The trading plan

Definition

A trading plan is the written constitution of an account: what it trades, why that should work, how positions are sized, what triggers entries and exits, what risk is never exceeded, and how performance is judged. Its function is not prediction but PRE-COMMITMENT — moving every decision possible from the moment of emotion to the moment of calm (bias-loss-aversion and its siblings operate at decision time; the plan is written before decision time). For this platform the concept is native: an agent’s directives and strategy parameters ARE a trading plan, machine-enforced.

How it works / structure

• The canonical sections (engine-mappable):
  • Universe & edge statement: what is traded and the falsifiable reason it should pay (every entry in this KB’s strategy pillars is a template).
  • Sizing rules: the fixed-fractional/vol-target core (risk-fixed-fractional, risk-volatility-targeting) plus the never-exceed caps (per-position, per-theme, account drawdown — risk-max-drawdown-budget).
  • Entry/exit specifications: setup conditions (disc-playbook-setups), stop and target logic (mgmt-stop-loss, mgmt-profit-target), and the time-based fallback (mgmt-time-based-exit).
  • Review cadence & amendment rule: when performance is examined (disc-journaling-review) and — critically — HOW the plan itself may change (amendments in calm, never mid-drawdown; the meta-rule that protects the rest).
• The evidence frame (labeled): the plan concept is practitioner canon (Tharp’s framework is the standard exposition); its support is indirect but strong — the documented performance drag of unplanned behavior (overtrading, disposition-effect exits, revenge sizing: pillar 13’s catalog) is precisely what pre-commitment removes.
• The platform translation: user directives (per agent) + strategy parameters + risk caps = the plan, with the enforcement problem solved by construction — agents cannot deviate from what is parameterized; the human’s plan-discipline problem becomes a parameter-honesty problem.

When it applies

Account inception (no live capital without the written plan — the platform’s structural equivalent: no agent runs without directives and caps); strategy addition (each new strategy amends the plan through the amendment rule); drawdown states (the plan’s never-exceed rules are exactly the clauses that bind when feelings argue otherwise).

Risk profile & failure modes

• The unwritten plan: intentions held mentally renegotiate themselves under stress — the documented gap between stated and executed strategy; writing is the technology.
• Mid-drawdown amendment (the signature failure): loosening stops or doubling size while losing is plan destruction wearing the plan’s clothes — the calm-amendment meta-rule exists for this.
• Over-specification: plans regulating every tick produce abandonment; the documented sustainable form specifies the RISK layer rigidly and the tactics layer loosely.
• Plan-reality drift: markets change; a plan reviewed never is a museum piece — scheduled review (quarterly) with the amendment rule is the maintenance loop.

Evidence & limits

The framework is practitioner literature (labeled); the bias catalog it counters is peer-reviewed (pillar 13 citations); direct RCT-style evidence on plan adherence in trading is thin (labeled honestly). The platform’s version is enforced by architecture, which converts the discipline claim into a design fact.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Trades taken inside plan parameters outperform out-of-plan trades in this account’s ledger (plan-value audit)” — falsified by the tagged-ledger comparison.
• “Accounts with hard-enforced caps show smaller worst-month losses than matched discretionary accounts at equal average exposure (pre-commitment check)” — falsified by the cohort comparison.

Cross-references

• The components: disc-playbook-setups, mgmt-stop-loss, mgmt-profit-target, risk-fixed-fractional
• The review loop: disc-journaling-review, disc-process-vs-outcome
• The failure catalog it counters: bias-overconfidence, bias-loss-aversion