Blind test results: 304,705 Nebula systems on sealed price history

Nebula's blind test results are a public record of how trading systems built by Nebula performed on price data they never saw while they were being built. As of 22 September 2026 it covers 304,705 systems. 45.0% of them held up on their sealed window. More than half did not.

We publish this because a backtest on the same history a system was built on proves very little. The useful question for any EA software, including ours, is what happens on data the system has never seen.

The results

The grade is the one each system was given during its search, before its sealed window was opened. The higher the grade, the more often the system held up, which is what a grade is for.

How the test works

1. The data is split before the search starts. Each run divides the price history into an in-sample window (the strategy is built on it), an out-of-sample window (used to check it during the search) and a sealed, blind window: the most recent stretch, locked away for the whole search.
2. Systems are built and graded without the sealed window. Nebula builds candidate strategies, tests them in-sample and out-of-sample, and grades the ones it keeps. Nothing in the sealed window can influence any of that.
3. The seal is opened once. Each finished system is then run on its sealed window, a single time.
4. "Held up" means a profit factor above 1 on the sealed window. Profit factor is gross profit divided by gross loss over the trades in that window. Above 1 means the trades on unseen data made more than they lost, in the backtest.

Where the numbers come from

The results are pooled from Nebula installs that share anonymised test outcomes with the Nebula network. New installs agree to sharing during setup, and existing installs accepted it with an update; it can be switched off at any time in one click. No names, account details, trades or profit and loss ever leave a member's machine: only anonymised test outcomes.

What these numbers do and do not show

• They are results of back-testing and blind tests on historical price data, not live trading.
• Holding up on one sealed window is evidence, not a guarantee. A system can pass its sealed window and still fail live, because markets change and live costs differ.
• They describe systems built by Nebula across many users, instruments and settings. They do not predict how any one system, or your account, will perform.
• More than half of the systems did not hold up. That is the point of the test: it removes them before anyone trades them.

Check it yourself

Every system Nebula gives you shows its in-sample, out-of-sample and blind results, and the parity check gives you the exact MetaTrader Strategy Tester settings so you can reproduce its backtest on your own broker. See how Nebula works or download it free.

Weighing up EA software, ours or anyone's? Our guide to the best EA software and how to choose it lists the seven questions to ask before you buy.

Nebula network results, 22 September 2026. Figures are from back-testing and blind tests on historical data, not live trading. Screens show backtests and simulations. Past performance is not a reliable indicator of future results. Trading involves risk to your capital.