Position Sizing
The formula is simple. What it protects us from is not. Position sizing is not an optimisation problem. It is an engineering problem built around one requirement: remain in the game.
A seawall does not stop the ocean.
That would be an absurd objective.
The ocean is larger than the wall. Storms will come. Waves will arrive that nobody predicted. Some will be larger than anything experienced for years.
The wall has another job.
It determines how much of that force reaches what lies behind it.
That is much closer to the role of position sizing in a systematic trend following programme.
Position sizing does not improve the entry.
It does not tell us which market will trend.
It does not make the exit smarter.
It does not increase the probability that the next trade will work.
In fact, viewed one trade at a time, position sizing is profoundly uninteresting.
Its importance appears somewhere else.
Across hundreds of trades.
Across losing streaks.
Across markets suddenly becoming violent.
Across years in which the strategy appears to have forgotten how to make money.
Across the inevitable occasions when several things go wrong together.
Position sizing determines how much damage each of those events is permitted to do.
That makes it one of the least glamorous components in the programme.
It also makes it one of the reasons the programme survives.
The Formula Is Not Looking for the Best Bet
There is a temptation in finance to treat position sizing as an optimisation problem.
How much should we bet given the quality of the opportunity?
How much capital should we allocate to maximise expected return?
Should a stronger signal receive a larger position?
Should a market with better historical performance receive more capital?
Those are reasonable questions.
They are not the questions our formula is trying to answer.
Our position sizing mechanism is deliberately more modest.
It asks:
Given the amount of capital we are prepared to expose to this trade, the market’s recent movement and the distance to the initial stop, how large can the position be?
That is a very different problem.
Suppose we are prepared to allocate $1,000 of planned loss exposure to a trade.
The market’s ATR is 50 points.
Our initial stop is two ATR away.
The contract is worth $10 per point.
One contract therefore has an intended stop distance worth:
50 × 2 × $10 = $1,000.
So we hold one contract.
Now suppose the market becomes twice as volatile and ATR rises to 100 points.
The same two-ATR stop now represents $2,000 per contract.
If fractional contracts were possible, the formula would call for half a contract. With futures, of course, they usually aren’t, so contract granularity becomes a practical constraint.
Nothing in that calculation says the market is more dangerous.
Nothing says the trade is less attractive.
Nothing says the trend is more or less likely to continue.
The formula simply observes that the market is now moving more per unit of time.
So less exposure is required to maintain the same planned loss allocation.
That is the mechanism.
And its indifference is one of its strengths.
“The formula does not find the optimal position. It finds the survivable one.”
ATR Is Not Risk
This distinction matters enough to stop here.
ATR measures movement.
It does not measure risk.
A market with an ATR of 100 is moving more than a comparable market with an ATR of 50. That tells us something useful about the scale of recent price variation.
It does not tell us everything that can hurt us.
It does not tell us whether liquidity will disappear tomorrow.
It does not tell us whether the market will gap through our stop.
It does not tell us whether five apparently different positions are actually expressions of the same underlying exposure.
It does not tell us whether the political regime supporting a currency is about to collapse.
It does not tell us whether a quiet market is genuinely benign or merely waiting.
That last distinction is particularly important.
Markets can become eerily quiet before they become violent.
Volatility contracts.
Ranges narrow.
ATR falls.
The sizing formula responds exactly as designed: for a given planned loss allocation and stop multiple, a lower ATR permits a larger position.
Then something changes.
The market gaps.
Liquidity vanishes.
Correlations that looked harmless suddenly converge.
The realised loss can be very different from the neat number produced by the formula.
Was the formula wrong?
No.
We asked it a question it was never capable of answering.
ATR normalises exposure to recent movement.
It does not reveal the full structure of uncertainty hiding behind that movement.
This is why I resist calling ATR-based sizing “risk normalisation.”
It is too generous a description.
Risk is larger than volatility.
And survival requires more than a formula.
The Quiet Market Problem
Consider two markets.
One has been moving violently for weeks.
Everyone can see it.
ATR is elevated. The stop distance is wide. The sizing formula gives us a relatively small position.
The other has barely moved.
ATR is tiny.
The formula allows a larger position.
Which is safer?
We don’t know.
The quiet market may genuinely remain quiet.
Or its apparent stability may represent compression.
Participants may have converged around the same assumptions. Positions may have accumulated. Liquidity may appear deep because nobody currently needs to cross it.
Then the underlying condition changes.
Suddenly everyone wants the same door.
This is one of the recurring traps in markets.
Observed calm and structural safety are not the same thing.
The sizing formula cannot solve that problem because the information required does not exist inside the ATR calculation.
Nor should we keep adding clever adjustments until it pretends that it can.
Instead, the architecture solves the problem elsewhere.
We diversify.
We limit individual exposures.
We spread positions across markets and sectors.
We recognise that apparently independent markets can become correlated under stress.
We design the portfolio so that no single sizing calculation is being asked to protect the entire programme.
That is the subject of Episode 4.
And it reveals something important about the anatomy of the system.
Every component has boundaries.
Robustness comes partly from respecting them.
Different Markets. Same Arithmetic.
Now consider the problem the formula can solve extremely well.
How do we put cocoa, bonds, currencies, equity indices, metals and energy markets into the same portfolio without allowing their completely different price scales and contract specifications to dictate our exposure?
We translate them.
Not into the same price.
Not into the same volatility.
And certainly not into the same underlying risk.
We translate them into a common planned loss budget.
A violently moving commodity may require very few contracts.
A quieter financial future may require more.
Contract counts can therefore look wildly different across the portfolio.
That doesn’t matter.
A contract is not a unit of exposure that can be meaningfully compared across different futures markets.
The formula performs the translation.
It asks what the stop distance represents in dollars per contract and then determines how many contracts fit within the allocation.
Same arithmetic.
Different market.
Again.
And again.
And again.
There is something almost boring about that.
Good.
Robust systems contain a surprising amount of boredom.
Position Size and the Stop Are One Structure
Episode 2 dealt with the initial stop.
Now we can see why it cannot really be separated from position sizing.
Imagine widening the stop but leaving the number of contracts unchanged.
You have changed the planned loss exposure.
Now imagine widening the stop and reducing the number of contracts proportionately.
The price has more room to move, but the planned capital exposure remains broadly consistent.
The stop and the size are therefore not two independent decisions.
They are connected through the same arithmetic.
This matters because people often talk about tight stops as though they are inherently conservative.
They aren’t.
A tighter stop combined with a correspondingly larger position can produce exactly the same planned dollar loss as a wider stop with a smaller position.
What changes is the geometry of the trade.
One gives price less room with more units.
The other gives price more room with fewer.
The sizing formula connects those choices.
This is another reason isolated rules tell us very little about the behaviour of the programme.
We have to look at the architecture.
Entry.
Initial stop.
Position size.
Trailing exit.
They are separate mechanisms, but they are not independent mechanisms.
Change one and something elsewhere usually moves with it.
The Position Changes. The Formula Doesn't.
There is another distinction worth making.
The sizing formula establishes the position at entry.
Then the market begins doing whatever the market is going to do.
Suppose the trade moves strongly in our favour.
The trailing stop advances.
The relationship between current price and exit level changes.
The market’s volatility may change too.
The position we now hold is no longer economically identical to the position we opened.
But we do not need the original sizing formula to describe every subsequent state of the trade.
That is not its job.
It answered the question it was given at entry.
Other mechanisms now take over.
This is a recurring theme in the series.
Robustness does not come from building one magnificent formula that knows everything.
It comes from giving narrow jobs to different pieces of machinery.
The entry detects.
The initial stop defines the intended failure boundary.
Position sizing determines exposure.
The trailing exit manages the developing trade.
The portfolio controls aggregation.
The Cut Back Rule responds when the programme itself comes under pressure.
The intelligence is distributed.
Why We Don't Size Up Because We Feel Clever
A profitable trend creates an interesting temptation.
We entered.
The market moved our way.
Surely we now know more than we knew at entry.
Why not increase the position?
Sometimes there are systematic strategies for which adding exposure is explicitly part of the design. Pyramiding has a long history in trend following.
But that is different from discretionarily increasing exposure because a trade now feels confirmed.
The distinction matters.
A market that has already travelled a long way has demonstrated trend.
It has not demonstrated what happens next.
The future remains unavailable.
Indeed, some of the most violent reversals occur after trends have become obvious enough that almost everyone agrees they exist.
So our basic sizing philosophy remains deliberately unexciting.
We do not look at a wonderful trend and say:
This one deserves more.
Nor do we look at an ugly recent sequence and decide:
This one deserves less.
If changes in exposure are part of the system, they must themselves be systematic.
Otherwise position sizing quietly becomes another forecasting mechanism.
And we are back where Episode 1 began.
Pretending we know more than we do.
Setting the Scale
Eventually every sizing formula reaches a question mathematics cannot answer for us.
How much capital should we allocate to each trade?
The formula needs that number before it can calculate anything.
Suppose we choose 1% of current equity.
On a $200,000 account, that creates a planned allocation of $2,000.
Choose 0.5% and it becomes $1,000.
The arithmetic is trivial.
The consequences are not.
Because the number determines how quickly a sequence of ordinary losses becomes an extraordinary drawdown.
And trend following will produce sequences of losses.
Not might.
Will.
If a system wins fewer than half its trades, consecutive losing trades are not evidence of malfunction. They are part of the expected path.
Five.
Six.
Eight.
Ten.
The precise sequence is unknowable.
That is why the allocation cannot be chosen by asking how much we are comfortable losing on one trade.
The better question is:
How much can we allocate to one trade while remaining functional through the sequence we have not yet seen?
That is a survival question.
And survival has geometry.
Lose 10% and you need 11.1% to recover.
Lose 20% and you need 25%.
Lose 50% and you need 100%.
As drawdown deepens, recovery becomes increasingly demanding because capital compounds multiplicatively.
This is why position size matters so much.
The first job is not maximising return.
The first job is ensuring there is enough capital left for compounding to continue.
But that leaves one important question unanswered. Which equity are we applying that percentage to?
Why We Size From Closed Equity
There is another decision buried inside the sizing calculation that matters far more than it first appears.
What do we mean by equity?
Suppose the programme begins with $1 million and a series of strong trends produces $200,000 of open profit.
The account now shows total equity of $1.2 million.
But the closed balance remains $1 million.
Which number should determine the size of the next trade?
We use the second one.
Position size is calculated from closed balance equity, not total equity.
The distinction sounds like accounting.
It isn’t.
If we size from total equity, the $200,000 of unrealised profit immediately becomes part of the capital base used to establish new positions. A 0.5% allocation calculated on $1.2 million is $6,000 rather than the $5,000 calculated on the $1 million closed balance.
So something subtle has happened.
The existing trends have increased the size of positions elsewhere in the portfolio before their profits have actually been realised.
We have begun spending the harvest while the crop is still in the field.
That creates a feedback loop.
Open trends generate unrealised profits.
Those profits increase total equity.
Higher total equity produces larger new positions.
Larger positions increase aggregate exposure.
Then the original trends retrace.
And they will retrace. Episode 2 established that giving back part of an open profit is the unavoidable price of allowing an outlier room to become an outlier.
The unrealised equity that supported those larger position sizes can now disappear.
The larger positions it helped create do not.
This is why we separate the two.
We compound what has been realised. We allow what remains unrealised to breathe.
Open equity belongs to the trend-harvesting machinery. It is allowed to expand, contract and retrace according to the exit logic. We do not want a large open trend continually changing the capital base from which unrelated new positions are sized.
Closed equity belongs to the capital-allocation machinery.
When an open profit is finally realised, it crosses that boundary. It becomes part of closed balance equity and can legitimately increase the size of future positions.
The reverse is equally important.
When losses are realised, closed equity falls. The sizing base falls with it. Future positions become smaller.
So the programme compounds, but it compounds from what has actually been banked.
This creates a clean separation between two different jobs inside the architecture.
Open equity tells us what the programme is currently worth. Closed equity determines how much capital the programme is permitted to deploy next.
That distinction becomes even more important when we introduce the Cut Back Rule.
Because the Cut Back Rule also listens to closed equity.
That is not a coincidence.
The ordinary position-sizing mechanism increases or decreases exposure as realised capital changes. The Cut Back Rule provides an additional defensive response when deterioration in that realised path becomes sufficiently significant.
Both mechanisms are looking at the same thing for the same underlying reason.
We do not allow temporary expansion in unrealised profits to lever the programme up.
And we do not require healthy open trends to be cut simply because those unrealised profits breathe.
Closed equity governs the capital machinery. Open equity belongs to the outlier-harvesting machinery.
Keeping those two worlds separate is one of the quiet ways the programme protects itself from its own success.
More Capital Does Not Simply Mean Bigger Positions
There is one more step.
Using closed equity as the capital base does not mean every increase in closed equity should immediately produce larger positions.
That would simply replace one feedback mechanism with a slower version of another.
Suppose the programme begins with $200,000 and, over time, realised profits lift the closed balance to $220,000.
We have genuinely earned another $20,000.
It belongs to the capital base.
But that does not mean every position should immediately become 10% larger.
Our preference is for weak compounding.
We allow meaningful realised growth to accumulate before increasing the scale of the programme. And when additional capital becomes available, increasing the size of existing positions is not necessarily the first thing we do with it.
There is a hierarchy.
Markets first. Systems second. Position size last.
Why?
Because additional capital can buy us something more valuable than simply more exposure to what we already own.
It can buy us breadth.
A futures contract that was previously too large for the account may now become feasible.
Another market can enter the portfolio.
More capital may allow another robust system to be deployed across markets already being traded.
The search network becomes wider.
Only after those opportunities for useful diversification have been exhausted do we need to ask whether existing positions themselves should become larger.
This is an important distinction.
If a $200,000 programme becomes a $400,000 programme, the objective is not necessarily to build exactly the same portfolio at twice the size.
We have another option.
We can build a richer portfolio.
More markets.
More independent system logic.
More places for an outlier to appear.
More ways of encountering it when it does.
This is why compounding in the programme is deliberately restrained.
We do not want every fluctuation in wealth mechanically transmitted into leverage.
Open profits do not increase the sizing base at all.
Realised profits can increase the capital available to the programme, but we prefer meaningful thresholds rather than continually resizing around every incremental change.
And when capital does become available, breadth gets first claim on it.
There is a simple principle underneath all of this:
More capital should first buy us more ways to be right, not merely a larger version of the same bet.
That is weak compounding.
It allows the programme to grow without allowing growth itself to become a source of fragility.
The Cut Back Rule: When the Programme Responds
There is a second layer.
The individual trade has a sizing rule.
The programme has one too.
That is where the Cut Back Rule enters.
The basic position sizing calculation determines exposure trade by trade.
The Cut Back Rule responds to deterioration in the programme’s closed equity by reducing the scale at which those positions are taken.
That distinction is deliberate.
We do not want to punish a portfolio simply because open profits are fluctuating while large trends remain alive.
Unrealised equity needs room to breathe if we are going to harvest outliers.
But when closed equity deteriorates sufficiently, something has changed in the realised path of the programme.
Exposure is reduced.
Not because we know the future will be bad.
Not because the system has stopped working.
And not because we are frightened.
Because survival becomes more important as the capital base becomes impaired.
Then, as the programme recovers, exposure can be rebuilt according to the same predefined machinery.
No prediction is required.
No committee meeting.
No sudden burst of confidence.
The system responds to its own path.
This is the time dimension of position sizing.
A single trade asks:
How much can this position hurt us?
The Cut Back Rule asks:
How much can the programme afford to expose now, given what has already happened?
Different scale.
Same philosophy.
The Formula Must Be Allowed to Be Boring
Position sizing creates two powerful temptations.
After a run of losses, the positions feel too large.
After a run of winners, they feel too small.
That is human.
Recent experience changes our perception of what comes next.
The market has hurt us, so we want less of it.
The market has rewarded us, so we want more.
But notice what has happened.
We have turned a normalisation mechanism into an opinion.
That is precisely what the formula was designed to avoid.
If the programme contains a mechanical response to drawdown, use it.
If it contains systematic rules for changing exposure, use them.
But once we begin overriding the formula because the current environment feels unusually frightening or unusually attractive, the return stream is no longer being generated by the system we researched.
It is being generated by the system plus us.
And we are not neutral additions.
We arrive carrying recency bias, loss aversion, overconfidence, fear, regret and a remarkable ability to discover compelling reasons why this time the rule should not apply.
The formula has none of those problems.
It is dumb.
That is useful.
The Seawall
So return to the wall.
It does not predict the next storm.
It does not know which wave will be largest.
It does not know whether tomorrow’s sea will be calm.
And it certainly does not optimise itself every time the weather changes.
It simply imposes a boundary between an uncertain force and what lies behind it.
That is position sizing.
We cannot control how many trades fail.
We cannot control the order in which losses arrive.
We cannot control when markets gap.
We cannot control when correlations converge.
We cannot control when the great trend finally appears.
What we can control is how much capital we expose while waiting.
That sounds almost disappointingly modest.
But there is a profound asymmetry hiding inside it.
The programme that survives gets another trade.
And another.
And another.
Eventually, perhaps, it gets the outlier.
The programme that optimised itself into extinction does not.
The formula does not find the optimal position.
It finds the survivable one.
And survival is what allows everything else in the system to matter.
Next we move one level higher.
Because sizing one position correctly does not tell us what happens when we hold fifty of them at once.
That is the problem of portfolio construction.
READ DEEPER
→ Position Sizing: The Geometry of Survival
→ Dynamic Position Sizing: Insights, Implications, and the Outlier Hunting Perspective
→ The Cut Back Rule: Engineering Survival in a Fractal World
→ Unlocking the Magic of Fat Tails: Position Sizing and Outlier Hunting
Previous: System Anatomy 2: Exit Logic | Next: System Anatomy 4: Portfolio Construction
Richard Brennan writes on systematic trading, complex adaptive markets, and the philosophical foundations of trend following at atstradingsolutions.com. His books include The Fractals of Finance, Complex Adaptive Markets, Carved by Impossibility and The Aussie Turtles Trend Following Guide.
Want to explore why structure exists at all?
Carved by Impossibility: What Remains When Everything Else Is Eliminated
The book explores the architecture of constraint, emergence, and reality itself, and what it means for how we understand markets, life, and the universe.
Available now on Amazon in paperback, hardcover, and Kindle.
Want the theoretical foundation for why trend following works?
The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.