Brennan, Richard

               Hour 1: Overview and Theory  - attached

               Hour 2: Practical Examples

  • Generating Random Data and Serially Correlated Data - Interactive Demonstration
  • How Sample Size is used to Detect the Presence of an Edge  -Interactive demonstration
  • Overview of an example workflow process - Interactive Summary
  • A 50 year walkthrough of an example Sub Portfolio - Real Time Demonstration
  • A Sample Portfolio and How it was Generated - Interactive Demonstration

A stimulating discussion is proposed whereby we look at concepts of 'noise' and 'signals' in real market data. Having this understanding then allows us to apply a quantitative approach to isolate meaningful data from noisy data to give us an edge in our trading systems

The presentation steps through important considerations to assist us to perhaps at least consider why it is essential to empirically test our models in a quantitative and systematic way to discover whether or not we can capture causal signals existing in otherwise noisy market data with our systems.

Initially the content to be discussed in this excellent series will form the basis of the introduction to noise and its impact in market data:

Part 1: https://www.youtube.com/watch?v=1z09UFOJ-G4&t=45s

Part 2: https://www.youtube.com/watch?v=Ejk7BxP263Q&t=40s

Part 3: https://www.youtube.com/watch?v=_K1FOr-jR1c&t=52s

Once members have an understanding of these broad concepts, I will the showcase the content in the following 3 videos using a real interactive process:

Part 4: Creating a random market data series and a random market data series with inserted bias  https://www.youtube.com/watch?v=dCJ_u8b8BRU&t=35s. We will also examine a generated random market data series and see how traditional indicators such as MACD, Support and Resistance, RSI and Stochastic indicators which still seem to work when applied to random market data.

Part 5: Assessing how this simulated data performs in tests undertaken with real trend following models https://www.youtube.com/watch?v=QRk0pPmr7dI&t=373s.

We will run real simple trend following models on this data to see how we can be lulled into a false sense of an edge even when we use random market data and why a large trade sample size is essential in determining whether or not an edge exists. We will then compare the performance of these models when applied to random market data and real market data to understand there is only a weak edge in real market data but this is all we need to find a profitable system in the long term provided we can strictly manage risk at all times.

Part 6: Reshuffling real market data to observe the effect on our Trend Following Models.: https://www.youtube.com/watch?v=HIUReA7fstM&t=28s. This introduces members to the concept of serially correlated data and it's importance for all trading models.

This presentation will conclude with a discuss of the important factors raised in the following excellent episode with Tim Masters. Tim worked very closely with David Aronson ("Evidence Based Technical Analysis"): https://www.youtube.com/watch?v=1RKz9v_0WDo