Markets are not uniformly efficient. They are very good at processing information that is obvious, timely, and easy to act on. They are systematically less good at processing information that requires breadth, depth, and time — information that exists in plain sight, but that no human team has the bandwidth to read comprehensively.
That gap is where durable alpha lives.
What Markets Price Well
Headline information gets priced fast. An earnings beat, a guidance revision, a macro surprise — anything that arrives in a simple, digestible format and points clearly toward a trade is processed by thousands of market participants within minutes. By the time a human analyst has formed a view, the market has usually moved.
This is not a failure of human intelligence. It is a natural consequence of markets being full of intelligent, fast-moving participants all looking at the same obvious signals.
The implication for investors who want genuine edge is straightforward: the obvious cannot be the source. Something that any reasonably attentive market participant can observe and act on is already priced.
What Markets Miss
The market’s blind spots are not in hidden information. The blind spots are in publicly available information that is too voluminous, too fragmented, or too pattern-dependent to be fully processed by any human research team operating under real-world constraints.
Human analysts are bottlenecked by time. They cover a limited universe. They read selectively — the filings that are flagged, the companies in their coverage, the data points that match the current thesis. What they cannot do is read everything, continuously, across a universe of thousands of companies, with no bias toward what is interesting and no fatigue filtering what gets processed.
The information that sits in this gap is not obscure. It is the same information every market participant theoretically has access to. The difference is in who has the infrastructure to read all of it.
The Processing Advantage
An AI-native research system reads everything. Not selectively — everything. Continuously, as new information becomes available, across the full investable universe.
It applies no cognitive bias to what deserves attention. It does not prioritise the companies currently in favour with the market narrative. It does not skip the disclosures that seem routine. It builds a picture from every data point, compounding the depth of its understanding with every new input.
This is not a marginal improvement on what human analysts do. It is a different capability — one that only exists at scale, and only produces reliable signal after years of continuous operation and calibration.
The Validation Gate
Most apparent patterns in data are noise. The discipline that separates a genuine AI research engine from a data-mining exercise is the rejection rate — the fraction of apparent signals that fail rigorous out-of-sample testing before being admitted to the live book.
At AvB Capital, the vast majority of what we research does not survive validation. What passes is what has demonstrated genuine predictive power on data that was not used to discover it. The rejection process is not a failure — it is the point. It is what ensures that the signals which reach capital are real, not curve-fit artefacts.
The moat is not in reading the market differently. It is in having the infrastructure to read it comprehensively, and the discipline to act only on what genuinely survives the test.
AvB Capital is an AI-native systematic manager based in London. Track record available to qualified investors on request.
Frequently Asked Questions
Do markets miss information?
Yes — systematically. Not because the information is hidden, but because comprehensively processing everything that is publicly available is beyond the capacity of any human research team. Markets are very good at pricing what is obvious. They are less good at pricing patterns that only become visible when you process a large volume of information over a long period of time.
What is an information processing advantage in investing?
It is not about having access to better data. Most information that generates durable alpha is publicly available. The advantage is in the depth and consistency of how that information is read — processing more of it, more continuously, with no cognitive bias filtering what gets attention and what doesn't.
Why doesn't the market arbitrage away AI information advantages?
Because the advantage is not in an insight that can be copied. It is in the infrastructure that produces the insight — built over years of continuous operation, calibrated against live outcomes, and validated against a rejection rate that eliminates most apparent patterns before they touch capital. That infrastructure cannot be replicated quickly.
How durable are AI-derived investment signals?
Durability depends on whether the signal reflects a genuine, persistent relationship or statistical noise. Rigorous out-of-sample validation — the kind that rejects the majority of apparent signals before deployment — is what separates durable alpha from curve-fit artefacts. The validation discipline is at least as important as the discovery.