Technique · Instructions
Strategy Tester: MT4 & MT5 Step by Step
Set up a test reproducibly, test its assumptions and correctly limit historical results.
On this page
First formulate the test questionMT4: Preparing a single testMT5: Choosing data and environment consciouslyHow price modelling changes a testReading the reportOptimisation and separate control periodsTest protocol and practical limitSourcesFirst formulate the test question
Do you want to check whether the program correctly executes the described rules or whether the strategy would be economically viable under certain assumptions? These are different questions. Our start: a short test question, a manageable period of time and an unchanged initial configuration.
Before starting, write down program version, platform build, icon, data source, period, time frame, parameters, starting capital, account currency, account model, and cost assumptions. Subsequently selected profit results are harder to judge if the original test question is missing.
MT4: Preparing a single test
- Open Strategy Tester via View.
- Select the installed and compiled EA program.
- Set icon, time frame and test period.
- Choose model: Each tick, checkpoints or only opening prices, according to the offered settings.
- Open expert properties; control initial capital and input parameters. Turn off optimization for the first run.
- Record spread assumption and data base, then start.
- Check results, report and journal; use visualization when necessary.
A model with few price points can produce different results for strategies trading within a candle than a more detailed history. „Every tick in MT4 is not equivalent to the assurance of a complete real tick history.
MT5: Choosing data and environment consciously
- Open Strategy Tester and first select a single test.
- Set EA, Symbol, Period and Parameters.
- Select the offered tick mode and document the data reference.
- Check start-up capital, account currency, leverage and required instrument data.
- Control existing extended cost, account and delay assumptions instead of adopting their default unchecked.
- Start the test; examine the result and journal after the data load.
When testing several instruments, their data and trading conditions must also fit. A seemingly fast calculation can be based on a simplified model.
How price modelling changes a test
| Model | Significance for interpretation |
|---|---|
| Real Ticks in MT5 | Uses available historical tick data; further examine data gaps and model assumptions |
| Generated ticks | The intermediate progression is modeled from existing data |
| 1-minute OHLC | Few bases per minute; no complete observed tick sequence |
| Opening prices only | Highly reduced progression; separately problematic for intrabar-dependent rules |
The MT5 help describes differences in triggering and execution between simplified and detailed test modes. Therefore, a good result in the rough mode must not be considered unchecked as confirmation of real stop execution.
[FEATURES-TEST5]Editorial counter-test: compare the same unchanged EA on the same period with two suitable models. First, examine individual deviant trades. Do not immediately choose new parameters to compensate the difference cosmetically.
Reading the report
Tester reports provide, among other things, results and decline indicators. Read their definitions and separate balance, equity, number of trades and costs. The journal helps to detect faulty or incomplete runs.
[TESTREPORT5]Our evaluation questions: How many independent situations have been tested at all? Does a single big winner carry the result? How low was the account in the meantime? Are open positions taken into account at the end of the test? Are the costs consistent with the intended use?
A self-selected example strategy with only three trades allows for less robust statements than a broad observation over different conditions. A higher historical profit alone does not say which variant would be more robust in the future.
Optimisation and separate control periods
In an optimization, several parameter combinations are tried out. Our method: define reasonable areas in advance, use a development period and withhold a later control period. Each new selection based on this control period makes it a bit of a development data set.
Don’t just look at the best individual result: Are similar parameters in a comparable range of results or does the strategy break down if there is a small change? Subsequently observe an unchanged configuration in demo forward operation. This is an editorial test method, not a guarantee.
Test protocol and practical limit
| To be archived | Why? |
|---|---|
| Program and parameters | So that the same variant can be retrieved |
| Data model and period | So that the result remains comparable |
| Costs and contract assumptions | So that the economic basis is recognizable |
| Report, journal and selected individual operations | So that the key figure and the actual process fit together |
| Separate forward observation | So that later experience is not optimized into the old history |
A tester reproduces a model of the environment. Real network disruptions, actual brokerage decisions, changing liquidity and future quotes are not fully predicted. Subsequently for operational preparation, EA Guide and Contract specification use them.
Sources & scope
Check the evidence.
- MT4 · Tester settings
Concrete MT4 tester aid. Advertising blocks for the new platform may not be read as MT4 functions.
Open the source ↗ - MT5 · Strategy Tester
Test modes and multi-symbol tests; data and assumptions determine the statement.
Open the source ↗ - MT5 · Real and Generated Ticks
Data Models and Limitations of Tick Reconstruction.
Open the source ↗ - MT5 · Tester specifications
Triggering and price processing are dependent on the test model and instrument.
Open the source ↗ - MT5 · Tester report
Test indicators are to be read within their model assumptions.
Open the source ↗