A profitable day can hide a broken process
A green account balance answers one question: how much did the recorded activity gain or lose? It does not tell you whether the setup was valid, the planned risk was respected or a mistake happened to work. Treat those as separate questions.
This article is a review procedure, not a promise of becoming profitable. If you need to write the rules first, start with the trading plan guide. If the rules already exist, compare the intended decision with the actual order and result.
| Question | Required evidence | What it cannot establish |
|---|---|---|
| Did the strategy qualify? | The frozen setup definition and information recorded before entry. | That the trade should have won. |
| Was the plan followed? | Order, size, stop, timing and management records. | That the strategy has positive expectancy. |
| What did the account earn? | All fills, fees, financing and reconciled account changes. | That good results came from good decisions. |
Four invented trades show why the distinction matters
The following ledger is deliberately invented. Each trade starts with USD 100 planned risk, and USD 10 is assumed for its total round-trip costs. Costs are simplified for the exercise; no broker, strategy or historical performance is represented.
| Trade | Gross | Costs | Net | Net R | Plan followed? | Record |
|---|---|---|---|---|---|---|
| A | 250.00 | 10.00 | 240.00 | 2.40 | Yes | Eligible setup; planned execution |
| B | -100.00 | 10.00 | -110.00 | -1.10 | Yes | Eligible setup; planned stop |
| C | 80.00 | 10.00 | 70.00 | 0.70 | No | Entry outside the approved session |
| D | -180.00 | 10.00 | -190.00 | -1.90 | No | Stop widened after entry |
The complete account result is USD 10.00, or +0.10R across the four equal-risk trades. The compliant group contributes USD 130.00; the non-compliant group contributes USD -120.00. Plan adherence is 50%.
Trade C made money despite an ineligible entry. Trade B lost while following the declared plan. Calling C a good decision and B a bad one simply because of their outcomes would teach the wrong rule. The sample is far too small—and entirely fictional—to establish an edge or a reliable cost of mistakes.
example_id,initial_risk_usd,gross_usd,costs_usd,net_usd,net_r,followed_plan,note A,100,250,10,240,2.40,true,Eligible setup; planned execution B,100,-100,10,-110,-1.10,true,Eligible setup; planned stop C,100,80,10,70,0.70,false,Entry outside the approved session D,100,-180,10,-190,-1.90,false,Stop widened after entry
Capture the decision before memory edits it
Before entry, save the setup version, eligibility checks, intended order, initial risk and the conditions that cancel the idea. After execution, add actual fills and costs. After exit, reconcile the result. A screenshot taken after price moved cannot replace missing evidence of the original decision.
Define adherence in advance. For example, a record may count as compliant only when all required entry, size, stop and session checks pass. Mark unavailable evidence as unknown instead of silently counting it as a pass. Use the journal guide to distinguish the plan, transaction log and review.
Keep 1R tied to the initial monetary risk. Moving a stop after entry must not change the denominator to make the result look cleaner. A loss worse than −1R may reveal costs, a gap, slippage or an override. The record should preserve which explanation is supported.
Run a review that ends with one testable task
- Reconcile first. Match the journal against orders and account records. Missing trades or omitted costs make the rest of the review unreliable.
- Classify the difference. Separate strategy losses, execution differences, rule violations and records you cannot explain.
- Look for a repeatable failure. A recurring wrong size calculation needs a control. One losing eligible trade does not automatically need a new indicator.
- Choose the correction. Name the action, evidence to collect and next review point. “Be more disciplined” is not an operational instruction.
- Keep research separate. A change to eligibility, entry or exit creates a new strategy version. Test it separately rather than mixing old and new results.
For trade D in the invented ledger, the recorded problem is a widened stop. A concrete next task could be to document the allowed stop changes, test any supported platform restriction in simulation and log every override. That is different from assuming the original strategy would have won with a tighter stop.
The risk-system guide explains what a warning, local guardrail and broker-side control can actually enforce. An app notification is not automatically an order block.
Know when the evidence is still insufficient
A small sample can reveal a broken routine but cannot reliably rank several strategies. Splitting it into many markets, sessions and setup grades can make a fortunate subgroup look important. Write down the comparison before selecting the most attractive result.
If correct execution is still negative after costs, do not assume more willpower will fix it. Check data quality and the underlying hypothesis. If a backtest looks positive but incoming observations deteriorate, inspect fills, costs, regime changes and selection bias—not only psychology. The validation process keeps those explanations distinct.
A useful weekly conclusion may be to continue simulation, pause an unexplained process or reject an idea. It does not have to end in a new trade. There is no guaranteed relationship between a tidy journal and future income.
Questions worth answering
Does following a trading plan make the strategy profitable?
No. Adherence lets you identify which rules were actually executed; it does not create an edge. A consistently followed strategy can still lose after costs. Review both the strategy evidence and execution record.
Should I remove rule-breaking trades from my results?
Keep them in the actual account record. You may also show separate compliant and non-compliant groups for diagnosis, with the classification method stated. Presenting only the favorable group as the account's result is misleading.
How many trades prove that I have improved?
There is no universal count. Outcome variation, costs, frequency, market conditions and the size of the suspected improvement matter. A short sample can identify a recurring error without establishing future profitability.
What should a weekly review change?
It should identify a specific problem, supporting records, an owner and a next check. A broken execution routine may need correction; a strategy change requires a separate version and validation. Do not make every losing trade a reason to redesign the setup.
Sources and scope
- CME Group: Risk Management and Your Trade Plan
Primary educational questions about planned loss and account exposure.
- Bailey, Borwein, López de Prado and Zhu: The Probability of Backtest Overfitting
Research on selection among repeated tests; it does not validate this fictional ledger.
The review categories and worksheet are educational tools prepared for this page. No personal trading experience, customer outcomes or independently verified performance is claimed.