What risk management actually is

The same trade, twice 

Two accounts holding identical equity open the same instrument, in the same direction, at the same price, and close it at the same price the following day. One records a debit that barely registers on the statement. The other records a debit that changes what the account is able to do next. Nothing about the market differed between them: the quote was the same quote, the move was the same move, and neither account had the slightest influence over either. The only quantity that differed was the number of units each contract covered, and it was fixed before the position existed. The whole subject sits in that observation. The outcomes diverged, the price path did not, and what produced the divergence was decided in advance of the market rather than in response to it.

The block is symmetrical on purpose, because the point it carries is not that a smaller size is better. The larger size produces the larger credit in the favourable case by precisely the factor that produces the larger debit in the adverse one. Size scales both columns identically and favours neither. What makes it the first subject of this module is narrower than a preference: the multiplier is stated by the trader and the move is not, so it is the one term settled before the market has been consulted at all.

What is settled in advance, and what is not 

The word control is used loosely in trading writing, and the separation underneath it is sharper than the word suggests. One list holds the quantities that are typed into a platform before anything happens. The other holds the quantities that arrive from the market and are read rather than written.

Stated before a position exists:

• The instrument, and whether any position is taken at all.
• The direction, long or short.
• The number of units the contract covers.
• The level at which the position is closed if the market moves against it, and whether that level rests on the server as an order or exists only as an intention.
• How many positions are open at the same moment, and how closely related the instruments behind them are.
• Whether a position is held across a scheduled release, a session close or a weekend.
• The point at which trading stops for the day, and what condition defines that point.

Arriving from the market, and read only:

• The next price, and every price after it.
• Whether a level is reached at all, and which of two levels is reached first.
• The distance a market opens away from its previous close after a gap.
• The price a resting order actually fills at when the book is thin, which the costs module covered as slippage.
• How far a move continues once it has begun.

The first list is not a list of things that work. It is a list of things entered into a form, and no entry on it makes an adverse move less likely or brings any part of the second list closer to being known. What it fixes, in advance, is what the account looks like once the second list has done whatever it was going to do. That is the entire claim, and it is narrow on purpose.

Managing risk is not predicting outcomes 

A forecast is a statement about what a price will do. A size decision is a statement about what an account will look like if the forecast turns out to be wrong. They are answered at different times and by different parties: the forecast is settled afterwards by the market, the size beforehand by the trader, so however much conviction sits behind the first, it changes nothing about the arithmetic of the second. The two are frequently collapsed into one, usually by letting confidence in a view select the size.

Two traditions argue about where effort belongs, and a reader will meet both. One holds that selection is the substance of the work, on the reasoning that a position never taken cannot produce a loss at all, and treats sizing as bookkeeping that follows the real decision. The other holds that any individual result is close to indistinguishable from noise, and that the only quantities a trader states in advance are size and exit. Neither position can be settled by examining one trade, and neither is falsified by a run of results in either direction, which is why the argument persists.

Where the two arguments touch is worth stating plainly. A plausible reason for entering a position places no constraint on the sequence of results that follows it. A run of losses is not by itself evidence that a method was chosen badly, and a run of gains is not evidence that it was chosen well, because a run in either direction is compatible with a great many descriptions of what is happening. The account records the run either way, and the decline from an equity peak that a sequence of losses produces is arithmetic rather than opinion.

Why the close out arithmetic forces the question 

The margin module built the account panel field by field and ended at the level where the platform stops reporting and starts closing positions on its own. Two features of that arithmetic bear on this lesson. Used margin is fixed when a position opens and does not respond to price, so the denominator of the margin level does not move, and everything that moves the reading moves it through equity. And the unrealised result that moves equity is calculated on the full contract value rather than on the collateral held against it, so an adverse move is measured against the whole position, a loss can exhaust the margin posted, and losses are not limited to the amount deposited.

Together those produce the sentence this module is built on. The distance between an account's state when a position opens and the level at which the platform begins closing positions itself is a function of the size chosen at entry, not of the reason for the trade or of how carefully the chart was read. The arithmetic below is one account and one instrument at one price, with a single term changed.

The two positions are the same instrument, at the same price, in the same account. Changing the multiplier alone moves the level at which the firm's check intervenes from a price the market would have to travel almost the entire way to, down to a move of well under a tenth of the opening price. That check is a floor a firm enforces on its own exposure, not a plan and not a method. It runs on the server whether or not anything has been decided, and where an account sits relative to it was settled by the size decision.

What it does not do 

A description that stops at the previous section overstates the thing. Deciding a size in advance does not make an adverse move less likely, does not make a favourable one more likely, and says nothing about which of the two arrives. It fixes a multiplier, and the market supplies the other term without reference to what was decided.

Nor does an exit level cap a result. A stop loss is an instruction to close a position once a price reaches a stated level, and it is not a guarantee of the price at which that closing happens. In a market that gaps over a weekend or moves faster than the book can be refreshed, the position closes at the first available price beyond the level rather than at it, so the realised result can be worse than the intended one. The costs module covered that as slippage, and it applies to an exit as much as an entry.

Where practitioners disagree 

Three arguments run through everything that follows, and none resolves. The first is what size should be derived from. One convention derives it from a fixed fraction of account equity, on the reasoning that a rule with one input is a rule that gets followed, and is criticised for treating a quiet instrument and a violent one as the same object. Another derives it from a measure of the instrument's recent range, which answers that criticism, and is criticised in turn because a range measured over the past describes the past and can be wrong about the present at the moment it matters.

The second is whether an exit level belongs on the server as a resting order or in a note beside the screen. A resting order executes without anyone present, and is criticised because a brief excursion through the level closes a position the market then leaves behind. A level watched without an order avoids that, and is criticised because it depends on attention being available at the exact moment attention is hardest to give. Both criticisms are accurate, so the disagreement is about which failure is preferred rather than about which claim is true.

The third is whether limits measured over a calendar day or week correspond to anything real. One tradition holds that a boundary drawn somewhere is what stops a bad session continuing into a worse one, and that an arbitrary boundary is still a boundary. Another points out that a market has no interest in a Tuesday, that a limit reached at noon moves the same activity to a different day, and that the calendar belongs to the trader rather than the instrument. Each of those arguments, and the arithmetic underneath it, is the subject of the lessons that follow.