Mistakes

Closed losers

No closed losers in window. Either we got lucky or we didn't trade enough.

Retired rules

When a HOLD decision is triggered by an explicit, named rule gate (e.g., Rule 41 stabilization criteria unmet, Rule 27 macro cap reached, earnings binary event proximity per Rule 39), the decision log must include the triggering rule as a signal entry in the format "rule_gate:[rule_number]:[criterio…
retired 2026-06-08T19:06:12.522448+00:00 — Zero binds in 40 days (added 2026-05-11). The June 5 narrowing created an OFF-PLAYBOOK exemption so broad it eliminated all binding surfaces. Today's June 8 HOLDs (GOOGL, AMZN, ANET) were not labeled OFF-PLAYBOOK yet sti…
When a symbol accumulates 6 or more round-trip pairs (opposing BUY/SELL actions within 7 days of each other) within a 30-day window, flag the symbol for a "churn review" before executing any further BUY trades in that symbol. The churn review must answer: (1) Is the position thesis still intact, or …
retired 2026-05-28T15:22:28.309094+00:00 — Auto-lift condition met: GOOGL 30-day round-trip pairs fell to 3 (<6 threshold). Flagged by reflection memos 299/301/305 without action; retired manually per supervisor memo 1 (2026-05-28).
When attribution data shows the "thesis" signal has a beat-market rate below 55% over 20+ samples, treat thesis-only decisions (signals=[thesis]) as requiring at least one corroborating signal from a separate category (news_signal, macro_signal, technical_signal) before execution. A thesis-only deci…
retired 2026-05-21T19:06:27.028986+00:00 — Thesis signal has continued strengthening: now 64% beat-market rate at n=25, up from 48% (n=29) when the rule was first triggered and from 59% (n=27) at last update. The corroboration constraint itself remains valid and …
When the trailing 30-day attribution data shows any news sub-signal category (news_signal:analyst, news_signal:earnings, news_signal:product, or similar) with a beat-market rate below 35% at n≥5 samples and average 7d alpha below -1.5%, demote that sub-signal to "supplementary context only" status. …
retired 2026-05-20T19:06:10.198396+00:00 — Superseded by Rule 47, which carries updated attribution evidence from session 2026-05-20. All three demoted sub-signals have worsened since Rule 46 was last updated (analyst: 24%→18%, earnings: 25%→18%, product: 23%→14%…
When attribution data shows the "thesis" signal has a beat-market rate below 55% over 20+ samples, treat thesis-only decisions (signals=[thesis]) as requiring at least one corroborating signal from a separate category (news_signal, macro_signal, technical_signal) before execution. A thesis-only deci…
retired 2026-05-12T19:11:01.194095+00:00 — Superseded by Rule 45, which carries the same thesis-corroboration constraint but with updated attribution evidence (thesis signal improved from 48% to 59% beat-market rate, n=27, as of 2026-05-12). Rule 45 also reflects…
When a BUY or SELL decision is executed with signals=[] (empty signal log), the decision must be automatically downgraded to HOLD-review and returned to the decision loop for corroboration. An executor must not proceed to order submission with an empty signal log. If no corroborating signal can be a…
retired 2026-05-12T17:08:44.087721+00:00 — Replaced by audit-observable signal_gap marker (v2/trader.py emits a signal_gap decision_signals row whenever a buy/sell has no concrete signal_refs). Enforcing the original "must not proceed to order submission" gate wo…
When logging signals for a decision, deduplicate within the same signal sub-type: count each distinct sub-type (e.g., news_signal:analyst, news_signal:product) at most once per decision log for attribution scoring purposes. If multiple instances of the same sub-type exist (e.g., five separate analys…
retired 2026-05-12T17:00:08.724870+00:00 — Signal deduplication is enforced at the storage layer via decision_signals primary key (decision_id, signal_type, signal_id) with ON CONFLICT DO NOTHING in insert_decision_signals_batch. The rule restated a database inva…
Before submitting any SELL order, validate that the target position exists in the current portfolio state with sufficient available shares. If a prior-session exit was already executed for a position, do not re-attempt the same exit in a subsequent session. Cross-reference pending/executed orders ag…
retired 2026-05-12T16:56:43.857231+00:00 — Hard-coded pre-SELL validation exists in v2/executor.py:submit_order (position-existence check) and v2/risk.py. The rule restates a code invariant rather than gating decision logic; zero citations in 30d confirms it is d…
Treat thesis:buy as a degraded signal requiring corroboration. Do not act on thesis:buy alone — require at least one additional signal (news_signal, technical confirmation, or sector catalyst) before executing a buy.
retired 2026-05-02T07:53:54.114925+00:00 — Built on broken signal-citation pipeline. The executor structurally could not cite news_signal IDs (the executor context never surfaced them and validate_signal_refs stripped every fabricated citation since 2026-03-22). …
Before executing any SELL order, require at least one attributed, corroborating signal (news_signal, technical breakdown confirmation, sector deterioration catalyst, or stop-loss trigger) beyond a bare thesis or empty signal log. Do not exit a position based solely on unlogged or unattributed signal…
retired 2026-05-02T07:53:54.114925+00:00 — Built on broken signal-citation pipeline. The executor structurally could not cite news_signal IDs (the executor context never surfaced them and validate_signal_refs stripped every fabricated citation since 2026-03-22). …