Skip to content
← Back to 12-Step Operating Checklist

Operator brief · 271

The first three steps write. Nothing is read until the fourth.

The key idea

Two phases

A write phase and a read phase, with a hard boundary between them.

Steps one through three touch the ledger only: each trade entered or updated with its dropdowns and inclusion flag, floating and active risk refreshed on every open position with its stop note, and the exposure column confirmed to be counting the right rows. Steps four onward consult outputs — control selections, the gate block, the primary output, the exposure audit, the packet. The boundary is not a stylistic preference. Everything in the read phase is a formula evaluated against the write phase's contents at the instant it is read, so a read performed mid-write is not a partial answer or an approximate one. It is a complete, correct, confidently formatted answer to a question about a ledger that is halfway between two states.

FigureThe phase boundary, and what crosses it
no reads above the lineUpdate every tradestatus, branch, inclusionRefresh open positionsfloating R, active risk, noteConfirm exposure columncounting the right rows— boundary —ledger now currentSelect controls, read gatefirst readsRead output, copy packetvalues leave the console

Every arrow below the boundary carries a value computed from the ledger above it. Nothing recomputes when the ledger is corrected afterwards — the packet has already been copied and the decision already made.

Why the order is not obvious

The natural way to work a spreadsheet is to update and glance, update and glance.

The habit the sequence corrects is a reasonable one everywhere else. Editing a cell and watching the dependent figures move is how a person builds confidence that a workbook is wired correctly, and it is how most operators learn a new tab. It becomes a liability here because the console's outputs are not diagnostics of the workbook, they are evidence bound for another workbook, and the glancing habit means the operator forms an impression of the cycle from readings taken partway through the update. The impression is what survives. By the time the ledger is complete and the figure has settled to its true value, an expectation already exists, and a settled figure that contradicts an expectation formed ten minutes earlier gets checked far more sceptically than one that confirms it.

The dependency

Exposure has to be current before the gate reading means anything.

Beyond the general principle, the specific chain matters. Open exposure is computed from active risk on included open rows, and it feeds remaining capacity, suggested fresh trades and the whole smart sizing block. The gate is computed from equity against the peak, and it selects the capital-state row that prices the pool those calculations run against. So an exposure figure taken before the ledger is current is subtracted from a pool that may itself be about to change, and the resulting capacity number is a composition of two figures from different moments. Neither is wrong individually. Their combination describes no state the account was ever in, which is the characteristic signature of reading a calculation surface partway through an update.

The irreversible step

Copying the packet is where a mid-write reading stops being recoverable.

Up to the moment the packet leaves the console, an incorrect reading costs nothing — the ledger gets finished, the figures settle, and the operator reads again. The copy is the point of no return, because from there the values exist in a second workbook that has no connection to the first and no way to learn that the source has since changed. The throttle then decides from them, the decision is logged, and the log records a snapshot of evidence that the console itself no longer agrees with. This is why the checklist places the packet copy late and behind an explicit exposure audit: the last thing before the transfer is a check that the thing being transferred is finished.

In practice

Treat the ledger as closed before opening the calculator.

The workable version of this discipline is to think of the two phases as two sittings rather than one continuous pass. Bring every trade in the cycle current — status, inclusion, floating and active risk, stop note — and only then move to the calculator. If something in the read phase reveals a ledger problem, which happens, the correct response is to return to the ledger, fix it, and re-read from the start of the read phase rather than continuing forward with one corrected input among several already-read figures. It costs a minute and it is the difference between a packet describing the cycle and a packet describing the cycle as it stood at several different times.

  • The console answers every question at every moment, from whatever the ledger holds.
  • Impressions formed mid-update survive the correction that follows them.
  • The packet copy is irreversible — the throttle cannot learn the source changed.

The key idea

Live calculation is a feature that has to be paired with a discipline about when to look.

A surface that recomputes instantly is enormously useful and quietly hazardous for the same reason: it never signals that its inputs are incomplete, because incompleteness is not a state it can represent. The workbook cannot know whether the operator has finished entering trades. The only place that knowledge exists is in the sequence the operator follows, which is why the sequence is written down and numbered rather than left to judgement. It is a protection implemented in procedure because there is nowhere in the software to put it.

Connected inside MARS

Every brief documents the same shipped system.

The complete MARS package — eleven workbooks, three TradingView indicators, the full manual library — $497.