Trend Prophets Academy: June 11, 2024

Hi reader

Welcome to another edition of the Trend Prophets Academy Newsletter!

In this edition, we look at:

  • What to look for when considering AI in investing.
  • We provide a great checklist for you.


Since we know your time is valuable, we summarize each section with its key points on top so that you can get all the important information in less than 2 minutes. Those that want to learn more and see the graphs and visuals, can continue reading further down.

Understanding AI and Machine Learning

The summary:

  • Understanding AI and ML: AI encompasses various technologies, with LLMs and ML being crucial subsets. ML allows computers to learn from data and make predictions.
  • Investor Use Cases: LLMs can analyze financial statements and earnings calls, saving investors time and effort. However, predicting stock performance with ML is highly unreliable.
  • Key Considerations: Investors should critically evaluate ML systems, asking about overfitting, backtesting, data types, retraining frequency, performance metrics, transparency, interpretability, handling anomalies, and risk management.

 

Why is this important?

  • Informed Decisions: Understanding AI and ML helps investors make better decisions by leveraging these technologies effectively.
  • Risk Awareness: Being aware of the limitations and challenges of ML models can prevent reliance on unreliable predictions and promote more cautious investing strategies.
    See below to read the full article.

 

Did you know that in the past 5 years, NASDAQ produced over 100 different signals on when to buy and sell? We did. That’s how we beat QQQ by over %125 in 5 years. Subscribe to Trend Prophets today.

The Big Picture

AI Versus Machine Learning

In this week’s edition of the Trend Prophets Academy newsletter, we are going to look at how do-it-yourself investors, or all individual investors for that matter, can incorporate AI into their portfolio management activities.

But before we can discuss this, it is important for everyone to understand what exactly AI is and what is accessible to most investors. Artificial intelligence can be subdivided into many different areas which are described by the types of data, algorithms used, and the objective of the AI model. The main issue today is that since the launch of ChatGPT, large language models (LLMs) have become synonymous with AI. Thus, when someone says AI, they immediately think of generative AI, where the AI generates content. This could be text, videos, or images.

But the AI world contains many more applications. Another term you will frequently hear is machine learning (ML). Many practitioners will debate whether machine learning is actually AI, even though LLMs are built using machine learning techniques.

Machine Learning is a subset of AI that focuses on the development of algorithms that allow computers to learn from and make predictions or decisions based on data. ML algorithms build a mathematical model based on sample data, known as “training data,” to make predictions or decisions without being explicitly programmed to perform the task.

Machine Learning is a part of Artificial Intelligence. While AI encompasses a broader spectrum of technologies that simulate human intelligence, ML specifically refers to the systems and algorithms that learn from data.

Getting back to the point of this article, how can investors make use of AI and ML?

Use Case 1: Using LLMs to Read Financial Documents and Earnings Transcripts
The first use case is centered around LLMs given how much attention they are receiving. Most do-it-yourself investors do not have the time, resources, and possibly the skill to analyze financial statements and pore through earnings call transcripts. This is a perfect task for LLMs. They can quickly (and given good prompts, of course) summarize all key points and allow investors to get a better understanding of corporate financial performance. This will, in turn, hopefully allow for more informed investment decisions.

For example, one can use ChatGPT to analyze earnings transcript calls and isolate key financial metrics, key statements made by the officers, and ask the LLM for the overall sentiment of the call in relation to a specific topic. One could ask, “What was the sentiment of the CEO during the conference call?” after uploading the earnings transcript.

Use Case 2: Predicting asset prices (the holy grail)
Everyone wants the holy grail: a system that will tell you which stocks to buy that will make you a lot of money. This has historically been the domain of hedge funds like Renaissance Technologies. But predicting stock performance is incredibly difficult and most models fail.

This has to do with how the models are created.

For example, you can peruse Medium articles daily and see hundreds where the authors have produced amazing models that beat the S&P 500 using deep learning ML models. They are using historical prices to predict future prices and using complex ML techniques to do it. This is all false. For the most part, every one of these models tends to overfit. And that is a key term everyone must understand when looking at machine learning. Overfitting means that the models are trained so well on historical data that when the model sees data it’s never seen before, they produce terrible results. So, when building machine learning models, the choice of algorithm (model), the data, and how the model is trained are crucial.

The point is, don’t believe everything you read. Predicting the markets with ML is incredibly unreliable. Sophisticated hedge funds succeed because they don’t just use historical prices to predict future prices. They will use alternative data, such as analyzing satellite imagery of how many cars are in a Walmart parking lot, or how many tankers are passing through the Suez Canal in a given period.

There are some services out there that purport to use machine learning and technical analysis. The results are mixed. The test is really time. Subscribers don’t see how the backtest was conducted, or how overfitting was dealt with. You must have faith in a black box system that is the majority of machine learning models, especially when using sophisticated methods like deep learning.

When deciding on what kind of machine learning algorithm to use for a given project, you must deal with a bias-variance and interpretability trade-off. The more complicated a model, the more you can fit the training data, but then you tend to overfit and have little idea why the model is making the predictions.

Compare this to the most basic form of machine learning, linear regression. Linear models impose a strict structure on the data. In this case, you are forcing a linear relationship on the data, which may or may not be correct. By forcing this structure, it is a biased model, and you can’t capture the full range of the variability in the data. More complex models like deep learning have very low bias since it isn’t forcing a structure on the data, and it can better model the variance of the data. However, this is why it will overfit.

This is the main challenge of any ML model. So, when someone says we have a great ML system, there are many questions that you should ask.

An AI/ML Due Diligence Checklist

  1. How do you deal with overfitting?
  2. How was the backtest conducted? You can’t simply use a machine learning model’s result. You need to run the trading strategy through data the model has never seen before and use the actual prices one would receive on the trade.
  3. What type of data do you use? Is it purely historical price data, or do you incorporate alternative data sources like sentiment analysis, macroeconomic indicators, or other non-traditional data sets?
  4. How often is the model retrained? Markets evolve, and a model that performed well in the past may not perform well in the future if it isn’t regularly updated.
  5. What metrics do you use to evaluate model performance? Beyond just returns, consider risk-adjusted metrics like Sharpe ratio, maximum drawdown, and volatility.
  6. How transparent is the model? Is it a black box, or do you have some understanding of how it makes decisions? Transparency can help in understanding the model’s strengths and limitations.
  7. What is the interpretability of the model? Can you understand and trust the model’s predictions and the reasoning behind them?
  8. How do you handle market anomalies or black swan events? It’s crucial to understand how the model responds to unexpected market conditions.
  9. What is your risk management strategy? Even the best models can fail, so having a solid risk management plan is essential to mitigate potential losses.


There are some good services out there that are delivering results. The more successful ones are more centered around picking companies based on financial metrics, such as finding the best companies with low debt, high return on equity, and consistent earnings growth. I have also seen some good managers that use machine learning to improve portfolio construction. I have also seen amazing results on the fixed income side. However, the question comes down to if these services are available to individual investors or just the pros. And of course, the cost is a huge detail. If you are using a service that charges $1,000 a month, you might need to invest a significant amount of capital so that your fees don’t represent more than 2% of the committed capital.

How Does Trend Prophets Use Machine Learning?
The Trend Prophets system is driven by signal processing: filtering out the market’s short-term volatility to isolate the mid to long-term trend. We use machine learning to help determine the optimal parameters of the signal processing technique and to detect the volatility regime that dictates trading frequency. We aren’t trying to predict where the market is going; we are identifying entry and exit points for various ETFs which are designed to participate in the upside and protect on the downside.

Over the past five years, we have identified close to one hundred entry and exit points for NASDAQ, which would be otherwise undetectable without our system. As a result, we are up a total of 275% instead of just 156% had you not followed our signals.

Trend Prophets Strategy Performance (as of 2024-06-07)

Table 1: Trend Prophets top 5 performing strategies. Source: Trend Prophets, EODHD. Please see our website for all performance information. Inception dates for all strategies is 2010-03-10 except for MARA which starts in 2017.

Don’t wait. Subscribe today and see what we can do for you.

That’s it for this edition of the Trend Prophets newsletter! Please contact us at info@trendpophets.com for any questions.

Cordell L. Tanny, CFA, FRM, FDP
President & Founder

Join the Trend

Disclaimers: Past performance is no guarantee of future results. This newsletter should not be considered as investment advice and is intended for information purposes only. Please see our Terms and Conditions for all disclaimers.