross_15m (rate 1.00)Dimvest Judge
9/23/2026, 7:19:14 PMETHholdpolicy: no triggerDimvest Judge
9/23/2026, 7:19:14 PMBTCholdpolicy: no triggerDimvest Judge

Risks

  • Past paper performance is not evidence of future real returns — the gate measures a paper record, not the market you will trade in.
  • Trades recorded before slippage was introduced carry no slippage cost, while the gates treat every row uniformly — pre-slippage PnL is slightly flattered. Slippage applies going forward only.
  • Slippage is modeled as a fixed bps per fill (default 10), not market impact — real fills in thin books will move the price.
  • Paper fills include fees and slippage but not partial fills or funding.
  • Single-lineage evidence: one version's record in one market regime — not an ensemble result.
  • The gate's thresholds are v1 constants, pending review after the first generation of candidates exists.
LLM off — the engine is running mechanically