Skip to content
B21 Decision Desk Real money

Market awareness

Two readings decide how hard this engine is allowed to push. The regime decides whether it may buy at all. The posture score decides how large each buy is. Everything below is read only — nothing on this page changes what the engine does.

Engine state Gate mode — Regime — Posture — Brake — Entries — Read only
Position size multiplier
Loading the posture record…
Pressure
Market half of the score
Book half of the score

The market regime

A replica of the three-factor rule the backend runs: the equal-weight Nasdaq-100 against its 21-day average cloud, the VIX against its own 21-day average, and the ratio of QQQ to XLP as a measure of appetite for risk. All three have to agree before the regime changes. When they do not agree, the regime stays where it was last set — it latches. This replica refreshes about every five minutes.

The three factors, right now

Scrolls sideways — 4 columns
Each factor's current reading against the level that decides its vote. Readings are shown to two decimals; the stored values sit in the panel below at full precision.
Factor Vote today Reading Reference
Loading…

Why it says BEAR on a mixed read

Scrolls sideways — 3 columns
Regime changes on record in this payload, most recent first. Days held is calendar days, not trading sessions.
Set to When Days held
Loading…

How the posture score is built

Fourteen readings, in two independent halves. Each reading contributes nothing at all until it crosses its quiet level, then ramps in a straight line up to its full level, then stops — so no single reading can take over. Every contribution below is the reading's own weight multiplied by how far it has travelled between those two levels. The columns add up. The page checks that they do, and says so.

Scrolls sideways — 7 columns
The market half. Rows are in the order the engine stores them.
Market reading Now Quiet below Full at Weight Contribution Share of weight
Loading…
Share of weight is contribution ÷ weight. The engine also stores a whole-number version of it for the market half; this page shows the exact ratio.
Scrolls sideways — 7 columns
The book half — whether the edge is working. Rows are in the order the engine stores them.
Book reading Now Quiet below Full at Weight Contribution Share of weight
Loading…
Scrolls sideways — 2 columns
How each book reading was calibrated, in the engine's own words.
Book reading Calibration basis
Loading…

From the two halves to the headline

every step re-computed here and compared with what the engine stored

The two halves are scored separately and the more pressured of the two governs. That is the whole reason a calm market can still produce a small position: a losing book overrides a quiet tape, and never the other way round.

Engine reason string

What that buys per trade

Pressure lands in one of four bands, and the band sets the percentage of the account risked on a single trade. Sizing back up needs pressure under 10% on both halves, which is why the engine cannot grow its bets while it is losing.

Scrolls sideways — 4 columns
The four bands. The active band is marked and carries a red edge.
Band Pressure up to Risk per trade Status
Loading…

The rest of the sizing record

The trades behind the book half

Every closed trade in the window the posture model reads, oldest first, measured in R — the result divided by the money that trade was risking. A trade that loses exactly its planned stop is −1.00R. Anything past that line lost more than it was supposed to.

R multiple per closed trade

Drawn from the table below with no chart library and no CDN, so a blocked network cannot take a number off this page. Every bar has its exact value in the table.

Scrolls sideways — 4 columns
Closed trades, oldest first.
Ticker R multiple Exit reason Closed
Loading…

Where the posture has been

One graded reading per session, as stored. The strip shows the shape; the table carries every value.

Scrolls sideways — 4 columns
Graded posture by session, most recent last.
Session State Pressure Size multiplier
Loading…

Flags carried but not recorded

The posture row still carries ten single-signal on/off flags alongside the graded score above. They are listed because leaving them out would hide the fact that they are empty.

Scrolls sideways — 2 columns
Single-signal flags stored on the posture row.
Field Value
Loading…

How to read this page

This is a pressure gauge, not a set of switches. Each reading has a quiet zone where ordinary conditions contribute exactly zero, so nothing sits on the score as permanent drag. Past that it ramps, and past its full level it saturates. Size comes off gradually. There is no cliff.

The market half and the book half are scored independently, and the more pressured of the two governs — so a losing book can override a calm tape.

Every quiet level and full level is calibrated against this system's own 754 sessions of history, not against textbook numbers. Five textbook thresholds were tested and thrown out: the raw McClellan Oscillator, the IBD distribution-day count, a plus-two-sigma stretch, the raw SHY/HYG cloud break, and a stale constituent list. Each of them fired between 36% and 55% of the time, which is another way of saying they said nothing.

The regime latch is path dependent, and that is deliberate. The three factors are checked against live prices through the session, so a brief alignment that lasts minutes can set the regime for days even if the settled close would never have set it. The page shows the current reading of each factor next to the latched regime precisely so the difference is visible rather than confusing.