The Warning study measured a fitted percentile crossing that nothing consumes.
What reaches Telegram is a quadrant change: fixed 50/40 dividers, hysteresis,
two-session confirmation, 3-day cooldown. Those thresholds are constants, not
fits, so there is no training set to protect and all 11 detected corrections are
evaluable instead of the 4 that fell in a holdout.
Replaying it: 1/10 corrections, 0.9 false alarms/year. Random alarms at the same
firing rate match or beat that in 65% of draws. The panel now carries ablations
(does the quadrant machinery earn its place?), external baselines (does the score
earn its complexity?), and that null, because a bare "2 of 4" was unreadable in
either direction. Nothing in the alert path was retuned on the strength of it.
Fundamentals become a third channel rather than a term in either score. v3 cut
them arguing 12+8 of 100 points "could not change any published conclusion" --
true only when every technical sensor reads zero; weighted they moved the bar for
the 40 divider from 40 to 25. But no fusion weight is measurable either: with ~10
events and no fundamental history, any weight is a policy preference presented as
a measurement. So the read is a categorical state (supportive/neutral/adverse/
unknown) with an evidence grade, derived by fixed rules from stored facts, read
by confluence. The LLM extracts and explains; it does not score.
Absence stays absence throughout. `unknown` is unreachable by averaging, a stale
or empty observation may display but never confirm, extraction failures map to
`unknown` rather than `mixed`, and the study rows are coverage-matched and marked
not-measurable until enough corrections are covered -- otherwise a fortnight of
observations renders as 0/10 and reads as a failed test.
Observations become a real time series (migration 033); they lived in a single
overwritten settings slot, so no history existed to replay. Pre-rename snapshots
are adapted rather than discarded. METHODOLOGY stays v4 -- no score changed --
so no reseed; STUDY_SCHEMA moves to 3 and discards the cached report.
Post-deploy: re-run Event Study from Admin -> Jobs. The panel reads "not run yet"
until then.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
"Market Regime" and "Regime Monitor" sat next to each other in Admin -> Jobs
(pipeline steps 4 and 5) reading as the same job. They are unrelated, and the
names had it backwards: "Market Regime" is the SPY 50/200 guard that drives the
TopBar trend dot and the counter-trend warning on setups, so it changes what a
setup shows; "Regime Monitor" is the observational AI/Tech thermometer that
explicitly feeds no trades. The more consequential job had the vaguer name.
market_regime "Market Regime" -> "Market Trend (SPY)"
regime_monitor "Regime Monitor" -> "AI/Tech Risk Monitor"
Display strings only. The job *ids* are persisted -- they key the pipeline step
list, cron config, runtime tracking and run history -- so they are untouched,
as is the /regime route, which keeps existing links working.
The label the admin UI renders comes from JOB_LABELS in admin_service (via
routers/jobs.py), not from the scheduler's APScheduler `name=`. Both are updated;
only the former is user-visible.
Carries the vocabulary through the rest of the surface so it does not half-land:
page title, nav ("Regime" -> "Risk"), the empty-state instruction that names the
job to run, the quadrant alert toggle, the morning-pipeline hint, and the
Telegram alert headline ("Regime quadrant change" -> "AI/Tech risk quadrant
change"). No test asserts any of these strings.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The LLM-sourced capex/earnings observations carried 12+8 of 100 Warning points,
so both pegged at 100 produced a Warning of 20.0 -- below the event study's 25.3
alarm threshold and still inside the "stable" band. The reading was
arithmetically incapable of changing anything on screen, which is why refreshing
it appeared to do nothing. They are now a qualitative overlay reported beside
the scores rather than diluted into them.
Calibrated against the 408 v2 sessions to 2026-07-24, reproduced offline from
Alpaca + FRED; the harness matched the stored prod distribution exactly before
any parameter was changed.
State:
- P3 used dd_pct * 5, reaching 100 at a 20% drawdown -- the 90th percentile of
the observed distribution -- so 39/408 sessions sat at exactly 100 with no
resolution left during the part of a selloff that matters most. Replaced with
anchored breakpoints keeping headroom past the observed 36% maximum, blended
2:1 like P1/P2 instead of max(). P3's realized share of State falls from 65%
to 40%, matching its nominal weight.
- Credit level is now anchors-only. ICE capped FRED's BAMLH0A0HYM2 at a rolling
3-year window in April 2026, silently turning the 10-year percentile leg into
a 3-year one that scored 20 points of stress at an OAS of 3.5 -- the level its
own anchors call "mild". The anchors already encode the long-run distribution.
Warning:
- Added HY OAS 20-session widening (25%). The level is pinned at zero below the
3.5 anchor; its rate of change is not.
- Divergence tapers to a 0.35 floor instead of a hard price_ret >= 0 gate, which
zeroed the sensor through every decline: on 2026-07-24 the basket shed 10
points of participation in 20 sessions and Warning printed exactly 0.
- The event study and the live monitor now share one sensor definition, so they
cannot silently drift apart.
Bands are per axis (State 20/50/80, Warning 20/40/60) with quadrant dividers at
50/40; v2 Warning never exceeded 64.9 against a shared 60, leaving that half of
the quadrant unreachable. Realized shares: State 73/15/8/3%, Warning 69/20/8/3%.
Snapshots now record credit_history_days and vix_history_days -- the percentile
defect went unnoticed for months because nothing asserted the window the code
claimed.
Cutover: the first run rebuilds 400 sessions automatically; the Event Study job
must be re-run, as its cached report self-invalidates on the methodology check.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Review of the shadow book found seven ways the two books could leak into
each other; all are fixed here. The most serious silently invalidated the
comparison the shadow book exists to make.
- Shadow holdings no longer suppress the manual candidate list. The
open-trade exclusion filtered on any book, so shadow taking the
top-ranked names removed exactly those from the user's list and alerts,
confining the discretionary book to leftovers. Scoped to the manual
book. Closed-trade alerts and paper-book equity were leaking the same
way and are likewise scoped.
- Shadow sizing now matches _simulate_portfolio: min(1% risk, 20% notional
cap, available cash) from marked equity, plus the sub- dust guard.
Previously risk-only from realized equity, so a tight stop produced a
multiples-of-equity leveraged position the strategy would never take.
- Shadow only trades setups from the scan that just ran (<6h old) with one
setup per ticker. A failed or disabled scan step could otherwise open
positions from a prior session at stale prices.
- Gate-reset transitions are observed for both books, so a shadow stop-out
completes fail -> requalify instead of staying locked forever.
- Manual list/close endpoints default to the manual book and reject
hand-closing shadow trades; the performance endpoint is scoped to the
caller so 'your picks' is not every user's book.
- run_shadow_book is registered as a paused job so Admin can trigger it.
Also anchors three pre-existing paper-trade tests (and the new alpaca
window test) on the UTC date. They build fixtures from the local date but
the service stamps opened_at in UTC, so they failed only between 00:00 and
02:00 in a UTC+hh timezone -- latent on ba2df8b, exposed by the clock.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Stored TradeSetup rows are point-in-time snapshots from the RR scan, so
the ticker page could show stale confidence/reasoning/composite (e.g.
sentiment=neutral in the setup card while the sentiment panel showed
bullish). Overlay current score/sentiment context onto the API payload
for GET /trades and GET /trades/{symbol}, gate and format Telegram
qualified-setup alerts on the same live values, and apply the
min_confidence/recommended_action filters after the overlay so they
judge what the caller actually sees. Stored setups stay frozen for
outcome analysis and backtests.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Applies the backtest-validated trailing stop to live paper trading, and surfaces
it transparently.
Exit (A):
- New paper-trade exit policy (paper_exit_mode=trailing, paper_trailing_pct=12),
tunable in Admin → Paper-Trade Exit. resolve_open_trades runs a trailing stop
(initial stop as floor, ratchets up from the peak; target ignored — the
validated rule) and records close_reason (trailing|stop|target|manual; +migration
013).
- list_trades enriches open trades with the live trailing-stop level + distance %.
Open Trades panel shows the active tactic and a Trail Stop column.
Alerts (B):
- Daily digest now lists open trades with unrealized gain, trailing stop, and how
far away it is.
- New "trade closed" alert: one summary per auto-close (trailing/target/stop, not
manual) — direction, reason, days held, P&L abs+%/R — covering wins AND
stop-loss losses. Deduped by trade id; toggle in Admin alerts.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fires once when the regime monitor shifts quadrant (regime index x early
warning), so you don't have to watch the tab. Two guards against spam:
- Hysteresis: each axis only flips once the value crosses its divider by a
margin, so a point parked on a boundary keeps its quadrant instead of
flip-flopping day to day.
- Cooldown: a genuine change stays quiet for a few days after the last alert.
Seeds the baseline silently on first run; reuses the existing Telegram dispatch
+ AlertLog. New per-trigger toggle in Admin → Alerts (on by default).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Behavior-preserving cleanup (345 tests pass, ruff clean):
- scheduler: replace 62 inline logger.x(json.dumps({...})) calls with a
_log_event helper, and collapse 11 identical _job_runtime dicts into an
_idle_runtime() factory over _JOB_NAMES.
- settings: add app/services/settings_store.py (get_setting/get_value/get_map/
upsert_setting) and route ~13 hand-rolled SystemSetting queries + two
identical _settings_map helpers through it.
- scoring.get_rankings: collapse the per-ticker N+1 (3-4 queries + a commit each)
into 2 bulk reads + a single conditional commit; drop the redundant re-fetch.
Lazy recompute-on-read is preserved. Adds first tests for get_rankings.
Net ~ -245 lines across the touched modules.
Qualified tickers already get their own "qualified setup" alert, so an S/R
proximity ping on them is redundant noise. Drop the watchlist ∪ qualified scope
(remove now-unused _alert_scope_tickers) and alert only on watchlist tickers.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three fixes to over-firing S/R proximity alerts:
- Route through cluster_sr_zones (the same merger the chart uses) instead of raw
SRLevel rows, so near-duplicate levels (e.g. CVX 183 + 185) collapse into one
zone and one alert.
- Alert only the single NEAREST strong zone per ticker, not every nearby level.
- Scope to watchlist + qualified-setup tickers via _alert_scope_tickers (was
iterating all watchlist entries only; qualified setups are now included too).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Closes the action loop — instead of polling the dashboard, the platform pushes
actionable signals to Telegram. New hourly 'alerts' job dispatches four
toggleable triggers, deduped via a new alert_log table (cooldown-based for
qualified/S-R/digest, watermark-based for score deterioration). Admin → Settings
gains a Telegram panel (write-only bot token, chat ID, per-trigger toggles, Send
Test). Credentials follow DB > env precedence (TELEGRAM_BOT_TOKEN / _CHAT_ID).
Backend: alert_service + AlertLog model + migration 005, scheduler job, admin
endpoints/schema. Frontend: AlertSettings panel, hooks, api, types.
Deploy: run alembic upgrade (new alert_log table).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
