Live rung: strategy out-of-sample holdout, then separate paper performance, and the standing live-readiness instruction

[seathatflowsinourveins]

What

Adds two required live-rung gates to catalogs/us-equities/gates-20260922.json. Their criteria are fixed before any strategy qualifies. They were revised before merge after an independent review (four majors, five minors). strategy-historical-holdout is renamed and redesigned as strategy-out-of-sample-holdout.

strategy-out-of-sample-holdout

Established by one of two routes:

• (i) Historical holdout. The strategy's frozen protocol passes its own holdout rule, including that protocol's minimum trades and multiple-testing adjustment. Any lower bound is computed at the family-wise adjusted level.
  • For mover-early-entry v1 (protocol.json selection_and_pass_criteria.holdout), that means at most 5 rule-exits, at least 50 trades each, and a Holm-adjusted one-sided p ≤ 0.05.
  • The paper candidate is the one named by paper_e2e.
• (ii) Preregistered prospective paper holdout. A forward paper study passes its own preregistered test.
  • Its protocol, configs and code are pinned by sha256 before the first order.
  • It is analysed exactly once, at its preregistered analysis point.
  • For the shape, see incentive-monitor/forward-protocol-v1.json on claude/incentive-monitor-20260924.

Both routes need:

• at least 100 round trips and 20 sessions, or the protocol's larger minimums;
• returns net of costs from a cost table and a fee schedule, each pinned by sha256. The current instances are cost-table-run-v1.json be50cbdf… and fees.json fa1f468b…. Cost per side is the quoted half-spread plus the protocol's slippage rule;
• recorded market data or real Alpaca paper fills, pinned by sha256. Synthetic, offline or fixture evidence never counts.

strategy-paper-performance

This gate needs a separate, later paper period:

• the same strategy and the same frozen configs, pinned by sha256;
• starting after the out-of-sample gate's analysis point, so no session counts toward both gates.

Pass criteria:

• Analysis point: one, preregistered. It is the first session end with ≥ 20 sessions and ≥ 100 round trips. It is analysed once, with no repeated testing.
• Estimator: pooled mean net return per round trip. Fees come from the pinned fee schedule. Alpaca paper charges no commission, and the spread is in the fills.
• Lower bound: a one-sided 95% circular 5-session block bootstrap lower bound > 0.
• Drawdown: max drawdown of the strategy's paper equity ≤ 20%.
• Receipts: zero needs_attention.
• Slippage: adverse slippage is measured per side, in bps: entry (fill/model − 1) and exit (model/fill − 1). Its median must not exceed the pinned per-side cost assumption. With a bucketed table, the median of each fill's excess over its own bucket's cost must be ≤ 0.
  • The comparator is the strategy's own paper-vs-model script, named in its protocol.
  • For mover, that is paper_compare.py. Its exit_diff_bps uses the opposite sign, so the adverse exit value is recomputed rather than copied.
• Sessions: a session is a scheduled trading day of the strategy. Days without trades count.
• Leverage: leverage above 1x counts only at an established ladder rung.

Receipt and checker

Both gates share blueprints/us-equities/strategy-performance/receipt.json, read at /out_of_sample/status and /paper_performance/status.

The notes preregister that both sections carry the same strategy_id, protocol_sha256 and config_sha256. The checker verifies only the status strings, and a reviewer verifies the rest before a dated flip commit.

live-go note

Readiness is now stated as: the live lane is ready for the user's decision when the checker's blocking.live equals ["live-go"]. The user's 2026-09-24 instruction stays recorded:

• paper and simulation are not subject to human approval steps, and their rung gates still apply;
• the user stores the live key and gives the explicit go;
• automation never flips live-go or places live orders.

"Never gated" is removed.

Handbook

• The gate table is dated September 24, with the counts unchanged: 22 gates, 21 checked.
• The live row names the out-of-sample and paper-performance gates.
• The paper row now matches the checker: the paper rung has been ready since Establish adaptive-paper-broker-trial on a main-built 1x paper trial (paper rung ready) #128.

Tests

• tests/test_trading_gates.py adds both gates to the GOOD/BAD receipt fixtures.
• A new test checks that a receipt holding only the out_of_sample section flips only that gate.

Evidence (rebased on origin/main 5af3581, raw)

• python3 scripts/trading_gates.py --check: passed. 22 gates; rung_ready is sim true, paper true, live false. blocking.live lists the five open live gates, both strategy gates and live-go. There are no flip candidates.
• python3 scripts/evidence_manifest.py --check: passed (5,235 files).
• python3 scripts/validate.py: passed (5,235 hashed files, 143 receipts).
• python3 -m unittest tests.test_trading_gates: 19 tests OK.
• Full python3 -m unittest under ecosystem-bounded-run: Ran 3945 tests, OK (skipped=419).
• Guarded gitleaks git --log-opts=origin/main..HEAD --no-banner --redact: 1 commit scanned, no leaks found.

These are synthetic fixtures and local integration checks. No receipt exists at the shared path.

Status

No strategy qualifies today: mover v1 and v2 had no development pass (#162). The prospective route is not yet on main: the incentive-monitor study proposes it on its own branch, and no gate status follows from it until this PR merges.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

[seathatflowsinourveins]

Closing: this PR adds required live-rung gates and a standing live-readiness instruction. Live trading is out of scope for this repository; the owner's rule is paper lane only here, with live handled elsewhere. The branch is kept, so the gate criteria can be reopened or carried to the private trading repository. Part of the 2026-09-25 noise cleanup approved by the repository owner; the trading-lane session acknowledged.