Train bespoke ML models to generate custom signals for your strategies

The ML Quant Lab is TrendSpider's platform for training custom machine learning models that generate trading signals. You choose a prediction target (for example, whether an asset will close higher over the next N bars), define input features using TrendSpider's market data (technical indicators, fundamentals, custom formulas, price action, custom indicator outputs, and alternative data), and select a machine learning model type. If you are not sure which inputs to use, the Lab's built-in LLM prompt box can suggest and configure features for you. Trained models can drive scans, alerts, rules-based strategies, or automated Strategy Bots.

No. The ML Quant Lab is built for traders without a computer science background. You configure training runs through point-and-click menus, pick input features and a model type from built-in options, and let the ML Quant Lab handle the training and validation. Coding is never required at any stage.

The ML Quant Lab supports several standard machine learning model types out of the box, including Random Forest, K-Nearest Neighbors (KNN), Naive Bayes, and Logistic Regression. Each suits a different kind of prediction problem: Random Forest handles non-linear relationships well, KNN is useful for pattern-similarity matching, Naive Bayes is fast for basic classification tasks, and Logistic Regression is interpretable for probability estimates. You can train multiple models on the same prediction target and compare their performance.

You can train models on technical indicators, corporate fundamentals, price-action features, custom formulas and computations, custom indicator outputs, and the alternative data available elsewhere in TrendSpider. Because the lab uses the same data layer as the rest of the platform, almost anything you can put on a chart also available as a model input, which lets you engineer feature sets that go beyond what pre-packaged ML trading products expose. Please note: certain asset classes cannot be used for training models due to compliance restrictions. Please contact TrendSpider support if you have questions about which asset classes can be utilized in the ML Quant Lab.

A trained model outputs signals you can apply across TrendSpider: paint them on charts, scan the market with them, use them as conditions inside rules-based strategies, trigger cloud-based alerts, and forward-test or back-test them before committing real capital. You can also deploy the model as a fully automated Strategy Bot that places trades on your behalf through a supported broker integration via webhook, or runs in alert-only mode and notifies you via Discord, your phone, or email.

Yes. You can validate any ML strategy by running historical backtests, then forward-test it live by deploying it as a Strategy Bot. Backtests return the same performance signals you get for rules-based strategies (win rate, drawdown, equity curve), so you can compare ML and rules-based approaches apples-to-apples before committing real capital.

They solve different problems. Sidekick AI is a conversational assistant you chat with to analyze charts, retrieve data, build scans or alerts, and get insights in real time, with the same data access as the platform itself. The ML Quant Lab trains custom machine learning models on historical data to generate predictive trading signals you then use in charts, scans, alerts, or bots. In short: Sidekick helps you think and research; the ML Quant Lab builds ML models that produce signals for you.

Cross-breeding is an advanced feature that combines multiple trained models and retrains the result, producing a new strategy with reduced bias and stronger performance than any of the individual models would produce on their own. It is part of how the ML Quant Lab helps traders find progressively better models, rather than settling for the first model that trains successfully.