A rehearsal is not a guarantee
One successful driving rehearsal does not establish readiness for every road. Demo trading is useful for practicing a process and discovering errors, not guaranteeing live execution. We design a small predefined experiment. Thirty observations is the assignment size, not a magical profitability threshold or automatic permission to trade real money. Data quality, dependence between observations and conditions not yet encountered require separate examination. The count is an organizational choice, not proof of an edge.
30 observations = assignment size Not proof of future returns
The experiment agreement
Before starting, record version, symbol, hours, required data and the definition of a valid observation. Include no-trade reasons and safety stops. Do not tolerate corrupt data or dangerous execution just to reach the count; stop and record why. The sample endpoint is for analysis, while a safety stop may happen earlier. A rule change midway creates a new version. Do not silently combine versions and describe them as one frozen experiment.
Version • period • valid observation Safety overrides the target count
A ten-trade arithmetic example
Use six hypothetical winners of two R and four losers of one R. Gains total twelve R and losses four, leaving eight R gross. Assume an additional cost of one tenth R per trade: total cost one R, net result seven R, average zero point seven R. This arithmetic describes only the example. Do not subtract costs twice if they are already included. A positive result here does not establish the future effectiveness of any actual strategy.
6×2R - 4×1R = 8R Costs 1R → net 7R
Examine the path
A final total does not reveal the path. Loss ordering, maximum drawdown and open positions matter. Closed-trade balance differs from current equity; floating losses can hide beneath a positive balance result. Trades sharing one event or trend are not necessarily independent. Every extra row is therefore not automatically a new independent experiment. Record both the drawdown path and the conditions producing the sample, and preserve the resulting limitations in the report.
Closed balance ≠ current equity Row count ≠ independent information
Three review decisions
Continue when data and execution are reviewable but evidence remains limited. Revise when a specific defect can be addressed in a separately tested version. Stop when the premise or controls have failed. A profitable sample may justify only more demo work. This lesson does not prescribe a universal live-trading threshold. The decision should depend on what was and was not tested, not merely whether the most recent profit number appears green.
Continue / revise / stop Sample profit is not every answer
Report what remains unknown
Report count, adherence, net outcome, drawdown and errors together. List missing observations, unseen conditions and possible broker differences separately. The next test needs a clear question. Replace the market was bad or the robot was great with reviewable observations. Label unmeasured items unknown. A useful report does not fill gaps with unsupported confidence; it identifies the evidence needed next and explains which conclusions are currently justified.
Count • adherence • net • drawdown Unknowns + next experiment
Complete the foundations
Check the example: eight R gross, one R cost, seven R net. Then design thirty demo observations with a frozen version and safety stops. These seven lessons should produce a playbook, journal and better questions, not pressure to risk real money. The paid workshop develops a market-condition map with sample data and review criteria. Complete the experiment plan before starting, and tie the decision to continue or revise to specific evidence.
Playbook + journal + review Next: a conditions map
