Decision rights · measurable rules · controlled change
Build an Auditable ES Trading Plan
“Trade pullbacks, use discipline, stop after a bad day” is not a plan. Nobody can reconstruct what was permitted, what triggered the order or whether the stop was changed. An ES plan becomes useful when a second person can apply its rules to the same timestamped evidence and reach the same trade, wait or stop result.
One plan, one stated job
Define What the Plan Is Trying to Control
An objective should name the instrument, holding horizon, decision frequency and evidence standard. “Make money trading ES” cannot distinguish a process failure from ordinary outcome variance.
Unusable objective
Capture good moves
“Good,” “move” and “capture” have no timestamped definition. Any outcome can be rationalized afterward.
Auditable objective
Execute one defined setup
For example: test a named intraday setup in a declared session, with no hold through scheduled exclusions, fixed permission limits and order-level review.
It controls how a hypothesis is executed and measured. Profitability requires separate, forward-safe research with costs and out-of-sample evaluation. A clean plan can still produce a negative result.
Set gates before the chart becomes persuasive
Build a Permission Matrix
| Gate | Required pass | Reduce branch | Stop branch |
|---|---|---|---|
| Product | Exact ES or MES month/year, live data and routing permission | Use smaller product only after fresh sizing | Unknown symbol or entitlement |
| Session | Permitted clock window and CME status | Narrower subwindow supported by evidence | Maintenance, halt, holiday uncertainty |
| Event | No excluded release or hold risk | Lower cap if prewritten and tested | Unbounded scheduled or unscheduled event risk |
| Market quality | Spread, depth and intended quantity within thresholds | Smaller whole-contract quantity | Stale feed, wide spread or insufficient depth |
| Account | Daily, drawdown, portfolio and live-margin gates pass | Lower independent cap | Any breached or unknown limit |
| Operator | Connection, platform and required attention available | Simulation-only session | Fatigue rule, interruption or unresolved incident |
A permission matrix should return a machine-like outcome: GO, REDUCE, WAIT or STOP. “Use judgment” may be unavoidable in a narrow field, but the plan should identify the evidence and who has authority to exercise that judgment.
Turn the setup into observable fields
Write a Setup Card That Can Be Replayed
The setup card separates facts available before the order from labels assigned afterward. Every level and state needs a calculation clock, source and expiry.
Hypothetical setup card
Context: exact dated ES contract; session window; no excluded event. Candidate: price returns to a timestamped reference after a forward-defined impulse. Trigger: a specified close or order-flow condition known at bar completion. Invalidation: a predeclared price proving the setup no longer applies. Expiry: cancel if not triggered by the stated clock or if market quality fails.
- Data clock
- Known at decision
- Entry
- Exact branch
- Invalidation
- Price + reason
- Expiry
- Time/state
Context
What must already be true?
Dated contract, session, event state, market quality, volatility state and any directional filter.
Trigger
What new observation acts?
Exact price, timestamp, completion rule and permitted order type. No future bar may redefine the trigger.
Invalidation
What proves the idea wrong?
A market observation independent of the desired contract count. It is not “when the loss feels large.”
Expiry
When does the idea become stale?
Time, event, roll, liquidity or structural change that cancels the setup even without a fill.
Risk permission comes before order permission
Separate Trade, Day, Portfolio and Collateral Limits
| Limit | Plan field | Required action at boundary |
|---|---|---|
| Per decision | Stressed dollars from stop, slippage, fees and gap reserve | Floor to whole contracts; zero is valid |
| Open aggregate | Stressed loss across all current positions | Block new correlated exposure |
| Daily | Realized loss plus open stressed risk | Cancel new entries and follow reduction protocol |
| Drawdown | Distance to account or program hard boundary | Stop trading before forced breach |
| Margin | Timestamped CME and broker requirement plus buffer | Use as an independent cap, never as maximum loss |
final size = min(trade-risk size, daily cap, portfolio cap, margin cap)|unknown = zeroA signal does not choose an order
Write the Execution Branches Before Entry
CME Globex supports defined order types, while platforms may add synthetic behavior. The plan must describe the actual exchange or broker order, its trigger, price protection, duration and failure branch.
Entry condition passes
Is price control more important than immediacy?
Yes: use the specified limit logic, accept non-fill and set cancellation time. No: use only an approved executable order with a maximum cost and protection rule.
Partial fill
Is partial quantity permitted?
Yes: recalculate risk from actual average fill and protect filled quantity. No: cancel remainder and follow the written flatten-or-keep branch.
Stop trigger
Did the exit behave as modeled?
Yes: record trigger-to-fill slippage. No: follow emergency exposure reduction; do not widen the stop to preserve the thesis.
- Before submit: symbol, side, quantity, order type, price, duration and attached orders read back correctly.
- After submit: exchange or broker acknowledgement received; rejects and pending states handled.
- After fill: actual average price, remaining quantity and protective order reconcile.
- After cancel: acknowledgement proves no working order remains.
- After exit: account position is flat or equals the explicitly approved remainder.
Define emergency authority
An Incident Plan Is Part of the Trading Plan
| Incident | Immediate control | Evidence to retain | Resume gate |
|---|---|---|---|
| Market data stale | Stop new orders; verify position through an independent channel | Last update time, connection state, screenshots/logs | Fresh synchronized data and reconciled position |
| Order state unknown | Do not resubmit blindly; contact broker or use approved order-status channel | Client ID, acknowledgements, rejects and timestamps | One authoritative working-order inventory |
| Protection missing | Use prewritten reduce/flatten procedure | Fill, attempted child order and error | Cause repaired and tested in simulation |
| Risk limit breached | Cancel entry authority and reduce under policy | Equity, positions, fills and cause | Formal incident review; no same-session improvisation |
| Wrong contract | Stop automation and correct exposure under cost controls | Old/new symbols and every fill | Symbol mapping and roll process repaired |
Name the action, account, communication route, authority and evidence. If the plan assumes a tool remains available during the failure that disables it, the branch is not operational.
Judge process before outcome
Review Every Decision, Including No-Trade Outcomes
| Record | Minimum fields | Question |
|---|---|---|
| Permission | All gate values and pass/reduce/wait/stop output | Was the decision authorized? |
| Setup | Context, trigger, invalidation, expiry and screenshots/data IDs | Could another reviewer reproduce the label? |
| Risk | Budget, stress inputs, floor math and binding cap | Was size computed before entry? |
| Execution | Decision, arrival, submission, acknowledgement and fill timestamps | Where did cost enter? |
| Management | Every modification with reason and timestamp | Did the plan change after risk was known? |
| Outcome | Gross, costs, net, maximum adverse/favorable path | What happened, without confusing it with plan quality? |
| No trade | Binding rejection gate | Are filters blocking too much, too little or appropriately? |
Change one controlled version at a time
Freeze the Plan Before Evaluating It
- Assign an ID.Version, effective date, instrument set, data sources and author.
- State the hypothesis.What behavior is expected and what would falsify it after costs?
- Freeze the rules.Do not edit triggers or exclusions after seeing the same-period outcomes.
- Collect a declared sample.Include rejects, misses, partials, incidents and no-trades.
- Evaluate separately.Measure adherence, cost estimate error and outcome uncertainty.
- Retire or revise.Create a new version with a reason; preserve the old record unchanged.
Minimum viable plan state
Simple means few rules, not vague rules
A production-quality first plan may trade only one setup, one session and one product. It still needs complete permissions, sizing, execution, incident and review branches before live risk.
Sources, methods and editorial disclosure — reviewed August 28, 2026
- CME Position and Risk Management for scenario-based position and risk controls.
- CME Futures Order Types for current exchange order-type mechanics and protection behavior.
- CME Pre-Trade Risk Management for permission and order-management control concepts.
- CFTC futures risk disclosure for leverage, margin-call and loss-beyond-deposit risks.
- ES/MES contract-mechanics authority and ES/MES sizing workflow for current contract inputs and whole-contract math.
Sources were reviewed August 28, 2026. The setup language is an illustrative control template, not a strategy recommendation or empirical finding. No profitability, optimal session, order type, risk percentage or setup edge is claimed.