Identify what produced an EA result, compare its assumptions and record the evidence still missing before you evaluate a trading robot.
A trading robot shows a rising equity curve. Before asking how much it made, ask how that curve was produced. A historical simulation, a reconstructed portfolio and a funded account record answer different questions.
This guide gives you a practical sequence for checking Expert Advisor (EA) results. Start with the evidence label, inspect the setup, and finish with a list of conclusions the record actually supports.
1. Identify the evidence type
Keep the following categories separate when saving or comparing results. Forward testing describes testing prospectively on new observations; it can occur on a demo or a funded account. It does not automatically mean live trading.
| Evidence | What it describes | What to check |
|---|---|---|
| Backtest | Rules simulated on historical data | Data, code version, settings, costs and execution assumptions |
| Portfolio reconstruction | Separate results combined into an estimated portfolio | Time alignment, weighting, shared capital and omitted interactions |
| Modelled study | Alternative outcomes generated under model assumptions | Method, inputs and sensitivity to those assumptions |
| Demo account | Prospective operation in a simulated trading environment | Signal behavior and differences from funded execution |
| Live account | Trades recorded in a funded account | Verification, full history, cash flows and account configuration |
2. Save the test conditions with the chart
Record the dates, symbols, timeframes, broker data, starting capital and EA settings. Add spread, commission, swap and execution assumptions where applicable. Identify known gaps. Without these details, a second person may be unable to reproduce or interpret the result.
Also establish what the vertical axis represents. Balance usually reflects closed transactions and cash flows; equity includes the current value of open positions. A balance curve can therefore leave open-position losses less visible. A reconstructed curve needs its own calculation method.
3. Read return and drawdown together
A return total does not describe the route taken to reach it. Inspect drawdown depth, time spent below previous peaks, recovery periods, trade count and concentration in a few trades or market periods. State whether a drawdown measure uses balance or equity.
Illustrative example: a report covering five historical years and a live account covering three months have different observation windows. Even if their charts look similar, the live account does not demonstrate that five years of simulated trades occurred. Label and assess each record separately.
4. Ask what changes outside the original setup
Investigate sensitivity to costs, execution conditions and parameter choices. Separate the data used to choose settings from data reserved for evaluation. If many alternatives were tried, record that search process; the winning historical configuration alone does not reveal how it was selected.
For a worked testing reference, see MetaQuotes' guide to testing a trading robot before buying. It demonstrates practical checks on an example EA. Its screenshots and platform details are dated, so verify the controls in your current terminal before following them.
5. Make the evidence request explicit
When a presentation leaves a question unanswered, request the missing item rather than filling the gap with an assumption. Use this checklist when reviewing a report or discussing it with its author:
- What produced this record, and what period does it cover?
- Which code version, parameters and account settings were used?
- Does it include open-position equity, costs and cash flows?
- What data was used to develop or select the strategy?
- What changed in an independent or prospective evaluation?
- Can a claimed live record be opened and checked independently?
What to do next
A useful review ends with an evidence label and a clear account of what remains uncertain. A strong historical result can justify further research; it cannot guarantee future returns. Use the portfolio checklist before combining individual systems.
