Using technical data and AI to enrich Fundamental Investing part 1
This article is the first in a series of articles that will explain how to analyze financial data in new, meaningful ways and how the assistance of AI can play a crucial role in your decision-making process and portfolio management. We’ll go over many topics, starting with considering technical data when deciding the best strategy for managing your portfolio.
For many fundamental investors, relying heavily on technical data is considered a big mistake. So the first point I want to get across is essential and can be a summary for the entire article: Technical data can provide crucial fundamental insights into market behavior.
This is especially true when you dive deep into the behaviors and relationships between different technical data sources and analyze their correlation. In fact, correlation analysis and examining how specific technical data sources behave together is one of the essential building blocks in constructing a correct portfolio. Looking in the right places can give you insight and understanding of the current market state and exactly where we are heading.
One of the most critical skills in portfolio management is knowing when to hedge. Successful timing for correct hedging can be a fantastic tool to avoid potential market crises. This is precisely why the accurate assessment of the market state is so critical and why leveraging insights from technical data can be an essential tool in developing a proper investment strategy.
We’ll review several examples of insights derived from the correlation of different technical data entities.
Bonds- Equity correlation -
Let’s explore the relationship and collective behavior of government bonds and equities (for example, TLT vs. SPY) in different market states. Government bonds can provide deep insights into the current market state since they are the main run-to-safety asset in extreme market conditions. We can see the correlation of these assets with the market as time progresses around -0.3. Still, in Graph 1: when the market reacts to an incoming financial crisis (see 2008 and 2020), we can see that there’s an extreme inverse correlation between bonds(TLT) and equity (SPY), reaching a correlation of around -0.8.
Graph 1-
But what about 2022? Even though it can be said that there’s a financial crisis going on right now, we don’t see the same behavior as we’ve seen in 2008 or 2020. That’s because there’s one exception to this rule. When a financial crisis is driven by inflation, we will not see an extreme inverse correlation between the two.
So, let’s review the current cheatsheet for this insight:
To get a clearer picture of how the market views a downturn, we also need to examine corporate bonds (LQD red line). Graph 2 examines the correlation between corporate and government bonds as time progresses. As we can see, corporate and government bonds are highly correlated most of the time. But, in extreme market conditions, something happens, and the correlation becomes negative(strong signal) or still positive but with a big difference in the relative return(weaker signal). So, adding this rule into consideration, we can almost say with certainty that when all three conditions apply, the market views the current downturn as a crisis.
Graph 2-
In past crises, there was a significant change in the collective behavior of corporate and government bonds; Although this is a purely technical observation, the rule itself has a fundamental reason: when the economy is under extreme stress driven by a financial crisis, investors view corporate bonds as a risky investment. The probability of corporations defaulting rises exponentially. That, in turn, causes large amounts of money to flow from corporate bonds into government bonds due to the lower likelihood of government insolvency. This kind of logic doesn’t apply to the general state of the market. Generally, we expect investors to invest mainly in corporate bonds due to their higher yield.
This adds another rule to our cheat sheet:
FinityX cheat sheet-
Insights from Commodity (DBC) state on inflation -
Commodity prices are less informative about the general market state, but they offer some information about inflation, especially when you consider them with bond prices. Looking at available commodity prices, There could be two main reasons why they should rise:
- Inflation
- The growth of the economy
A good assessment of the situation as commodity prices rise will help the investor to adjust their portfolio accordingly. Graph 3 examines commodity prices as time progresses. We can see that there was significant growth in commodity prices in 2006–2007 and 2021–2022.
Still, examining commodities prices alone can’t help us to determine the reason for the steady growth. Is it because of global inflation or because of rising GDP? To answer this question, we need to examine bond prices as well. Graph 4 looks at both commodity and bond prices.
Examining the graph, we see that in 2006–2007 both commodities and bond prices went up. It suggests that the main reason for the rise in commodity prices wasn’t inflation but the expansion in the economy. On the other hand, we can see that in 2021–2022 commodity prices went up while bond prices went down- which means that the reason for the rise of commodity prices was inflation. It is extraordinary. This is the strength of the relative macro analysis. It helps you extract fundamental facts about the state of the market using the combined technical information of several key ETFs.
This adds another rule to our cheat sheet:
FinityX cheat sheet-
So up until now, we examined several key ETFs that added more insight into the current state of the market, but we still need to talk about the VIX index.
VIX-
The VIX index has the potential to be the most informative index when it comes to the market state assessment, but it is also the most tricky of them all. But what’s wrong? Historically, we know that the VIX index goes way up whenever there’s stress in the market, so- we should look out for a steep rise in the VIX index and hedge accordingly. What you should know is that most of the time, looking for a potential crisis by examining the steep incline of the VIX index will result in a false alarm. The market will usually reach ATH a few months later, so using the VIX index as a stand-alone risk metric can significantly hurt your return.
Let’s say you want to hedge your portfolio whenever the VIX reaches a certain threshold, and as we’ll see, historically, this threshold should optimally be around 32. Up to that exact point, your portfolio has already suffered significant losses. That’s because the market has already suffered significant losses without you doing any hedging. Starting to hedge at that point might recover all losses if you are, in fact, right about the market being in a crisis. But historically — it’s entirely possible that the market will recover after the VIX reaches this predefined threshold. In fact, when it does recover — that recovery is usually significant and fast. In these cases, you won’t be able to earn back your losses. This is the first main point to know when examining the VIX. A steep rise is necessary for market crises but is insufficient.
For instance, this kind of strategy historically failed when deployed in 2011,2013,2018, instance. So how can we still leverage the technical data of the Vix index to our advantage? Again, we must consider it with accurate information about other asset classes. This will be explained in the next section.
Looking at the graph and examining the different values, you can’t use a predefined threshold that will get consistent and reliable results as a condition for hedging your portfolio.
Insights from analyzing the “yield curve.”
The term “Yield Curve” refers to the relationship between the short-term and long-term interest rates of fixed-income securities issued by the U.S. Treasury. Under normal circumstances, long-term interest rates exceed short-term interest rates. So when short-term interest rates exceed long-term rates, we’ll define the yield curve as “inverted.” Under normal circumstances, the yield curve is not inverted since long-term debt typically carries more liability and risk, which translates to higher interest rates than short-term debt.
So when can we be in a state where the yield curve is inverted? Let’s follow the definition and think — does it make sense that short-term debt is riskier than long-term debt? When the market predicts a recession, it makes a lot of sense. Think about it — long-term securities are still risky, but a possible recession will not have a long-term effect compared to short-term securities that may mature while a recession is still happening.
Let’s look at the historical yield curve. As you can see in the chart below, the marked gray zones are times of market recession, the red lines are times when the yield curve was inverted, and the overall graph shows us the yield curve over time. We can see that the yield curve is inverted right before these gray zones. It happens in months and behaves precisely as we expect it to.
So, let’s add another rule to our cheat sheet:
FinityX cheat sheet-
As we’ve demonstrated, technical data can hold significant fundamental conclusions about the current and future market state. Many people mistakenly think that technical data is useless because they can’t see meaningful patterns in the technical data of singular stocks. As we’ve shown, examining joint behaviors of specific ETFs can provide significant insights into the market state and how the market views different economic events. Correlation analysis of technical data does add complexity, but it also adds depth and meaning. This is why we believe that making intelligent and informed decisions must include the proper technical analysis in any investment methodology.
One important thing to note is that this article only demonstrated the power of technical analysis on the collective behavior of only 2–3 asset classes together. But, that alone brought a lot of complexity to our analysis. Imagine how robust technical analysis can be if we incorporate more asset classes and look for more subtle changes in their collective behavior. The thing is that the complexity of the analysis rises exponentially with the number of assets. It’s simply not practical. This is where Machine Learning comes into play.
We at FinityX have developed a financial AI infrastructure for developing and researching deep learning models, capable of analyzing the technical data of thousands of assets and output quality predictions for active trading. This infrastructure has several key features that make it successful:
- Generic design: It allows us to train models with different types of neural net architectures and train on any subset of desired stocks.
- Reliability: This infrastructure features a robust method of reliable backtesting, which in turn enables us to assess the theoretical performance of any model ready for deployment
- Diversity: with highly customizable configurations, we’re able to train many diverse models, each with its strengths and weaknesses — creating consistent results when combined into one single portfolio
- End-to-end training: unlike many designs aimed at training a model to output stock price predictions or only to predict some specific insight that could be beneficial, our models are trained end-to-end. This means that the model is trained to provide the finished portfolio without human intervention while optimizing unique loss functions to maximize the profit and stability of a final product — a daily/weekly/monthly portfolio.
This design produced reliable financial models with a proven track record.
Explaining exactly how we train our models is a subject for a set of articles that will come out in the future. Still, in the meantime, you can get a sneak peek if you read the article written by our CTO on defining a reliable backtest methodology (https://medium.com/@shiloabram/algo-trading-backtest-shenanigans-part-1-ecd0fec1f317).
But back to our subject: we wanted to build a specific model that encompasses all of the knowledge mentioned above while enhancing it with a relative analysis of all data points together, instead of only looking at different pairs. This led to the development of what we call the "FinityX state machine," a unique proprietary ML solution specifically designed to tackle this problem which will be the subject of my next article.
This article is a part of a series of articles on various subjects and fields by the FinityX’s team.
We believe that sharing knowledge and helping others is a part of our essence and our ability to thrive in a world full of investment (but not only) opportunities.
If you like this, please feel free to look for more content from our team on our Linkedin page: https://www.linkedin.com/company/finityx-ltd/
And please feel free to ask me any question and discuss those matters with me: