Algo Trading Space Backtest Manager: launch review

What the Backtest Manager is

Testing an expert advisor properly takes many backtests: every symbol it trades, several timeframes, a few set files, more than one broker’s data. MetaTrader 5’s Strategy Tester does that one run at a time, each set up by hand. The Backtest Manager is meant to turn it into a queue: you describe the tests once, the app builds every combination and runs them while you do something else.

What matters is what the app does not do, per ATS: it does not simulate. Every test runs in a real MT5 terminal with the real Strategy Tester, and the app reads the report MT5 writes, deal by deal. A result from the Backtest Manager should therefore be reproducible in your own terminal with the same settings and data. That sets it apart from generators and simulators with an engine of their own.

ATS says it has used the tool internally for a long time and quotes more than 14,000 backtests run, including those published on algotradingspace.com. The desktop app opens to the public in November 2026; until then there is a waiting list. How to read backtests and where they mislead is covered in our guide to backtesting and forward testing.

What ATS promises (vendor claims)

These features are listed on the product page. We could not run them ourselves before launch, so we report them explicitly as ATS’s statements.

• –Batch mode: pick an EA and a set file, then several symbols and timeframes; the app creates one test per combination, with modelling, latency, deposit and leverage you set once.
• –Optimisations, across many symbols too: start, step and stop per input, genetic or complete search, ranking criterion. Results as a filterable table (say profit factor above 1.2 and more than 500 trades), each pass promoted to a full backtest in one click. An MT5 optimisation cache (.opt) can be imported as well.
• –Library for EAs (.ex5) and set files, with licence keys kept with the robot, versions and variations of set files; every test records which version ran.
• –Results with return, drawdown, return to drawdown, win rate, profit factor and Sharpe, filterable and searchable; the full report with equity, deals and analysis by symbol, hour and weekday.
• –Combining: replay up to ten results on one account, with the real joint drawdown instead of a sum of separate curves, and each strategy’s contribution per month.
• –Portfolio Builder (variants from lowest drawdown to aggressive), Robustness Suite (other data, Monte Carlo) and Workflows (beta) that chain backtests, optimisations and reviews.
• –Queue to CSV: export, edit in Excel or Google Sheets and import again.
• –Equity, not just balance: the curve is rebuilt from deals and price bars so floating drawdown shows. Results are available through an API.

On the technical side ATS says: the app runs on Windows with your own MT5 installations, makes its own copies of the terminal and leaves your trading terminal untouched. Tests run on the history of the terminals you add, so on your broker’s data. Backtests run in parallel, as many as cores and memory allow; optimisations run one after another because each already uses every core.

Seen: the test library in the Account Tracker

The Backtest Manager is connected to the Account Tracker. In an account’s Backtest Comparison tab there is an Import from Backtest Manager area: it lists the tests that ran in the Backtest Manager and attaches one of them to the live account instead of uploading a report. The operator’s account had 38 tests there.

1. 11. History with saved comparisons and Run AI Analysis for a new AI review.
2. 22. Import from Backtest Manager: the list of all tests from the desktop app, paged.
3. 33. Search by name, robot or symbol.
4. 44. Test name carrying everything that describes the run: EA and version, symbol, timeframe, set file, risk, starting capital, modelling (real ticks, random delay) and a ten-year period.
5. 55. Key data: symbol, timeframe, number of trades, result and drawdown of the test.
6. 66. Import attaches the test to this account. We imported nothing.

The test names show how a proper backtest should be documented: real-tick modelling with random delay, ten years of history, fixed risk and capital, and the same strategy run separately on each symbol instead of one polished combined result. That discipline is exactly what many sales backtests lack, as we often note in our EA reviews.

Seen: backtest versus live account

The feature that reveals most about an EA is the check between test and reality. You upload a Strategy Tester report or import it from the Backtest Manager, set filters and start the analysis. Metrics, deviations and anomalies appear at once; the AI summary is an extra step.

1. 11. Backtest Comparison tab in the account view.
2. 22. History (saved reports) and Run AI Analysis.
3. 33. Import from Backtest Manager, collapsed.
4. 44. Loaded report: an HTML file from the MT5 Strategy Tester. According to the docs MT4 reports work too.
5. 55. Testergraph (.csv), optional: the tester’s balance and equity curve. Without it the curve chart stays empty later.
6. 66. Account data is pulled by the tracker from the history, here with number of trades and days. Two backtests can be compared instead.
7. 77. Filters for magic number, symbol and the account’s period.
8. 88. Normalise capital by lot size or capital. The note below says by which factor the test’s amounts are scaled.
9. 99. Run Analysis starts the check.

1. 11. Confidence: whether the sample is large enough for the metrics to be reliable.
2. 22. Periods of account and backtest in the legend.
3. 33. Match with the backtest from 0 to 100 and a label, here “Needs Attention”.
4. 44. Metric with assessment: account against backtest, below it “in line”, “better” or “worse than expected” with the deviation in percent.
5. 55. Maximum drawdown worse than expected: the most important warning in this example.
6. 66. Second row: expected payoff, trades per month, Sharpe and recovery factor.

1. 11. Equity curves on top of each other: empty in our example because no testergraph file was loaded.
2. 22. Detected anomaly with severity, here a drawdown well above the backtest’s.
3. 33. Holding time of trades in five classes, account against backtest.
4. 44. Win rate overall and split into buys and sells.
5. 55. Per symbol: win rate and profit factor in account and test.

1. 11. AI Analysis Summary: the assessment in words.
2. 22. AI label, so it is clear the text is machine-generated.
3. 33. Summary: periods, win rate, profit factor and drawdown compared.
4. 44. Recommended actions, such as reassessing after another 50 trades.

Two observations from the test. First, the software recalculates the metrics every time you open it, but the AI summary is a saved text from the day of the last analysis, in our case 2026-03-31. The account had kept trading since, and text and current metrics no longer matched (on profit factor, for example). Always read the summary together with its date and regenerate it after larger changes; that costs one AI credit, which is part of the Pro allowance.

Second, the result is honestly uncomfortable: for our sample account the check flagged a drawdown well above the backtest as a high-severity anomaly. That is exactly what the feature is for. Why EAs so often run worse live than in testing is covered in why EAs fail live.

Where it fits: backtest, demo, small live account

ATS describes an order we would recommend for any EA: backtest first, then demo, then a small live account. The Backtest Manager covers the first step, the Account Tracker records the other two, and the comparison links both once the account has enough trades. If you want to combine several EAs, per ATS you can replay up to ten tests on one account in the Backtest Manager and check later in the tracker whether the live portfolio follows the plan.

One caveat applies to any backtest tool: more tests do not make a strategy better, they only make overfitting easier. Whoever tries a hundred set files and keeps the best one has usually optimised chance. The Robustness Suite is meant to test exactly that, per ATS (other data, Monte Carlo); whether it does so well we will check after launch. More on lot size and risk per trade in our guide to position sizing.