Tasilab

Tasilab is a simulation-based platform designed for research, education, and strategy validation on the Saudi Exchange (TASI). It enables users to conduct backtesting, machine learning experiments, and paper trading using real TASI data, with an emphasis on transparency and reproducibility. The platform is accessible via a Python SDK and does not require advanced qualifications to use, aiming to make quantitative analysis and strategy development more approachable.

Key features include the ability to run backtests on both indicator-based and machine learning-based trading strategies. Users can train scikit-learn classifiers offline and evaluate them against historical TASI bars, with out-of-sample accuracy and other metrics logged for each experiment. Tasilab supports several pure-numpy indicators and allows strategies such as threshold or moving-average-crossover, with options for stop-loss and take-profit. Every backtest and experiment is logged with parameters, metrics, and trades, and users can compare runs side by side, pin highlights, and publish results to their profiles. The platform models real-world trading conditions by incorporating actual TASI fees, VAT, slippage, and trading hours, ensuring that simulations reflect realistic constraints. All trades are simulated; no real orders are placed, and no real money is at risk.

Paper trading is available with support for both market and limit orders, executed against live Saudi Exchange quotes with 15-minute delay, and users receive Telegram alerts when orders fill. Tasilab provides up to five years of daily OHLCV data for every listed equity on the exchange, normalized for splits and dividends. An AI co-pilot is previewed, which will explain every metric in plain language and allow users to ask follow-up questions about models and results. The platform operates entirely in a simulation environment and is not licensed to provide regulated financial services or investment advice.

Tasilab is currently in a free Founding Period, during which all features are unlocked on 15-minute delayed data with no credit card required. Paid plans are planned for future release but are not yet available. The service is delivered through a Python SDK and web interface, with all experiment runs being fully reproducible and shareable by run ID.