The review intentionally separates the opening move from the full-day outcome. Open Direction measures whether the morning call anticipated the first market move; Session Resolution measures whether that broader signal still described the day by the close. The review UI is designed to render these dimensions from normalized stored comparison rows so the table text stays aligned with the actual grading logic.

The evening review checks whether the morning confidence level was appropriate for the realized tape. A 7/10 bearish call that only gets the open right but misses a strong rebound should lose calibration points even if the open score is strong.

The review table uses score-aware verdict bands to summarize grading strength. Strong means 80% or better, Good means 65-79%, Mixed means 40-64%, and Weak means below 40%.

A normalized evidence row stored with the review that captures the morning expectation, realized outcome, score, verdict, and explanation for each scored category. It keeps live review and archive views aligned because both pages read from the same comparison payload rather than rebuilding the language independently.

A concise after-close narrative that summarizes how the session actually unfolded, which drivers dominated, and where the realized tape diverged from the morning framing. It complements the numeric rubric by preserving the sequence and texture of the day.

A proposed review-logic or prompt adjustment surfaced during meta-analysis for human inspection. It is advisory until explicitly accepted, and it exists to capture recurring grading blind spots in a form that can be reviewed, discussed, and tested.

Strong directional bias with increasing intensity throughout the session. In the morning workflow, this is an expected session shape, not a guaranteed realized outcome. It is usually associated with a high Volatility Ratio (>0.6), a significant directional move, broad participation, and lower-timeframe evidence that the tape is accepting away from fair value rather than rotating back and forth.

A violent expansion of the daily range. Price moves decisively beyond previous day limits, often triggered by surprise news. In practice this is associated with extreme close-versus-range behavior, a large realized move relative to expected range, and strong participation rather than just a small premarket gap.

A quiet or conflicting narrative where the market lacks a clear leader. Price is more likely to mean-revert within range than to sustain directional expansion. A session can still be classified as CHOP even when it has a visible upward or downward bias, if volatility and participation never build enough to confirm a true trend or reversal.

Market exhaustion at major psychological or data-driven levels. Price moves strongly in one direction before "snapping back" sharply as the narrative shifts and materially rejects the earlier move. A mere afternoon fade is not enough; the reversal needs meaningful intensity, participation, and intraday sector rotation to change the character of the session. If the tape only looks reversal-like visually but realized intensity stays muted, the system should keep it in CHOP with directional bias rather than promote it to a true REVERSAL.

A high-level structural assessment of the market environment (e.g., Structural Bearishness). It anchors the daily briefings, providing a broader context for short-term news events.

ALIGNED: The daily sentiment matches the 90-day thesis (e.g., Bullish Daily vs Bullish Strategic).
DIVERGING: The daily sentiment contradicts the thesis (e.g., Bearish Daily vs Bullish Strategic), signaling a potential "Counter-Trend Breakout" or a "Deep Dip."

FRESH: Generated within the last 10 days; high relevance.
AGING: 10–17 days old; thesis still active but the next biweekly refresh is due.
STALE: 17–27 days old; thesis still renders, but a fresh strategic update is increasingly preferred.
EXPIRED: 28+ days old; thesis logic is excluded from daily synthesis until a newer strategic view replaces it.

A synthesis layer in the outlook workflow that explains how the current strategic thesis, recent review evidence, and regime-health read either reinforce one another or pull in different directions. It helps separate a structurally aligned setup from a tactically noisy one.

The explicit list of conditions that could weaken or invalidate the current strategic thesis. These risks focus the outlook on what would change the structural case, not just what would create a noisy single-session deviation.

The subset of strategic conditions the system can check automatically. If one is breached, it can request a strategic refresh.

The broader set of strategic conditions kept for visibility and context. These stay on the page even when they are descriptive, manual-only, or not yet checked automatically.

The transition layer between structural outlook and the next trading session. It translates the longer-horizon thesis into concrete near-term conditions, trigger zones, and evidence to watch so the outlook remains actionable instead of abstract.

The raw RVOL metric compares current traded volume against the average of the prior 10 full daily sessions. The system also uses a contextual premarket interpretation so extremely early snapshots are not judged against full New York session participation as if they were directly comparable.

The raw Volatility Ratio compares the current realized move against a 10-session average daily range baseline. During premarket phases, the system also uses a contextual interpretation so a small London or early pre-open move is judged against expected premarket behavior rather than against a completed New York cash session.

A compact lower-timeframe layer derived from S&P 500 futures bars. It summarizes session shape through features like opening drive, trend consistency, VWAP behavior, reversal evidence, and chop score. It is used as an overlay for archetype selection, not as a replacement for the broader snapshot.

The close-side review uses regular-session lower-timeframe features, not just daily bars, to classify the realized session shape. It looks at realized opening drive, VWAP behavior, reversal strength, and chop score to judge whether the day actually behaved like trend, chop, breakout, or reversal. This is distinct from the morning intraday overlay, which is based on premarket futures context rather than the realized cash-session tape.

A guardrail that tells the system whether lower-timeframe futures evidence is usable, thin, degraded, or unavailable. If the latest bar is stale or bar coverage is too sparse for the current New York session phase, the system lowers confidence and avoids letting weak intraday evidence dominate the morning call.

A 1-10 confidence scale assigned by the system. It combines narrative strength, price action, breadth, cross-asset agreement, and event risk. As a rough interpretation guide, 0-3 suggests low conviction, 4-6 suggests moderate conviction, and 7-10 suggests high conviction. It is not a literal probability of being correct. High conviction implies not just a strong move, but a move that is also internally confirmed.

A compact internal-strength layer built from sector breadth, average sector movement, dispersion, and participation flags. It helps the system distinguish a broad, confirmed move from a narrow or fragile one. For example, green index futures with weak cyclical sectors can still represent a low-quality or unstable opening tape. The system also treats these readings with caution because premarket quotes are not always perfectly time-synced across futures, ETFs, and other proxies.

The expected directional pressure into the cash open based on the premarket tape, overnight narrative, and internal confirmation. It focuses on the first meaningful move after the bell rather than the final close.

The broader expected direction and tone for the full regular session after accounting for opening setup, macro schedule, and follow-through risk. It is the directional anchor used for session-resolution grading.

A breadth-style read on how many of the key sector and risk groups are actually supporting the move. The system now tracks per-group breadth counts such as up, down, and flat members so it can distinguish genuinely broad leadership from a misleading average driven by only one or two sectors.

The ranked pattern of expected leaders and laggards across sectors, indices, and risk proxies. It helps define whether the setup looks cyclical, defensive, tech-led, broad, or unusually fragmented.

Agreement between direction, breadth, sector participation, and cross-asset behavior. A move with internal confirmation has supporting evidence beneath the headline index print rather than relying on one narrow driver.

The set of internal measurements that judge participation quality, including sector breadth, average sector move, dispersion, and related confirmation flags. These metrics help determine whether a move is broad, narrow, or conflicted.

The directional lean implied by the internal participation map. It can reinforce the headline session bias, soften it, or warn that the visible move lacks enough support to trust at face value.

A special morning-analysis posture used when high-impact macro data is scheduled during the session. On these days, the pre-data tape is treated as provisional rather than fully settled, and the briefing is expected to use more conditional scenario language around the release time and its possible surprise outcomes.

Measures the expansion or contraction of the Average Daily Range (High-Low) relative to a 10-day baseline. A drift of >30% indicates that the market has entered a new volatility regime.

The mathematical spread between the best and worst performing sector ETFs. High Dispersion (>2.0%) suggests sectors are moving independently, making standard broad-market calls more complex.

A meta-analysis signal that compares average open-direction success with average full-session resolution success. A large positive gap means the system is reading pre-open momentum well but over-projecting that opening tape into the close.

A 1-5 assessment of whether the current Market Regime still explains the session. Very weak readings can accelerate a meta-analysis run via the hysteresis trigger, but they do not automatically rewrite the live system.

An advisory LLM-based check that asks whether a proposed prompt change would likely have improved a prior miss. It is stored as supporting evidence only and does not count as a true historical backtest or auto-promotion rule.

The average price range (High to Low) over a specified period (usually 10 days). It provides the baseline for expected daily movement. A 10-day ADR of 120 points means the market "normally" moves 120 points in a session.

A session-context layer that adjusts how pre-open activity is interpreted across overnight, early premarket, and late pre-open phases. Its purpose is to reduce false low-conviction signals caused purely by the fact that liquidity is naturally thinner before the New York cash open.

A practical limitation of the pre-open snapshot: different instruments may update on slightly different clocks. Index futures often move first, while sector ETFs and some cross-asset proxies can lag. The system therefore treats early contradictions as tentative unless they are broad, persistent, or confirmed by multiple instruments.

Bitcoin is tracked as a high-beta risk appetite proxy. Its behavior helps test whether equity strength is being confirmed by broader speculative capital or whether risk is fragmenting beneath the surface.

The system checks whether Bitcoin confirms or contradicts the day's equity risk tone. Confirmation strengthens the risk narrative; divergence suggests a thinner or more fragile move even when stocks are green.

Morning story choice is AI-led but constrained by per-driver candidate buckets. The goal is to preserve real cross-currents across active drivers instead of collapsing to one dominant headline cluster. The system prefers fewer, stronger stories and is allowed to leave weak buckets blank rather than forcing filler coverage.

The specific releases, headlines, flows, or price levels most likely to redirect the session. They give the briefing a concrete map of what can accelerate, invalidate, or reframe the current narrative once trading gets underway.

The review page tracks two separate clocks: when the grading output was generated, and when the closing market data used for that grading was pulled. This avoids conflating report time with data freshness.

A stored informational review note that measures whether the displayed morning stories captured the session's most important drivers and contradictions. It is visible in review/archive history but does not change the weighted score.

A compact explanation of the most important evidence supporting the current call. It gathers the strongest confirming and conflicting inputs into one readable synthesis so the reasoning behind the bias is easy to audit later.

Triggered when inter-market correlations break down (e.g., Gold crashing while Bond Yields spike, or Nasdaq rallying while BTC lags). This signifies "Inter-market Decoupling" and often precedes a higher-volatility session as different asset classes re-price at different speeds.

A tactical state where traditional hedges (Gold/Bonds) fail to act as a defensive buffer, forcing a chaotic reorganization of defensive assets. BTC’s relative performance provides the primary directional bias in these states.

The system detects full US market holidays and shortened sessions using an ET-aware market calendar. During closures, it avoids producing stale live-session assumptions and keeps the dashboard oriented around the next valid session.

Morning and review workflows run on scheduled automation, but the system is also built to support manual reruns. Execution uses broad DST-safe polling plus ET-aware gate windows, so actual run time can vary within the approved window instead of landing on one exact minute. In practice, the morning briefing usually appears in two pre-open passes, with the later pass typically landing about 5 to 60 minutes before the open, while the review usually appears within about 120 minutes after the close. The strategic outlook refreshes on a biweekly cadence (every other Sunday), keeping the thesis within the fresh or aging window under normal operation.

Core text workflows use a three-model routing stack: Gemini 3 Flash Preview first, Gemini 2.5 Flash second, and Gemini 3.1 Flash Lite Preview as the high-volume fallback. This ordering keeps the strongest synthesis model in the lead while preserving a stable reasoning standby and a generous-capacity backup path.

Thinking levels are tuned by workflow rather than applied globally. Narrative, Meta, and Strategic workflows default to HIGH; Grading defaults to MEDIUM; Search defaults to LOW; and Simulation defaults to MEDIUM. Some model families express this as an explicit thinking level, while others fall back to their native automatic reasoning behavior.

A logical guard prevents the system from treating an already-open session as a fresh premarket prediction window. This preserves the historical integrity of the morning call and reduces accidental retrospective regeneration.

If a manual review run happens before the current session has actually closed, the review builder can fall back to the most recent completed market session for verification instead of grading incomplete data.

The bi-weekly review that labels the regime as VALID, DRIFT, or CHANGE. It can suggest prompt or config changes for human review, but it does not change the live system by itself.

The reviewed status of a suggested change. It records whether an idea is pending, accepted, rejected, or still held back from use.

The overarching thematic state of the market (e.g., Fed Centric, AI Euphoria, Reflation). Regimes determine the "weight" given to specific news drivers.

The logic bridge between daily execution and bi-weekly meta-analysis. Very weak regime-fit readings can force an earlier system audit, but they do not automatically rewrite prompts or configs. Human review still controls any accepted changes.

The system monitors primary indices (S&P 500, Nasdaq, Dow), sector ETFs, and a broader cross-asset risk map that includes Oil, Gold, Treasury Yields, USD, and Bitcoin. It also uses market internals, premarket breadth/dispersion signals, scheduled macro events, and lower-timeframe overlays to judge whether the tape is broad, fragile, event-driven, or internally contradictory.

The confidence attached to the current strategic outlook after weighing structural evidence, recent review performance, and regime stability. It expresses how strongly the system trusts the thesis itself, which is a different question from how forcefully it leans on a single session call.