XTester 0.0.70: A More Reliable Path from Idea to Validated Strategy
The new XTester release helps the AI lab reject non-working candidates earlier and return to a version supported by real trades and fresh out-of-sample evidence.
Each product in the EasyTrading ecosystem has a distinct role. XTester evaluates a trading idea on historical and unseen data. CopyTrader handles execution of selected logic. TradeStat is intended to bring the resulting performance into one statistics service.
The quality of that pipeline therefore starts with research. The first stage must distinguish real strategy progress from a candidate that looks plausible in code but never trades or relies on invalid evidence.
XTester 0.0.70 has now been released, with a primary focus on AI Research Lab reliability.
How AI research can enter a dead end
The AI lab creates several strategy versions, tests them, and develops the most promising directions. A version that is formally valid, however, is not always operational.
It may:
- never reach its first entry condition;
- use future information;
- calculate risk incorrectly;
- omit a protective exit;
- report an expected risk stop as if it were a software failure.
If such a candidate becomes the base version, later experiments inherit the same problem. The research continues without moving closer to a deployable strategy.
What XTester 0.0.70 changes
The new release checks candidates before a full backtest and evaluates more than profit or fitness. Research Lab looks for real signals, orders, and trades, verifies hard requirements, and requires fresh Walk-Forward or Out-of-Sample evidence.
When a branch stops producing useful results, the lab can return to the objectively strongest working version and continue improving its weak point, such as drawdown.
At a plateau, the lab now waits for the user’s decision. Direct guidance can be carried into the next hypothesis, while normal chat messages should no longer be lost.
Research is easier to inspect
Versions can be filtered into working and failed groups, viewed as a tree or sortable table, and compared by key metrics. This matters beyond automation: the user can see why one version became the base and another was stopped.
Developers also get a line-by-line Git diff preview, making it easier to review exactly what changed between strategy versions without leaving XTester.
What this means for a trader
The release does not claim that AI can automatically discover a profitable strategy. It addresses a more fundamental problem: reducing runs that extend a technically invalid candidate and making the selection process easier to understand.
That matters before live execution. A strategy should do more than look profitable on familiar history. It needs stable behavior on unseen data, explicit risk limits, and explainable signal execution.
XTester 0.0.70 is available for Windows. The release passed 478 automated tests and is distributed through the installer, portable build, and automatic update channel.
Historical testing cannot guarantee future performance. Always evaluate drawdown, out-of-sample stability, and live execution risk before deploying a strategy.