Decision Timing Model

Assumptions

Everything the model believes, stated before the math. If you do not buy one of these, do not buy the output.

1. Every deadline cites its source or term; unsourced or malformed deadlines are rejected and listed.
2. WAIT requires a named, dated, monitored information event - generic market calls are refused.
3. An expired option is reported EXPIRED and cannot be recommended.
4. Waiting value is UNKNOWN without economics inputs; staging is offered only when its cost is documented.
5. Nothing here is a price forecast; timing classes come from deadlines, evidence calendars, and stated economics.
6. The model classifies only - the user's money or commitment requires separate approval before any action.

Derivation

Deadlines enter only with a source or term reference and a well-formed date; rejected deadlines are listed with reasons. The classification ladder is deterministic:

• EXPIRED - every sourced deadline has passed; an expired option cannot be recommended.
• WAIT - a named evidence event is expected at least 14 days before the next deadline; waiting is option logic, not a market call.
• ACT - the next deadline is within 90 days and no named evidence arrives in time.
• STAGE - further out, and a documented staging cost preserves the baseline.
• HOLD - none of the above; hold with a dated recheck, since a wait without a named event would be a generic market call.

The cost of waiting is computed only when a deterministic daily cost is provided:

where the 14-day buffer keeps the last evidence event clear of the deadline. Without economics inputs the waiting value is UNKNOWN - no default is assumed.

Identification strategy

Why the estimate supports the claimed meaning - and where the model refuses to claim more than the data carries.

• Sourcing is the identification rule for time itself: a deadline exists because a document says so, which is what separates a decision clock from a rumor.
• WAIT is identified by a named information event, not by sentiment: something specific must be expected to resolve uncertainty, with a date and a monitor, or the wait is refused.
• The classification reason prints with the class - every ACT, WAIT, STAGE, HOLD, EXPIRED carries the rule that fired and the dates it fired on.
• Market timing is explicitly not price forecasting: nothing in the output implies a price path, and the model never commits money - it classifies.

Inputs and outputs

Calibration and gate evidence

Validation design

• Gate-verified classification paths: unsourced deadlines rejected, WAIT gated on a named monitored event, expired options un-recommendable, near-deadline-no-evidence forced to ACT.

Evidence on record

• Smoke gates at seal (9 m31-specific): unsourced deadline rejected and listed; WAIT requires a named monitored event; the trigger names the evidence event; waiting value UNKNOWN without economics; an expired option cannot be recommended; near deadline with no evidence is ACT; waiting cost deterministic when economics provided; determinism (identical input hash on re-run); audit row written for every run.

Gates at seal

• smoke: 9 m31 gates incl. EXPIRED / ACT / WAIT path gates, determinism + audit row (suite counts at seal per the build agent's seal report)
• sealed at d3b3875 on atlas-nl-shock-redesign

Failure modes

The ways this model declines, degrades or goes UNKNOWN - and what it prints when it does.

• No sourced deadline and no pending named evidence: HOLD - a wait without a named event is refused as a generic market call.
• All deadlines passed: EXPIRED; the option cannot be recommended.
• Economics absent: waiting value prints UNKNOWN rather than assuming a default cost.

Sensitivity

• The 14-day evidence buffer and 90-day act window are the policy dials: tightening the buffer converts marginal WAITs into ACTs, and both are versioned constants rather than per-case judgment.
• Waiting cost scales linearly with days to the deadline; a small daily rate over a long window still compounds into a visible total.