Major Buy Alert Issued for September 30th

THE HILL REPORT

A Signal Shared Is a Signal Consumed

Connor Hill · InsightfulWord · September 2, 2026

There is a property of market signals that has no equivalent in most other kinds of useful information, and it determines almost everything about what a distributed signal is worth.

A recipe does not stop working because a million people have it. A map does not become less accurate as it is copied. A trading edge does. It is consumed by being acted upon, and the consumption is proportional to how many people act.

The mechanism is not mysterious. An edge exists because an asset is mispriced relative to something. Buying it moves the price toward correctness. Enough buying eliminates the gap, and the gap was the edge.

This is not a theory about whether markets are efficient. It is arithmetic about what happens when purchase orders arrive. The price at which a hundred people can buy and the price at which a hundred thousand people can buy are different prices, and the difference is a cost borne by whoever is later in the queue.

The effect has been measured directly and at scale, in a setting where the timing of disclosure is precisely known: the academic finance literature, where predictive patterns are discovered, then published, and where returns before and after publication can be compared.

The result is one of the more robust findings in empirical finance, and it is unambiguous in direction. Published predictors decay, and the decay is larger than what mere sample variation accounts for.

A commercial signal service is the same phenomenon with a shorter timeline and a larger simultaneous audience. Publication in a journal reaches a specialist readership over months. Publication to a subscriber list reaches everyone within hours.

What follows is what happens to an edge after it is published, the volume signature that discovery leaves in the data, why capacity is the constraint nobody advertises, why the least tradable names decay slowest, and what a large simultaneous audience does to execution.

What Happens to a Published Edge

The definitive study on this took the question directly and answered it with an unusually clean design.

Two researchers assembled ninety-seven cross-sectional return predictors from eighty peer-reviewed studies — the accumulated set of published patterns claiming to forecast relative stock returns. For each one, they identified the sample period the original authors used and the date of publication.

That produces three distinct windows. The original in-sample period. The interval between the end of that period and publication, when the pattern existed in the world but was not yet widely known. And the period after publication.

Returns fell in each successive window. Out-of-sample but pre-publication returns were about twenty-six percent lower than in-sample. Post-publication returns were about fifty-eight percent lower than in-sample.

The gap between those two figures is the informative part. If the decline were purely statistical — the result of the original researchers having found patterns that were partly chance — the out-of-sample decline would capture it, and publication itself should change nothing. Instead publication produced a further large decline, which the authors attribute to investors learning about the pattern and trading it.

The estimate is that roughly a third of the total decline is attributable to publication specifically. The remainder is the correction of statistical over-fitting, which is a separate problem and was covered in this space before.

Two-thirds of an edge, then, tends to survive publication in a journal read by specialists. That is the optimistic reading. The pessimistic one is that fifty-eight percent of the original figure was never available to anyone who learned about it from the paper.

The Volume Signature of Discovery

The same study looked at what happened in the market around publication, and the trading data corroborate the return data in a way that is difficult to explain otherwise.

Trading volume in the affected portfolios rose measurably after publication, on the order of a fifth by share volume. The difference in short interest between the long and short sides of the portfolios widened by a multiple.

Those are the fingerprints of people acting on the information. Volume rises because trades are being placed; short interest diverges because the strategy involves shorting one side; and both changes appear after publication rather than before.

The cross-section of the decay is equally telling. Patterns concentrated in liquid, easily traded, low-idiosyncratic-risk stocks decayed more. Patterns living in names that are expensive or awkward to arbitrage decayed less.

That gradient is exactly what a mechanical explanation predicts and exactly what a purely statistical explanation does not. Over-fitting has no reason to be worse in liquid stocks. Trading pressure has every reason to be.

The implication for reading any performance claim is direct: the question is not only whether the pattern was real, but whether it lives somewhere that arriving money can reach without moving the price.

Capacity Is the Constraint Nobody Advertises

Every strategy has a maximum size beyond which its own trading destroys its returns, and the size is rarely stated because stating it caps the addressable market.

The constraint arises from market impact: buying pushes the price up, and the push grows with the quantity relative to what the market normally trades. The empirical regularity is that impact grows roughly with the square root of order size relative to daily volume, which means doubling the size increases the cost per share by about forty percent rather than doubling it.

That relationship is forgiving at small size and punishing at large size, and the point at which it starts to bite depends entirely on the liquidity of the names involved. A strategy in the largest listed companies has enormous capacity. The same strategy in companies trading a few hundred thousand dollars a day has almost none.

Institutional investors treat capacity as a first-order question, model it explicitly, and close funds to new money when they judge it reached. That practice is the clearest available evidence that the constraint is real, since closing a fund forgoes fee revenue.

The arithmetic that matters for a distributed signal is simple to set up. Take the number of recipients, multiply by a plausible position size, and compare the total to the daily dollar volume of the security. If the product is a meaningful fraction of a day's trading, the recipients are collectively the market for that day, and the price they receive is a price they created.

Nothing about that requires anyone to behave badly. It follows from arithmetic alone.

Why the Smallest Names Decay Slowest

There is a genuine counter-argument to all of this, and it deserves to be stated properly rather than dismissed.

If an edge lives in securities that are hard to trade — small, illiquid, expensive to borrow, or subject to constraints that keep institutions out — then the arbitrage that would eliminate it is costly, and the edge can persist for a long time.

The published evidence supports this. The predictors that decayed least after publication were exactly those in the hardest-to-arbitrage names, and this is the standard explanation for why some anomalies have survived decades of documentation.

But the argument cuts both ways, and the second edge is sharper. The same illiquidity that protects the edge from institutional arbitrage guarantees that a distributed retail audience cannot capture it either.

A pattern that survives because it lives in a stock trading two hundred thousand dollars a day is a pattern that cannot absorb a thousand buyers, let alone a hundred thousand. The protection and the impossibility are the same property.

So the two cases collapse into one answer. Where a signal is easily traded, it decays quickly and the early recipients get most of what there is. Where it is not easily traded, it persists — and cannot be acted on at scale. There is no configuration in which a large simultaneous audience captures a durable edge in a thin security.

What a Subscriber Base Does to Execution

The final mechanism is the one closest to the actual experience of anyone acting on a distributed recommendation.

Recommendations are released at a moment. Everyone receives them at approximately that moment. The orders therefore arrive together, in the same direction, in a compressed window.

Correlated order flow in one direction is precisely the condition that produces the largest price impact per share. A dispersed set of buyers across a week costs far less in aggregate than the same buyers arriving within an hour.

Order type compounds it. A market order in a thin security during a burst of correlated demand executes against whatever is resting on the offer, and the resting quantity in such a name is frequently small.

The pattern is observable from outside. A thinly traded security that trades many multiples of its normal volume on a single day, with a sharp move and a subsequent partial reversal, has a well-known shape, and it can be checked after the fact from public price and volume data.

The general observation is that a signal's value depends on how many people have it and how easily the underlying can absorb them, and that both variables are knowable. Neither appears in a performance figure. A past return is computed as though the trades were free and the trader was alone, and a distributed recommendation is neither.