To Understand an Investment Approach, Ask Three Questions

Hoppy placed four cards on the table:

value
long term
technical analysis
quantitative research

Then he carefully drew a tournament bracket.

Value would face long term in the first round.

Technical analysis would face quantitative research in the second.

The winners would meet in the final to decide the “best investment school.”

Dr. Hop watched for a moment, then asked:

Hoppy picked up his pen, then put it down again.

All four cards could describe the same person.

Making those labels fight is a bit like asking “Where are you going?”, “How long will the trip take?”, and “How will you get there?” to compete against one another. They all matter, but they answer different questions.

So when we meet an investment approach, we do not need to start by asking which school it belongs to.

Three questions will do:

Why might it work?
How long is it prepared to wait?
How will it find and check evidence?

Those three questions form the map we will keep using throughout the course.

Question One: Why Might It Work?

Suppose someone tells Hoppy:

I buy stocks after their prices have fallen a lot.

That tells us what the person watches. It does not yet tell us why the approach might work.

A big fall may mean the market became too pessimistic. It may mean the company genuinely got worse. Both may be true at the same time.

So we need one more question:

Why do I think this is more than a lucky guess?

The slightly more formal name for this is the source of a possible advantage.

Hoppy asked, “If I can come up with a reason for my idea, have I found the advantage?”

Not yet. A reason is only the beginning of the research.

Some approaches may appear to earn more because holding them is unpleasant. Their prices swing harder, they may be difficult to sell, and they can suffer especially badly when the whole market is in trouble. If extra returns really exist, they may be compensation for bearing risks that other people would rather avoid.

Sometimes an opportunity may come from the market becoming too excited or too pessimistic, or from people taking time to understand new information. But that idea cannot be shortened to “Everyone else is irrational, and I am the only one thinking clearly.” Without specific conditions and evidence, “the market is wrong” is just a very convenient way to comfort ourselves.

Another researcher may have no secret information at all. They may simply read public material more carefully. Where someone else stops at “demand is growing,” they continue checking competition, costs, production capacity, and whether the share price already reflects that growth.

Some trades do not happen because anyone suddenly changed their opinion. A fund may need to follow portfolio rules. An index may rebalance. An institution may trade because of redemptions, risk limits, or an urgent need for cash. When someone has to trade, they may leave behind a pattern worth studying. That does not mean everyone watching from the sidelines can automatically profit from it.

Finally, two people can discover the same pattern and still get different results. One changes the rules every few days. The other follows a process and honestly includes fees, slippage, and data errors. An advantage may come not only from what someone discovers, but also from whether they can carry it out consistently.

There is no need to memorize all those labels yet.

For now, remember one thing: an investment approach should explain why it might have a little more support than a random guess. That explanation must still face a test. It cannot award itself a certificate.

Question Two: How Long Is It Prepared to Wait?

Hoppy wrote “long term” next to “the industry’s future.” Then he wrote “short term” next to “price trend.”

Dr. Hop asked, “How short is short? And long compared with what?”

Hoppy had no answer.

Five trading days may feel long to an intraday trader. For someone studying a factory expansion, three months may be barely enough for the equipment to arrive.

Once research begins, it is better to replace “short term” and “long term” with a period we can actually check: the next five trading days, the next three months, one year later, or perhaps two to three years.

And a single research idea often contains three different clocks.

Imagine that HopPop Cola plans to build a new production line for sugar-free drinks.

The first clock hangs in the real world. Consumers may take years to change their tastes. The new line must go through approvals, purchasing, installation, and trial production. Opening a spreadsheet does not make the factory skip to the ending.

The second clock hangs in the market. The share price may move as soon as the plan is announced. Or investors may remain skeptical until sales actually improve. Price can run ahead while the real-world change is still unfinished, or reality can change before price fully catches up.

The third clock belongs to the researcher. They must choose when to start observing, whether to check the result after sixty trading days or after a year, and how long they will wait before admitting that the original idea may not be unfolding as expected.

The three clocks do not have to agree.

An industry shift may last three years. The market may start trading that expectation six months early. The researcher may look only at returns over the next sixty trading days.

This is also where people often make a simple mistake: using daily data does not mean you can study only short-term trading.

Daily data simply records one observation per trading day. More than two hundred daily observations can describe roughly a year. Several years of daily data can help us study much longer patterns.

How often the data is recorded and how long an idea is observed or held are two different questions.

Question Three: How Will It Find and Check Evidence?

Suppose three researchers all want to know whether growing demand for sugar-free drinks could create an opportunity for HopPop Cola.

The first takes out a pen and draws a path. A shift in consumer taste may increase sales. Sales volume and pricing affect revenue. Revenue, costs, and competition then affect profit and market expectations.

This researcher begins with: Why would this happen?

That approach starts from a real-world mechanism. Its strength is that every step has a meaning. Its weakness is that even a beautifully connected story may still be only a story.

The second researcher opens a historical dataset. Instead of adding more to the story, they first define “growing demand,” decide which companies to observe and for how long, and compare whether similar situations appeared repeatedly in the past.

This researcher begins with: Has this happened again and again before?

Statistical comparisons can expose conclusions built from a few memorable examples. But a repeating shape in the data does not tell us its cause, and testing too many variations can produce a beautiful result by accident.

The third researcher feeds company, industry, and price information into a model and asks it to estimate what happens next.

This researcher begins with: Given this information, can we make a better prediction?

A predictive model can work with many clues, but it may look impressive on old data and fail in a different period. Even an accurate prediction does not prove that we understand the real cause.

The three researchers do not need another tournament.

The first can use a mechanism to suggest a hypothesis. The second can use statistics to check whether the pattern appeared repeatedly. The third can ask whether adding more information improves prediction. The order can also run the other way: someone may find a pattern in data first, then go looking for an explanation.

One asks why. One checks whether it happened repeatedly. One asks whether we can predict better.

They can work together, but they cannot sit one another’s exams.

So Where Do Technical Indicators, Python, and AI Fit?

Hoppy pointed at the candlestick chart on the desk. “What about technical indicators?”

A technical indicator usually rearranges price, volume, or other market records according to a formula. It can become part of statistical or rule-based research. But calculating an indicator does not automatically calculate the source of an advantage.

Python can help us clean data, run rules, and produce results. AI can help us discuss a hypothesis, write code, and investigate errors.

Both are useful. Neither can answer the first two questions for us.

“I used AI” does not explain why a pattern might exist in the market. “I wrote Python” does not tell us how long that pattern may take to appear.

Tools shape how we work. They do not automatically explain why the research might work.

What does it mainly observe?

Where might the advantage come from?
How long is it prepared to wait?
How does it mainly find and check evidence?

What is it most likely to miss?