Elementary Academy · Second course created
Backtesting with AI: data, costs and validation
Learn how to ask AI for a reproducible test, check code and detect bias. A complete case with fictitious data, from protocol to report.
Consult programWhat includes the course created
16 solved exercises, 4 graphs, 10 questions assessment, 3 CSV files and a Python reference engine. You will learn how to check rows, add costs, separate periods and contrast AI responses.
Requirements: free Intraday Basean outside AI and Python 3 if you want to run the engine. No real money and no trading account. Materials written in Spanish.
The complete lessons and materials are included in Joint access to the 12 courses for 49 €, available after confirmation of payment. There's no proven cost-effective strategy: all candles and lab operations are fictitious.
Programme of the eight laboratories
- From an idea to a rule that can miss
- Requests an AI data audit
- Gather a small motor and check it out
- Modeling costs before concluding the result
- Divide time and freeze proof
- Calculates metric and draws an audited curve
- Test Sensitivity Without Following The Winner
- Deliver a report that AI cannot decorate
A free sample
Requests the AI to check costs
Six operations produce a fictitious total gross of 0.1 currency units. Each operation costs 0.2 of total fixed commission and 0.1 of slide on each side for a unit.
Calculate total costs and the net result. Show the calculations, distinguish the two sides and do not interpret the result as real market prices or evidence of profitability.
See solution
Operating cost: 0.2 and 0.1 and 0.1 = 0.4. Total costs: 6 × 0.4 = 2.4. Net: 0.1 − 2.4 = − 2.3 currency units. The positive gross disappears by including model costs.
Before continuing
Consultation free backtesting guide with AI or Retirement program of intradia.
View catalogue and offer of 49 €Sources and scope
The temporary division follows the principle described in official documentation of TimeSeriesSplit. To expand the selection and oversetting study, please consult The Probability of Backtest Overfitting, Bailey and Associates. We do not implement that estimator in this laboratory.
Original AI market material prepared with AI support. Synthetic data and reference calculations, revised with the engine included. It contains no actual quotations, predictions and individual advice.
Enter my courses with my private code · Buy the 12 courses · 49 €