What risks do we allow ourselves to take on when we analyze the analysis?

We’re treading a dangerous line between accepting a polished version of the truth instead of the full picture

Audio version of this Article/Commentary

THIS COMMENTARY IS FOR ENTERTAINMENT PURPOSES ONLY. THIS IS NOT INVESTMENT ADVICE.

The below text is a transcript (may not be word for word in all places) of this video.

Why it’s more important than ever to be analyzing the data, and not the analysis

We are at an inflection point in the economy, and the markets, and in the global geopolitical space. Each of these areas have drastic, but hard to measure impacts on the other parts of our world – including each other.

It’s why, right now, it’s incredibly important that we look at the data, and not the analysis, to determine suitability for strategy, and to determine risk in the various areas we cover. 

As an analyst, it’s hard to make this statement, but I believe it’s necessary to do so: We must focus on analyzing the base level data, and not just look at the analysis made by figureheads and vocal market leaders.

There are a few things that this applies to:

1. We need to ensure that our data sets are properly sanitized
2. We need to understand what the agenda, or the specific catalysts are, for the analyses made by different stakeholders
3. We need to look at the different outcomes and properly de-risk strategy by sourcing such opinion, but we must be careful to discount certain aspects to properly make recommendations
4. We must look at the mob mentality and the trend information to determine how that is improperly weighting movements towards specific recommendations  

Most importantly, we must ensure that we are aware of the actual data, and not just looking at what has been showcased as a relevant set of information. In a world where bulletpoints, and simple characterizations have become so commonplace, right now is the time to avoid cutting corners. The particular catalyst for this opinion piece is the rise of A.I. in the business space, and the myopic view of the markets based mostly on this catalyst. It’s a much broader world we live in, and I believe that we are doing ourselves a disservice by focusing so heavily on the soundbites, and the specificity of the commentary right now. 

As someone who slants heavily into the risk management aspects of the industries I consult on, it’s important for the reader here to understand that my opinion will lean a bit more into the over-protection arena. But it’s also important that we look at the holistic view of the impacts as a total scenario, and not just about what’s upcoming (or worse, what is potential) on a certain channel, or in a specific industry.  

Let’s take a 30,000 foot view rather than a pathway perspective, with blinders on, in other words. 

That’s what this commentary will be about.

What is this all about – what is analyzing the analysis?

A perfect example of what I’m talking about can be seen with how we have catered our own information gathering to a specific tooling, and a specific methodology. How many times have you accepted the google rich snippet (the top Search Engine Results Page for a given query) as fact? Often, I’m guessing. It’s that mentality: that once a resource is mostly trusted because of track-record, we tend to listen to it implicitly when no red flags are identified – that is dangerous.

We hear analysts talk about the market. They say they are bullish on semiconductors. Wow. What a take! You don’t say.. You are BULLISH on the component that essentially runs our entire lives from medical devices to transportation to the smartphone we are reading this commentary on? That’s a super hot take. How is one able to NOT BE BULLISH on semiconductors? What does that even mean? 

We hear soundbites, and as long as we agree with the general tone and basic information, we tend to let it drive our behavior, at least subliminally, and potentially much more than that. 

And while that’s an obvious thing, it’s not necessarily something we look at as in-depth as perhaps we should. 

Remember that telephone game where we say something in a whisper into someone’s ear, and they tell the person next to them, and then so on, and so on, and when it has gone through 15-20 people, it has completely changed the initial message. That’s an extreme example, but it’s used here to illustrate a point: we listen to what someone has produced as a public opinion, and we interpret that as a core piece to our analysis, as long as we generally trust the tone, and delivery, or the deliverer. 

This is dangerous. 

Why is relying on other analyses dangerous?

It’s a concern when redeploying assets, or changing strategy, or simply planning for the future. Now more than ever, we must be looking at base data, and making assessments on that data, not a filtered, focused (or even watered down) version based on the many different factors of the analyst covering the data. 

So in the markets, let’s say, we must be careful not to simply look at the cherry picked information that led a complex, emotional, human to derive a strategy that will do multiple things to improve their positioning. 

Some important things to remember:

• Humans who are publicly stating things, can benefit materially from adoption of their opinion (and let’s give the benefit of the doubt and state that even if we accept that the person has no nefarious intent, this is still a dangerous situation)
• A media soundbite is a tiny fraction of the data that has been evaluated, and it is not capable of being used as a full-view take on a certain position
• People are influenced by their cognitive biases

These concepts may help us to look at the general market conditions with a bit better clarity. Here are some examples of things you may be viewing improperly at the time of this writing:

A.I is seen as a great innovation, and it will be. But right now, we view A.I. as an overall, all-capable thing, given enough time. Specifically (for example, related to the markets), we see it as a solution in the stock market as being able to better make stock picks in the future. That it will be able to process more data, and make better predictions about how things move in the market. That given enough compute power, it will be able to react quicker than a human, and will somehow be able to make trading a better optimized experience. Perhaps even, that it will be able to bring back stock picking to the market because of its enhanced efficiencies. 

My take is not even that the above concepts are inaccurate, or not possible. Instead, I argue that we put too much value into A.I.’s unproven ability to do exactly what we say it can do. We’re looking at A.I. as something superhuman, because data processing is the litmus test for humans. But historical performance is still no indicator of future performance. Looking at a purely binary analysis of the ability of A.I. to be a greater stockpicker than currently exists in the market, it must also combat the same difficulties that a human stock picker does. And because the markets have not fundamentally changed, it doesn’t necessarily make sense to think that A.I. would be able to predict market changes better than any random human with some amount of data to support a position. 

We didn’t understand how things would be impacted, for instance, during the dotcom bubble. Or the financial meltdown as a result of derivatives being written on Mortgages.We didn’t know how the market would react during COVID. Not even the second wave of COVID impact on the market. Maybe we looked at historical recessions, or the tulip mania, or the Spanish Flu for inferences, but we still made mistakes. We still see inflation that is driven by market behaviors that are (completely?) unpredictable. 

We act as though A.I. will somehow be so embedded into the minds and lives of individuals that it will be able to predict mass market mentalities, and furthermore, accurate consumer behavior. This is a ludicrous take on what A.I. is – at least when dropped onto the landscape of the next 50 years. It would need such a large amount of data – something it has no access to, and likely will not anytime soon, at the micro (individual) level – in order to make such an analysis. And even then, it would have to be a quantum analysis of such data, collected at the human level, and placed upon a macro level of all humans within a spectrum; then placed upon the current and historical market factors then extant; and then placed upon the backdrop of the geopolitical currents.

And this is the most basic conceptual framework of how this A.I. might even be able to access enough information to make a prediction at all. Let alone an accurate one. Forget about the voluminous data that exists, and is created daily in the market – itself a system of informational digestion. Talk about meta. 

Also, what about:

• Individual emotions?
• Foreign powers in relation to each other and their independent roadmaps? 
• Business decisions made behind closed doors?
• Privacy legislation?
• Regulatory environment?
• Financial decisions that aren’t connected to the data streams digested by the A.I.

What will change this analyzing the analysis paradigm?

At least until we get quantum computing in place, this is an almost unrealistic idea. But it’s good to dream, right? 

And, probably most importantly, and rightly so, why are we not looking at the privacy concerns, and the protection of data, and the future regulations about where A.I. is able to gather data for digestion? This is a make or break component of this holistic analysis and automation that we are viewing A.I. as. 

So let’s look at this more realistically: 

And using a single concept: Picking stocks with A.I. (itself a myopic avenue of analysis).

We very much want A.I. to be great at stock picking, and everything else we are hearing it will be able to do, but not only for the potential to improve revenues in the future from a business offering perspective, but also from an efficiency perspective for our own workload and workflow

• We must look at things as increasingly interrelated, and interconnected. We tend to want to segmentize everything because it makes it easier to make more specific picks and define more specific strategies, when strategy in and of itself is a much broader concept
• We should be avoiding more proactively, the tendency to simply ride the waves of the current hottest things. That isn’t to say we cannot make money off of trends – that’s the best way to benefit from the market. Rather, we should be looking at how things interplay in the short-term, intermediate term, and the long-term. If for nothing else, but to avoid bubbles, and risk, and particularly, too much reliance on other people’s work or technology as a panacea
• We should be looking at the types of data we place value upon, and looking for biases in the data sets we provide to the A.I. that we employ in our business.
• We should be looking at how we manipulate the outcomes and data presentment of the findings of the A.I. we utilize. Rather than skewing our presentation of that data to fit a preconceived narrative, we ought to be looking at presenting the initial data as cleanly as possible, and rely upon the parameters of the computational structures we build that result in our final dataset.
• If we can understand how important the models and the data are that we feed the machines, we can begin to understand how to actually benefit from the computational abilities of A.I., which is really what A.I. is about. 
• We must resist the temptation to build A.I. upon potentially flawed strategies, perceived outcomes and our own agendas, in order to be able to trust the output. 
• In the case of the markets, we must resist more than ever, relying on historical performance in a continually more nuanced world, rife with trillions of macroeconomic indicators, influences and influencers, that are all acting independently from each other, and simultaneously in some form of harmony with each other. 

Now that we are already in this Google 1st result mentality, how do we get better?

And back to the original topic of this opinion piece – what we are doing now, trend-wise, is dangerous. That is: we are too accepting of a “soundbite”; ‘media clip”; “bullet point”, version of what matters to us. We must be evaluating things from a more comprehensive perspective, generally. 

How can we do that in this new age of A.I.? Stop thinking about the market movements from a trendline perspective. YES – we MUST understand why the trends move the way that they do, and how to utilize that trend information, or else we risk losing out on all the gains that are potential in the markets. Without trending information, we cannot possibly make a profit in the stock market – at least not without heavy exposure to undue risk. 

But, we must be looking at WHY something is happening, not just THAT something is happening. This can help to understand entrance and exit points.

Mostly, we need to understand how to properly discount the analysis made by other analysts, to ensure that we use their prior work as a way to segment our future research, instead of looking at their analysis as the thesis for our decision making. 

This is appropriate now more than ever, but also as an ongoing characteristic of our own analysis. For those who are on the retail end of the market space, this may not be possible. Afterall, they pay people like us to do that work for them. But as professionals employed in the position to lead, and guide and influence retail end users in the market, we have a responsibility to do better at analyzing potential benefits of a strategy, and even the underlying strategy in and of itself.