Research Article #36 - Fitting Signals to Random Data
Research Article #36 - Fitting Signals to Random Data
The Dangers of Parameter Fitting: A Practical Examination Using Random Data
👋 Hey there, Pedma here! Welcome to the 🔒 exclusive subscriber edition 🔒 of Trading Research Hub’s Newsletter. Each week, I release a new research article with a trading strategy, its code, and much more.
Join over 4K+ readers of the newsletter!
First and foremost, I want to say thank you for all the support you’ve been giving me during the initial stages of this project.
If you have specific requests of content, make sure you send me a DM and I’ll try to get to them.
Today I want to write about a fun topic. Generating trading signals on completely random data!
This is important because there’s so much randomness in financial data, how can we know that what we’re seeing is real?
Take a quick glance at that chart.
Do you see patterns? Support and resistance? What about that big double bottom?
What if I told you that this data is completely random?
How to navigate a business where what we see, can be random? And what steps we can take to avoid it?
This is what I will try to make really clear today.
Also we’ll touch lightly on the topic of fitting signals to a dataset.
A lot of quants, that have a more statistical or math focused background, like the idea of using machine learning models to fit the best rules to a dataset.
And the core idea behind it makes sense, from an efficiency point of view.
Imagine trying to find which rule works best by hand, in today’s day and age.
Why not just feed a bunch of variations into an algo, and maybe even allow that algo to go wild and search for the best variations with no limitation?
Looking through thousands of combinations a second?
Much more efficient ain’t it? Even myself, without an engineering background, have developed such an algo to do massive optimizations.
The problem is that, as we’ll see through the article, this is not a good idea.
We need something more substantial, about what we’re trying to capture, to give us better odds of avoiding the randomness in the data.
Let’s get into it!
Index
Introduction
Index
Generating Random Data
Analyzing the Data
Finding a Profitable Signal
Python Code
Generating Random Data
To test our model, first we need a process to generate the random data.