Although efficient-market theory argues against outperformance, earnings sentiment continues to create short-lived opportunities.
For more than half a century, investors have been pondering over the same uncomfortable question: Does it still make any sense to try to beat the market? On one side, there are the cheerleaders of passive investing, armed with decades of data and Nobel Prizes. They have a simple, almost moral argument that markets are brutally efficient, trading costs work against you, and most attempts to outperform the broader index end in disappointment.
On the other side, there are traders, hedge fund managers, and systematic strategists who continue to hunt for excess returns. They have a point: inefficiencies still exist for those who know where to look.
Both sides are partly right. Which is why we're still talking about it.
The historical case against outperformance
So, why is there a solid skepticism against beating the market? Well, there are intellectual roots. The Efficient Market Hypothesis (EMH), developed in the 1960s and refined over time, argued that prices rapidly incorporate all available information. If that is true, even approximately, then it becomes extraordinarily difficult to beat the market using public data.
Numerous empirical studies have reinforced this view over the following decades. Large-scale studies of active fund performance showed that, after fees and costs, most managers underperform their benchmarks over long horizons. Even among the minority that do outperform for a while, they are rarely persistent. So, while they have been successful in one period, the success is often followed by mediocrity. Does that indicate that luck plays a larger role than many are comfortable admitting?
This is why indexing became the popular choice, as it helps, at least theoretically, most investors stay on the right side of the market.
But why is this conclusion incomplete?
If markets were perfectly efficient in practice, as EMH argues, certain return patterns would not have survived for decades across geographies, asset classes, and market regimes. But they have.
Let's take the example of momentum. Stocks that have performed well over the past several months have tended, on average, to keep outperforming for a while longer. This same pattern has shown up not just in U.S. equities, but across international markets and even in other asset classes. It has survived multiple market cycles, recessions, bubbles, and crashes. The same is true for value, where cheaper stocks have historically delivered higher long-term returns than expensive ones, even though the idea has been widely known for decades.
These anomalies do not simply mean that it's easy to beat the market. But they indicate something more subtle: while markets are highly efficient in some dimensions, they are stubbornly inefficient in others. For instance, stock prices respond quickly to clean, numerical data. But markets tend to move much more slowly when the information is messy or unstructured — for instance, the change in tone of an executive in an earnings call.
This slow digestion of qualitative information explains a well-known phenomenon in finance: Post-Earnings Announcement Drift (PEAD). For decades, researchers have observed that stocks often continue to move in the same direction for weeks or even months after an earnings announcement, long after the initial reaction should have priced in the news.
Markets are efficient, but humans are not
Markets are not abstract machines processing clean numbers in isolation from the economy or society. They are social systems populated by humans and institutions with incentives, constraints, and psychological biases. Investors don't suddenly become perfectly rational just because they are looking at a trading screen. They still panic when the market correction is sharp, feel safer investing in familiar narratives, and often extrapolate recent trends too far into the future.
Another layer of friction in the market comes from the fact that portfolio managers are evaluated over short time frames, and they operate within mandates and benchmarks. As a result, decisions are not made solely on long-term expected value, but on what is defensible and based on consensus.
All of this slows down how the market digests certain types of information. Clean numbers — revenue, earnings, and margins — are easy to process and compare. More ambiguous signals, such as uncertainty in guidance, changes in tone, or shifts in how management frames the business, take longer to be reflected in prices.
This gap between how markets should work in theory and how people actually behave in practice makes market efficiency uneven. This means market-beating opportunities can still be found in the market corners where human judgment and interpretation slow down how information is reflected in prices.
Why earnings calls remain under-processed information
Earnings calls, in particular, sit right in the middle of this gap. We instantly know the hard numbers and prices are quick to adjust accordingly. But what executives say on the call isn't just about a single guidance figure. It shows up in their tone, the points they repeat, the questions they lean into (or avoid), and even in what goes unsaid.
Of course, markets react to these cues, but not always immediately, and not always consistently. Researchers have shown that the tone of executive language often correlates with future earnings surprises, analyst revisions, and post-earnings drift. In other words, how something is said can matter just as much as what is said.
Despite this, earnings calls remain cognitively expensive to process. A human analyst can listen to a few carefully, but they cannot do so across hundreds of companies, quarter after quarter, with perfect consistency and timing.
From language to signal: what changed with AI
For years, sentiment analysis was either too crude or too inconsistent to be of use to any real trader. Keyword counts or simple polarity measures produced confusing signals that hardly offered any value under real market conditions.
Modern AI changes that equation.
Advances in natural language processing now make it possible to extract sentiment in a more context-aware, financially literate way. Instead of vague labels, corporate language can be transformed into time-stamped sentiment measures that are consistent across calls, comparable across firms, and testable across market cycles.
When we apply that to live earnings calls, this turns qualitative nuance into structured data — which can be acted upon before prices have fully adjusted.
This is what platforms like AlphaPro are trying to achieve. While sentiment alone can hardly forecast markets with certainty, the aim here is information compression — reducing hours of listening and interpretation into clear, real-time signals that can feed a systematic decision process.
How sentiment actually helps performance in practice
In real trading systems, sentiment does not work as a standalone signal. But it can become a strong signal when it interacts with other information. For example, it can reinforce a trend when the story around a stock is getting stronger, or raise a red flag when management doesn't sound confident even though prices are still going up.
It can also highlight situations where the numbers look fine at a glance, but the language coming from management tells a different story. If we use sentiment in that way, it can help with timing, filter out false moves, and keep you out of trades where price and narrative are slowly drifting apart.
None of this, however, eliminates the hard truths. Finance is full of smart ideas that fell apart because people were too confident and underestimated how messy real markets can be.
But it would be equally naive to conclude that all opportunities are gone. Because language-based sentiment has properties that make it unusually resilient. It is high-dimensional, context-dependent, event-driven, and deeply tied to human behavior. These characteristics slow commoditization and make the signal harder to arbitrage away completely.
So, is it worth trying to beat the market?
For most investors working alone, the honest answer remains no. The statistical odds favor simplicity, diversification, and humility. That said, for systematic traders, quantitative researchers, and teams willing to invest in data, validation, and risk control, the answer can still be yes. This is because markets remain inefficient in specific, human ways.
As long as markets are made of people who speak, persuade, hesitate, and frame stories, sentiment will continue to matter.