I came to this question through Maxwell’s demon. The imagined being watches individual gas molecules and opens a small door selectively, sorting faster molecules from slower ones. Its knowledge appears to let it create a temperature difference from which work could be extracted.
That suggested a question about trading: how does knowing more become earning money? An observer notices a distinction and acts on it. In a market, that action meets a price, another participant and the practical conditions of a trade. The information has to make a difference somewhere along that path. And if the price already reveals what the observer learned, why pay to learn it?
The physical account
One cycle includes the observer.
Two observations, A or B; memory starts ready.
01–02 Observe & act
Record kept through the action.
03 Erase to reuse
Include the surroundings. Compensating entropy increase outside the erased record.
Memory ready for the next observation
The economic account
The physics gives this inquiry a starting point. It also requires care. In the Landauer–Bennett account, measurement can in principle be reversible; logically irreversible operations, including erasing an unknown memory record, require entropy accounting. A demon repeatedly completing its cycle has to include its own apparatus. That does not set the price of research or establish a law of trading profits. I take from the analogy the need to follow observation through action and account for the process that makes the result possible. Bennett, Notes on Landauer’s Principle, Reversible Computation, and Maxwell’s Demon
Information has value through a decision
Consider a constructed example. A claim pays either 100 or zero at a fixed future date. Each outcome initially has a 50% probability. Buying one claim costs 50, plus an execution charge of 5. The available choices are to buy one or keep the money. To isolate the decision, compare expected monetary payoffs and leave out interest, financing and risk preferences.
A report costs 2. It is equally likely to be favorable or unfavorable. Conditional on a favorable report, the probability of the 100 payoff is 75%; conditional on an unfavorable one, it is 25%. These probabilities are assumptions of the example, not evidence of forecasting ability. For now, assume the claim remains available at 50 after the report arrives. The trading gains below exclude the report’s price.
- No report. Buying has an expected gain of 50 − 50 − 5 = −5. Keep the money.
- Favorable report. Buying has an expected gain of 75 − 50 − 5 = +20. Buy.
- Unfavorable report. Buying has an expected gain of 25 − 50 − 5 = −30. Keep the money.
Before the report is purchased, the policy it enables has an expected payoff of half of 20 plus half of zero: 10. Subtract the report’s price of 2, paid whichever message arrives, and the expected net gain is 8. Without the report, the best available choice in this example earns zero. That comparison gives the information an economic value: it allows a different choice in different circumstances.
The claim’s unconditional expected payout is still 50. The report has not increased it. Nor has uncertainty disappeared: after a favorable report, the zero payout still has a 25% probability. In that outcome the purchase loses 55, plus the report’s cost. The expected gain comes from selecting when to buy, while accepting that an informed decision can lose.
The price can carry the same information
The example grants a willing seller at 50; it has not explained why that price would remain available. Suppose instead that, after a favorable report, the available price becomes 75. The expected payout is also 75, leaving a loss of 5 after execution. The forecast remains accurate under the assumptions; the opportunity to profit from it at the old price has disappeared.
Trading can reveal information to others. If prices revealed everything research uncovered, someone could learn from the price and avoid the bill. Who would then pay to discover the information?
Grossman and Stiglitz model this dependence. Some traders buy a signal; others observe prices. Informed demand moves prices, but so does random asset supply, unobserved by the uninformed. Price changes therefore reveal the signal only partially.
Researchers pay upfront. They gain through trading with the uninformed on terms more favorable, on average, than if information were shared equally. The uninformed learn from prices and save the research cost. Where both groups coexist in equilibrium, their expected utilities are equal after information costs. Remaining uninformed can be rational.
As more traders become informed, prices reveal more and the relative benefit of purchasing information falls. Under the model’s assumptions, complete revelation cannot sustain costly research. This is a mechanism for financing information through trade, without a promise that any particular trade will profit. Grossman and Stiglitz, On the Impossibility of Informationally Efficient Markets
An opportunity still needs an executable trade
The numerical example also granted a full purchase at the quoted price. A real order has to arrive while that price is available. Increasing its size may require paying more. A limit order may fill selectively when the market moves against it, so the expected payoff conditional on a fill can differ from the forecast made before submitting it.
These effects belong in the account of the trade itself. Subtracting a fee from a backtest cannot repair a purchase that could never have occurred. A forecast can survive examination while the proposed way of earning from it fails.
Costs can also change the decision. In the example, a higher execution charge narrows the occasions on which buying is worthwhile. If the report became more expensive than the expected improvement it enabled, it would no longer be worth purchasing. The comparison is between the feasible choices with and without the information, after their respective costs.
Competition changes that comparison too. Other traders may learn the same thing, or infer it from prices. What was useful yesterday can still predict an outcome while offering less advantage at today’s purchase price. The economic value belongs to the information in a particular trading situation, rather than to the forecast in isolation.
What did a profitable trade establish?
Following information into a trade explains how it could produce money. Looking at money afterward does not identify that path uniquely. A trader who held broad market exposure may have been paid for bearing risk. Someone supplying liquidity may earn a spread while carrying inventory and facing adverse selection: trading against a counterparty whose information makes the quote unattractive. Luck can produce a positive result too. These explanations can coexist with an information advantage.
Alpha makes a narrower claim than profit: a return remains unexplained by a chosen model of normal returns. Another model may explain it through an omitted exposure. Fama’s joint-hypothesis problem captures this difficulty: testing market efficiency also tests the pricing model used to judge returns. An apparent anomaly alone cannot identify which account failed. Fama, Efficient Capital Markets: II
I would therefore examine the proposed connection. Records of what was known at the decision time can rule out information that arrived later. Feasible fills and stated expenses determine which trades and net results are being claimed. Observations outside the data used to choose the strategy can challenge the forecast. Comparing with a feasible strategy that did not use the signal, while accounting for differences in risk and costs, helps test what the information added. The constructed example supplies these conditions by assumption; actual research has to establish them.
A trader can therefore be right about an outcome and still have no profitable trade to make. This is where the market changes the demon’s story: other observers act too, and their actions change the terms on which knowledge can be used. Research can help make a price more informative even as the opportunity to earn from that research shrinks. For the trader paying the bill, what matters is the improvement in a decision still available to them, after the costs of learning and acting.