Where it is genuinely used: execution, stops and sizing

Nine chapters of what price data cannot do. This one is what it can, and the distinction turns on a single question.

The question that sorts everything

Does this use require the data to predict the future?

Chapters 3 to 8 all described uses that do. A trend implies continuation; a level implies a reaction; a pattern implies an outcome. Chapter 9 reported what the evidence says about that class of claim.

But price and volume data have uses that make no forecast at all — and those uses are untouched by everything in chapter 9, because they do not rest on the thing chapter 9 undermines.

Execution

You have decided what to buy and how much. The remaining question is how to transact. That is not a forecasting problem; it is a logistics problem, and price data genuinely informs it.

Liquidity and spread. Volume data tells you how much normally trades. An order that is a large fraction of typical volume will move the price against you, which is a cost — and knowing the ratio in advance tells you whether to break the order up.

Timing within a session. Chapter 2 noted volume has structure: heavier at open and close, lighter in between. Transacting when liquidity is thin costs more. That is an observation about market microstructure, not a prediction.

Order type. Whether to use a limit or a market order depends on the spread and the urgency, both observable.

None of this says what price will do. It says what your own order will cost, which is a different question with a knowable answer. And it matters: the Measuring your return and Behavioural finance subjects both found that costs, not selection, drive individual underperformance.

Stops

A stop is a decision made in advance about the maximum loss you will accept. It is a risk-control device and it requires no view on direction.

What it genuinely provides:

A pre-committed exit. The Designing around yourself chapter in Behavioural finance argues that decisions made calmly in advance beat decisions made under pressure. A stop is that principle applied to a position.

A bound on a single outcome. Combined with sizing below, it makes the worst case known rather than discovered.

What it does not provide, and is often claimed to:

It is not a guarantee. In a gap, the exit occurs at the next available price, which may be far worse. The Financial institutions subject's forced-selling mechanism applies: everyone's stops trigger together, into the same thin market.

It does not make a bad position good. A stop limits loss on a position you should not have held; it does not substitute for the judgement about whether to hold it.

It generates trading. Every stop that triggers is a transaction with costs, and a stop placed close enough to be triggered by ordinary noise will trigger often. That turns a risk-control device into a cost-generating one — which is the trade-off to be explicit about rather than the hidden defect it usually is.

Sizing

The most defensible use of price data in this entire subject, and the least discussed.

Volatility — the dispersion of past returns, from the Quantitative methods subject — is computed from price data. And it has a use that requires no forecast:

Size positions so that comparable moves produce comparable outcomes.

A position in something that typically moves 1% a day and one in something that moves 5% a day are not equivalent at the same rupee value. Scaling by volatility makes the risk contribution of each holding comparable.

Why this is legitimate where the rest is not. It uses the past to estimate dispersion, not direction. Volatility is substantially more persistent than returns are — calm periods tend to be followed by calm periods — which is among the more robust empirical regularities in finance, and it is a far weaker claim than predicting where price goes.

The honest caveat is that volatility estimated from a quiet period understates what a turbulent one will deliver, and the Measuring your return subject's warning applies: volatility assumes thinner tails than markets have, and an illiquid asset's low measured volatility reflects stale prices rather than safety.

The cost test, applied to yourself

The literature's standard for judging a rule deserves to be turned on the reader:

A technical trading rule is profitable only if its risk-adjusted profits exceed transaction costs incurred from implementing trades.

Apply that to your own activity, with the method in Measuring your return: compute your XIRR, compare it against a benchmark fixed in advance, and subtract what you paid in brokerage, spread and taxes.

Barber and Odean's finding sets the prior: the most active quintile of households earned 11.4% a year net against 18.5% for the least active, and the gap traced to costs rather than selection. Any method that generates reasons to transact is working against that.

Working the problem

Sorting the four decisions.

Deciding what to buy — requires a forecast. You are asserting this asset will do better than alternatives. Nothing in chapters 3 to 8 has been shown to support that, and chapter 9 is the evidence. This subject cannot legitimately help here.

Deciding how to spread a large order over a day — does not require a forecast. You already know what you want. The question is how to acquire it without moving the price against yourself, and volume and spread data answer it directly. This subject helps, genuinely.

Deciding how much to buy — does not require a directional forecast, though it requires a dispersion estimate. Sizing to volatility makes positions comparable in risk without any view on which way they go. This subject helps.

Deciding when to sell — depends entirely on why. If the reason is "the chart says the trend has ended", that is a forecast and falls under chapter 9. If the reason is a pre-set loss limit, a rebalancing rule, or because you need the money, no forecast is involved and price data simply executes the decision. The same action is legitimate or not depending on which question it answers — which is the cleanest statement of this subject's boundary.

So: two of four, and part of a third. That is a smaller claim than this field usually makes for itself and it is a real one.

The point

The question that sorts legitimate uses from the rest is whether the use requires price data to predict. Execution — how to transact a decision already made — does not: volume and spread tell you what your own order will cost, which is knowable. Sizing to volatility does not: it uses the past to estimate dispersion rather than direction, and volatility is far more persistent than returns. Stops are pre-commitment rather than prediction, though they guarantee nothing in a gap and generate costly trades when set inside ordinary noise. Choosing what to buy does require a forecast, and that is the use chapter 9 found the evidence does not support.