Risk management: surviving long enough to win

The answer

Risk management is not about avoiding losses. It is about making sure that the ordinary, expected run of losses cannot end your participation. A method with a genuine edge still fails if the account is not built to sit through the losing runs the method produces.

Why this costs you money

Cluster 16 covered the arithmetic of a single position: risk 1% of the account, place the stop where your reason stops being true, cap the position value. This article is about what happens across many positions, over months, and it is where most accounts actually die.

The failure is not one large loss. It is this sequence.

A method wins 45% of the time with an average win twice the average loss. That is a good method. Over 200 trades it makes money reliably.

It also produces, somewhere in those 200 trades, a run of 9 or 10 consecutive losses. Not because anything went wrong. A 55% chance of losing, repeated, will produce a run of 9 within a few hundred trades. It is expected output.

At 1% risk, that run costs about 9% of the account. Unpleasant and survivable.

But almost nobody sits through it at the same size. What happens on trade 6 or 7 is that the trader concludes the method has stopped working. They reduce size, or stop, or change the rules. The recovery, when it comes, is taken at half size or missed completely.

The loss was survivable. The response to it was not. That is the shape of almost every retail failure, and no amount of stock selection prevents it.

How it works

Three quantities describe the risk of a method. Most traders know none of them about their own trading.

1. Risk of ruin. The probability that a sequence of losses reduces the account below the level where you stop, before the edge has time to appear. It depends on 3 inputs: the win rate, the ratio of average win to average loss, and the fraction of capital risked per position. The third input is the only one you control directly, and it has the largest effect.

The relationship is not gentle. Doubling the risk per position does not double the risk of ruin. Past a certain point it converts a profitable method into a losing one, because the compounding of losses outruns the compounding of gains. This is why the 1% convention exists.

2. Maximum drawdown. The largest fall from a peak in the account value to the low that follows it, measured as a percentage. Every method has one, and the one you have already seen is almost never the largest one available. If your worst drawdown in 2 years is 12%, the correct planning assumption is that 25% is coming.

3. Drawdown duration. The number of days or months between the old peak and the return to it. This is the quantity that ends careers, and nobody measures it. A 15% drawdown that recovers in 3 weeks is an inconvenience. A 15% drawdown that takes 14 months to recover is 14 months of following rules that appear not to work, while doing nothing else with the time.

Duration, not depth, is what people cannot absorb.

Loss limits are mechanisms, not virtues

A loss limit is a number that stops your activity for a defined period. Three of them, and each has a specific failure it prevents.

Per position. Prevents a single position from mattering. Usually 1%.

Per day. Prevents a bad morning from becoming a bad month. A common setting is 3 times the per-position risk. When it is hit, the session is over.

Per week or per month. Prevents a bad month from ending your participation. A common setting is 6% to 10% of the account. When it is hit, you stop, and you do not restart until you have reviewed every trade in the period.

None of these require willpower on the day, which is the point. The number is set when nothing is at stake. The only thing required later is that you do not override it, and that is a much smaller requirement than deciding well while losing.

The specific failure mode of a loss limit is that it can be overridden with one click. So the limit needs a physical form: an alert set in advance, a note where you will see it, or a broker feature that prevents further orders.

What it tells you, and what it does not

These numbers tell you whether your account can survive your method. They say nothing about whether the method makes money.

That distinction matters because it points at the correct order of work. Most people try to improve the win rate. The win rate is the hardest input to improve and the least useful once it is above about 40%. The risk per position is the easiest to change and it has the largest effect on whether you are still here in 3 years.

The numbers also assume your past trades represent your future trades. They do not, if the market conditions change or if you change the method. A drawdown statistic from a 2-year sample in a rising market is a description of a rising market.

The decision rule

Size your risk so that the worst run you can imagine costs you money and nothing else.

Work backwards from the run, not forwards from the trade:

1. Find your longest losing streak so far. Call it N.
2. Assume a future streak of 2N. That is the planning number.
3. At your current risk per position, calculate what 2N consecutive losses

does to the account.

1. If that number would make you change your behaviour, your risk per position

is too high. Reduce it until the answer is no.

The test is behavioural, not financial. The question is not whether you could afford the loss. It is whether you would still place the next position at the same size.

Try this now

Five minutes with your own trade history. This produces the single most useful number in the cluster.

1. Open your broker's tradebook or profit and loss report and download the last 12 months. Most brokers provide a spreadsheet.
2. Sort by exit date, oldest first.
3. Go down the list and mark each closed trade W for a gain or L for a loss, after costs.
4. Find the longest unbroken run of L. Write that number down. This is N.
5. Now calculate. If you risk 1% per position, a run of 2N costs about 2N% of the account. If you risk 3%, it costs about 3 times that. Do the arithmetic for your actual risk per position.

What you should see. Most people find N is between 4 and 9, and are surprised, because they remember individual losses and not runs.

Then look at the 2N number and ask the only question that matters: would you place the next position at the same size, on the same rule, after that?

If you hesitate, you have found the real limit, and it is not your capital. It is the size at which you can still follow your own instructions. Reduce the risk per position until the answer is yes without hesitation. That number is your actual risk limit, and it is usually much smaller than the one you were using.

If you have fewer than 30 closed trades, you cannot measure N yet. Use 8 as the planning assumption. It is close to what a 45% win-rate method produces, and it is not a worst case.

Three real cases

1. Victor Niederhoffer, 27 October 1997 (United States) — a good record ends in a single session Niederhoffer was a well-known fund manager with a long record. His funds had sold put options on the S&P 500 index, a position that profits in calm markets and loses without limit in a fall. On 27 October 1997 the US market fell sharply during the Asian financial crisis and the positions could not meet margin calls. The funds were closed. The record was real. The position size relative to a fall of that scale was the failure.

2. The XIV note, 5 February 2018 (United States) — the expected event arrives Credit Suisse's VelocityShares Daily Inverse VIX Short-Term exchange traded note was designed to move opposite to short-term volatility futures. On 5 February 2018 volatility rose sharply and the note lost the large majority of its value in a single session. Credit Suisse announced acceleration and the note was redeemed later that month. The product had disclosed that this could happen. Holders had seen 2 calm years and treated the calm as the base case.

3. Optionsellers.com, November 2018 (United States) — losing more than the deposit The firm, run by James Cordier, managed individual accounts that sold options on natural gas futures. A sharp rise in natural gas prices in November 2018 caused losses that exceeded the capital in many client accounts, leaving debit balances. Interactive Brokers, which held the accounts, disclosed a loss of approximately $113 million from customer accounts in that period. Selling options without a defined maximum loss means the position size does not cap the loss. Nothing does.

The question that resolves it

A novice asks: how much can I make on this?

An expert asks: what does the worst 3 months this method can produce look like, and will I still be executing it on the last day of that period?

The second question is answerable from your own records in 5 minutes and it settles the position size without any opinion about any instrument.

What would make this wrong

If losing runs of 8 or more were rare in methods with a positive edge, most of this article would be unnecessary. They are not rare. With a 45% win rate, a run of 8 is close to certain within a few hundred trades. That is arithmetic, not an empirical claim, so it cannot be wrong in the way a forecast can.

The honest limits.

Historical drawdown understates future drawdown. Every measured worst case is the worst case that happened to occur in the sample. Treating it as a limit is the specific error that ends leveraged strategies.

Loss limits can cost money. A per-day limit will sometimes stop you before a recovery in the same session. That is a real cost and it is the price of the protection. A limit that never costs anything is set too loosely to work.

Risk of ruin formulas assume independence. They assume each trade's outcome is unrelated to the last. In practice losses cluster, because the conditions that produce them last for weeks. Real losing runs are longer than the formula suggests, which argues for smaller risk, not larger.

Carry this

• The run of 8 losses is expected output, not evidence that the method broke.
• Assume the next drawdown is twice the worst one you have seen.
• Duration ends more careers than depth. Measure the months, not only the percentage.
• Three limits: per position, per day, per week. Set them when nothing is at stake and give each one a physical form.
• The correct risk per position is the largest one at which you would still place the next trade after 2N losses.

Knowledge check

Related

• Position sizing across a portfolio of trades
• Risk management and position sizing (cluster 16)
• Mastering your emotions while trading
• ← The trader's playbook