Computational Investing Part I (Coursera) - Week 8

Week 8

Module 3

3.2

• Covers back testing ideas left out of class
• How back testing works
• Only get information that would have been available at the time.
• Risks in back testing:
• The market is different now than during your test period
• "Data mining fallacy" your past test were just lucky!
• Over fitting: strategy is tuned to historical data or optimized to noise or randomness
• Components of back test:
• Data
• Definition of strategy -- thing that actually generates orders
• Market simulator -- simulates generated orders
• Analysis engine
• Be aware of market impact on your orders

3.3

• Goal: create a model -- takes input and transfers it into output or prediciton