How we test
AlgoToolbox publishes MT5 test reports and documentation for our own builds. This page explains exactly how we test MetaTrader 5 indicators and Expert Advisors, how a score is calculated, how we get each copy, and what we will never do. If a report does not follow these rules, tell us and we will fix it.
The performance results from our own tests come from the MetaTrader 5 Strategy Tester, not live or forward trading. We don’t sell signals or manage accounts. We show what each backtest can—and cannot—tell you.
Two kinds of pages
- Test reports cover tools made by other developers, free or commercial. They can carry a score out of 5.
- Our builds are tools we coded ourselves: MQL5 ports of published scripts and our own original tools. Their pages state which checks were actually performed. A visual or functional check is not a strategy-performance backtest. Our builds are never scored, because a site cannot fairly rate its own work.
How we get each copy
- Commercial tools: we download the free demo version from the developer’s official MQL5 Market page. MQL5 Market demos only run in the MetaTrader 5 Strategy Tester, so the performance results from our own tests of these demos are backtests, not live or forward trading. Developer claims and cited external results are labelled separately.
- Free tools: we download the free version from the developer’s official page (MQL5 CodeBase, the developer’s website or repository).
- We do not buy licences, accept free licences or accept payment from developers, and developers do not see or approve a report before it is published.
- We never host, sell or redistribute commercial software. Every Test report links to the developer’s official page.
The test bench
Each performance report states its test setup, including these details where available, so you can attempt to reproduce the run:
- Platform and version tested
- Symbol and timeframe (the developer’s recommended pair first)
- Test period, with an out-of-sample window when one is available
- Requested data mode and actual tick-data limitations where the report or journal lets us establish them
- Broker account type, deposit, leverage, spread, commission and execution assumptions where available
We start with the default settings or the developer’s own set files, and say which. When possible, we add a robustness run (another period, a wider spread or a developer set file). Each report lists exactly which runs were done.
A requested real-tick mode does not prove that the tester accepted real ticks throughout the run. Where the evidence shows generated ticks, discarded data or missing cost inputs, we disclose that limitation. We do not call conditions realistic merely because a test completed.
Indicator and utility pages distinguish visual or functional checks from strategy-performance backtests. If a performance test was not run, we do not imply one through a badge, score or illustration.
How the score works
A Test report score is the average of five criteria, each rated from 1 to 5 in half steps, rounded to the nearest half. The breakdown is shown on every scored report.
| Criterion | What we check | 1 means | 5 means |
|---|---|---|---|
| Results | Strategy Tester results on the default settings: profit factor, drawdown against profit, recovery. Indicators: how the signals performed in the test. | Loses money or needs perfect conditions | Profitable with a drawdown that is small compared with the profit |
| Robustness | Does the result hold beyond the first run, in the robustness runs we did (another period, a wider spread, another set file or symbol)? Fewer robustness runs means a lower score here. | Works only on the vendor’s preset period | Holds across symbols, periods and spread |
| Risk control | Stop loss, position sizing, exposure caps; no hidden grid or martingale. Indicators: no repainting, no look-ahead. | Martingale or grid with no hard stop, or repainting signals | Fixed risk per trade, hard stops, signals that never change after the bar closes |
| Claims vs results | Do the listing, screenshots and description match what we measured? | Claims we could not reproduce at all | Everything we tested matches the listing |
| Usability | Install, inputs, documentation, tester speed and CPU load. | Undocumented inputs, crashes or freezes the tester | Clear inputs and manual, runs smoothly |
A tool can be published as “Not scored” when the demo is too limited to judge it fairly. A score describes our test, not a promise of future profit.
We publish bad results too
If a tool fails our test, the report says so. We do not delete or soften reports because a developer asks us to. Developers can send corrections of facts (a wrong version number, a missed input) and we will update the report and its “Updated” date.
Affiliate links
Some official-page links may be affiliate links. When one is, the button is marked as sponsored and the report says so next to it. A commission never changes the test, the score or the verdict, and we do not accept paid placements. The plain broker reference on product pages is a separate, clearly labelled affiliate link.
Limits of our testing
- Backtests simulate trades on historical data. They are not live or forward trading; slippage, requotes and broker execution can make live results worse.
- A tester run does not verify live broker execution or an external signal, news or AI service that the test cannot reproduce. Each report distinguishes what we tested from what we could not verify.
- Past performance never guarantees future results. Always test on a demo account first.
- Read our Risk Disclaimer before using any tool on a live account.
Questions about a test? Contact us.