Money management and risk at the portfolio level
The answer
Portfolio risk is not the sum of your position risks. It is the sum of the position risks that share a driver, because those arrive together. The first job is to add up total open risk, the second is to group that risk by what would damage all of it at once, and the third is to have a written rule that reduces exposure after a decline instead of during one.
Why this costs you money
You have probably already done the position-sizing arithmetic. One percent of the account per position, stop from the chart, share count from the division. That is the right calculation and it is covered properly in cluster 16.
Now here is what it does not protect you from.
Ten positions at 1% risk each are not 10 independent 1% risks. If 8 of them respond to the same interest rate, the same currency, the same commodity or the same source of foreign buying, then one event takes 8 of them to their exits on the same morning. Your stated risk was 1%. Your actual risk was 8%, and nothing on any screen told you.
This is the most common way a carefully risk-managed account still suffers a large loss. The person did the arithmetic. They did it on the wrong unit.
The second failure is worse and it is about timing. Most people reduce risk after a drawdown, when their confidence has fallen. That is the exact moment when a recovery is most likely and when reduced size means you participate less in it. Meanwhile, during the calm period before the loss, confidence was high and position sizes were largest. So exposure is systematically highest before losses and lowest before recoveries, and this pattern is produced by feelings rather than by a rule.
The third failure is the one nobody counts. Your money sits with a broker. Your shares sit in a depository. Your funds sit with a fund house. Every one of those is a counterparty, and a counterparty failure is a portfolio risk that no stop order addresses. Most investors have never asked what happens to their holdings if their broker fails, and the answer is different in India and the United States.
How it works
Portfolio heat: the number to know
Portfolio heat is the total amount you lose if every open position reaches its exit at once.
For each position: shares × (current price − exit price). Add them all. Divide by the total account value including cash. That percentage is your heat.
Most professional systems cap heat somewhere between 5% and 10%. The number matters less than having one, and less than the fact that almost no retail investor can state theirs.
Two refinements make it honest.
Use the current price, not the entry price. A position that has risen and whose exit has been raised behind it now risks less than it did at entry. Sometimes it risks nothing. Using entry prices overstates heat on winners and understates it on losers.
Add the positions with no exit at all. If a holding has no written exit price, its risk is not zero. Its risk is the entire position. Either write an exit or count the whole thing.
Correlation: why heat alone is not enough
Heat treats every position as separate. Positions are not separate.
The correction is to group them by driver — the thing that would move them all at once. Sector labels are a poor proxy, because 2 companies in different sectors can share a driver, and 2 in the same sector can be affected differently.
A workable grouping for most portfolios:
- Sensitive to interest rates, including lenders and anything valued on distant future profits
- Dependent on a single commodity, as producer or as consumer
- Dependent on foreign revenue or on a currency
- Dependent on domestic consumer demand
- Small and illiquid, meaning you cannot exit quickly regardless of the reason
- Leveraged, including derivatives positions and anything bought on margin
A position may sit in 2 groups. Total the heat in each group. The largest group total is the number that describes your risk, not the overall heat.
And Francois Longin and Bruno Solnik's finding in the Journal of Finance in 2001 applies directly here. Correlations between markets rise in the tails, and the increase appears in declines rather than in rises. So the groups you constructed from ordinary-period behaviour will merge further on the worst days. Assume the grouping understates the problem.
Kelly, and why nobody uses full Kelly
John Kelly published a formula in the Bell System Technical Journal in 1956 for the fraction of capital to stake on a repeated favourable bet in order to maximise the long-run growth rate.
For a simple bet that wins with probability p and pays b times the stake, the optimal fraction is:
f = (b × p − q) ÷ b, where q = 1 − p.
Edward Thorp applied the idea to blackjack and later to markets, and wrote about the application in detail.
Three things matter about Kelly for an investor.
It maximises growth, not comfort. Full Kelly produces very large swings. A Kelly-sized portfolio can be expected to experience declines of 50% or more as a routine event.
It requires you to know your edge. The formula's inputs are your true win rate and your true payoff ratio. You do not know those. You have an estimate from a small sample, and the estimate is usually too high, because losing trades are remembered less precisely than winning ones. Overestimating the edge makes the formula recommend a size that is not merely suboptimal but destructive, because staking more than the growth-optimal fraction reduces growth and eventually drives capital toward zero.
So practitioners use a fraction of it. Half Kelly or quarter Kelly is the common practice. Half Kelly retains a large share of the growth rate with much smaller swings.
The useful takeaway for a reader is not the formula. It is the shape of the curve. Position size has an optimum, and being above it is worse than being below it. Below the optimum you grow more slowly. Above it you grow more slowly and you can be destroyed. Given that your edge estimate is uncertain and probably optimistic, the correct response to uncertainty is to be smaller than the formula says.
Volatility targeting
A second portfolio-level control is to scale total exposure inversely to recent volatility, so that expected portfolio volatility stays near a chosen level.
The justification is that volatility is persistent while returns are not. A volatile month tends to be followed by a volatile month, so you can forecast risk without forecasting return. Alan Moreira and Tyler Muir published evidence in the Journal of Finance in 2017 that this improved risk-adjusted returns across several asset classes.
The practical retail version needs no calculation. When markets are moving twice as much as usual, take half the position size. That captures most of the benefit.
Drawdown control, and the rule that must be written in advance
The arithmetic of recovery is covered fully in cluster 16 and it is worth restating in one line: a loss of L requires a gain of L ÷ (1 − L) to recover. A 20% loss needs 25%. A 50% loss needs 100%.
The portfolio-level response is a de-risking ladder: a written rule that reduces exposure at defined levels of account drawdown.
An example structure, and the numbers are yours to choose:
| Account drawdown from its high | Action |
|---|---|
| 5% | No action. This is normal. |
| 10% | New positions at half size. No additions to existing positions. |
| 15% | Close the weakest third of positions by relative strength. |
| 20% | Reduce to a minimum core. Take no new positions for a defined period. |
Two things make this work and both are about timing.
It must be written before the drawdown. A rule invented during a decline is a decision made under stress and it will be either too aggressive or ignored.
It must have a re-entry condition, also written in advance. A de-risking rule without one converts a drawdown into a permanent exit. The most common version is a return to a stated level of the account's high, or the broad market returning above its long-term trend filter.
Risk of ruin
Risk of ruin is the probability that a sequence of losses reduces the account below a level from which it cannot recover.
The precise formula depends on assumptions that do not hold in markets. The qualitative result is what you need, and it does not depend on the assumptions: the probability of ruin rises steeply with the fraction risked per event, and falls with the number of independent events.
That gives 2 levers and only 2. Risk less per event, or make your events more independent. Being right more often is a third lever and it is far weaker than either, which is why the professional emphasis sits where it does.
What it tells you, and what it does not
These measurements tell you your worst plausible day, before it arrives. That is the entire purpose. A portfolio's risk is knowable in advance and almost nobody computes it.
They tell you which of your positions are actually the same position, which is usually a surprise.
They do not tell you the worst possible day. Heat assumes every exit executes at the exit price. Gaps, price bands and halts break that assumption, and the real loss in a severe event is larger than the computed heat.
They do not include counterparty failure, operational failure or fraud. Those are portfolio risks and they are not in the arithmetic.
They do not tell you your edge. Every sizing formula takes the edge as an input, and if the input is wrong the output is confidently wrong.
And they cannot fix an account that is too small for the method. One percent of a small account is an amount that many exits cannot be constructed around. The honest answer for a small account is fewer positions, wider exits and lower frequency, not a smaller percentage applied to an impossible number.
The decision rule
Compute heat, group it by driver, and cap the group rather than the position.
If your total open risk is under your limit and no single driver group exceeds about half of that limit, the portfolio is within its design.
If one driver group holds most of your risk, the correct action is not to add a hedge. It is to reduce the group. A hedge adds a second position with its own failure mode and its own correlation, and hedges fail in the conditions they were bought for.
Unless you have deliberately chosen a concentrated position and sized the whole account for it. That is a legitimate decision. It has to be a decision, written down, not an accumulation.
And when the account falls, follow the ladder you wrote before it fell. The single most expensive moment in investing is the point where a person decides their pre-written rule does not apply to this particular decline.
Try this now
Fifteen minutes. This produces the 2 numbers that describe your actual risk, and most people have never seen either of them for their own account.
- Open your holdings. For each position, write 3 things: current value, your written exit price, and shares held.
- For any position with no written exit price, write the whole position value in the risk column. Its risk is not zero.
- For every other position, compute
shares × (current price − exit price). That is that position's money at risk. Write it down. - Add all the risk figures. Divide by your total account value including cash. This is your portfolio heat. Write it down with today's date.
- Now group. Beside each position write which of these it belongs to: rate-sensitive, single commodity, foreign revenue or currency, domestic consumption, small and illiquid, leveraged. A position can carry 2 labels.
- Total the risk inside each group. Divide each by the account value.
- Write down the largest group percentage. That is your real exposure to a single event.
- One more line. Write the name of your broker, and the total value held with that 1 broker as a percentage of your investable money.
What you should see. Heat above 10% is common and most people are surprised by it. The larger surprise is step 7. In most portfolios 1 driver group holds between half and three-quarters of the total risk, and the person had understood themselves to be diversified because the position count was high.
Step 8 is the one nobody does. If more than about 80% of your money sits with 1 broker or 1 platform, you have a concentration that has nothing to do with markets and that no exit rule addresses.
Finally, write 1 sentence: what event takes my largest driver group to its exits at once? If you can name the event, you can decide whether you want that exposure. If you cannot name it, you do not yet know what you own.
Three real cases
1. Amaranth Advisors, September 2006 — concentration disguised as a spread Amaranth Advisors was a multi-strategy hedge fund. In September 2006 it lost approximately $6 billion in natural gas futures positions over a period of weeks and closed shortly afterwards. Ludwig Chincarini published a detailed analysis in the Journal of Applied Finance in 2008.
The positions were largely spread trades between natural gas contracts for different delivery months. A spread looks like a hedged position, because one leg is long and the other is short. What the structure disguised was that both legs depended on the same relationship, that the position was extremely large relative to the market's open interest, and that exiting would move the price against the fund.
The lesson for a personal portfolio is exact. A position that contains an internal hedge is not thereby diversified. And any position large enough that exiting it moves the price is a position whose stated exit price is a hope.
2. The VelocityShares Daily Inverse VIX Short-Term ETN, 5 February 2018 — when the risk is in the structure and not the direction This exchange traded note, listed in the United States and commonly known by the ticker XIV, was designed to deliver the inverse of the daily return of a short-term volatility futures index. It had performed strongly for several years during a period of low, declining volatility.
On 5 February 2018 volatility rose extremely sharply in a single session. The note's indicative value fell by approximately 96%, an acceleration event in its terms was triggered, and the issuer redeemed and terminated the product.
Three portfolio lessons sit in this case.
The instrument did exactly what its documentation said it would do. Nobody was defrauded. The loss came from a feature disclosed in the prospectus that most holders had not read.
A daily-reset inverse product does not deliver the inverse of a long-term move, and a position that can lose most of its value in 1 session cannot be protected by a stop order, because there is no price in between.
And the position's history was the trap. Several years of smooth gains produced confidence, and confidence produced size.
3. Karvy Stock Broking, November 2019, India — the risk that is not in any chart In November 2019 SEBI passed an interim order against Karvy Stock Broking Limited, restraining it from taking new clients and from using client securities. The regulator found that client securities had been pledged to raise funds that were transferred to a group entity. Client holdings were later transferred back to the rightful account holders through the depositories over a period of months.
No chart contained this risk. No stop order addressed it. No amount of diversification across shares helped, because the exposure was to the intermediary, not to the market.
The protections that did matter were procedural, and every reader can use them. Holdings in a demat account are held with a depository, and the depositories send statements directly to the investor. A client can check their own holdings independently of what the broker's application displays. Power of attorney arrangements and pledge instructions are the mechanisms through which this kind of loss occurs, and they can be reviewed.
The question that resolves it
A novice asks: how much am I risking on this trade?
An expert asks: how much am I risking on this driver, across every position I hold, including the ones I forgot are exposed to it?
The move from the first question to the second is the whole of portfolio risk management. It converts a per-position discipline, which most careful people already have, into a per-event discipline, which almost nobody has.
There is a fast way to ask it. Name a single plausible event — a large rise in interest rates, a currency fall, a commodity spike, a foreign selling wave — and then go through your holdings and mark every one that the event touches. The marked total is the answer, and the exercise takes 3 minutes.
What would make this wrong
The claim is that grouped risk beats position risk, that size has an optimum above which it is destructive, and that de-risking rules must be written in advance.
The grouping claim would be wrong if correlations between positions were stable and low. They are not. The evidence that correlations rise in declines is well documented and it goes the wrong way for anybody hoping otherwise.
The Kelly claim rests on knowing your edge, and this is the honest weakness of the whole framework. Every sizing method takes an edge estimate as an input. If your estimate comes from a small sample of your own trades, it is unreliable, and possibly the strategy has no edge at all. In that case the correct size is zero and no formula will tell you so. Sizing mathematics answers how much to bet given an edge. It cannot tell you whether you have one.
The de-risking claim is genuinely contested. Reducing exposure after a drawdown means participating less in the recovery, and recoveries have often been fast. A reasonable critic argues that a mechanical de-risking ladder converts temporary declines into permanent underperformance, and there is evidence for that in markets that fall and rebound quickly. The counter-argument is not statistical. It is that a rule you follow is worth more than an optimal policy you abandon at the worst moment. If you would genuinely hold through a 50% decline without selling, you do not need a ladder. Most people believe they would and most people do not.
And the honest limit. All of these controls reduce the size of losses. None of them creates returns. A portfolio with excellent risk management and no edge converges slowly toward its costs instead of quickly toward zero. That is a real improvement and it is not a strategy.
In India
Five Indian features change portfolio risk in ways the American literature does not cover.
Price bands can prevent an exit entirely. Individual shares have daily price bands. A share locked at its lower band may have no buyers at all, so no trade occurs and a stop order does not execute. Your computed heat assumed it would. This is why position size, and not the exit rule, is the primary control in India.
Derivative lot sizes make small risk impossible in some instruments. The exchange sets the lot size. If 1 lot risks more than your limit, the position is not available to you at an acceptable size, and the correct action is to skip it.
Peak margin and upfront margin rules limit intraday leverage. SEBI phased in upfront margin collection and peak margin reporting requirements, which substantially reduced the intraday leverage brokers could offer. This is a regulator imposing a portfolio risk limit that most traders would not have imposed on themselves.
Broker and custody risk is addressable and rarely addressed. Shares are held in a demat account with a depository, either CDSL or NSDL, and both send holding statements directly to the investor. SEBI has restricted the use of client securities and moved the market toward pledge mechanisms that require the client's authorisation. Read the depository statement, not only the broker application.
The evidence about outcomes is unusually direct. SEBI has published studies of individual traders in the equity derivatives segment showing that a large majority lost money, across multiple financial years. Those losses are not mainly analytical failures. They are sizing failures in a leveraged instrument.
Taxation matters at portfolio level too. Short-term and long-term capital gains on listed equity are taxed at different rates with the boundary at a holding period of more than 12 months, and rates were revised in July 2024. Set-off and carry-forward rules for capital losses affect the after-tax value of a de-risking decision.
In the United States
The United States has more instruments, more leverage and a different set of protections.
Standardised leverage, and regulated limits. Regulation T sets an initial margin requirement for share purchases and brokers set maintenance requirements above the regulatory minimum. Portfolio margin is available to qualifying accounts and permits substantially more leverage based on a risk model of the whole portfolio.
A rule that forces position sizing on active traders. FINRA's pattern day trader rule requires an account executing 4 or more day trades within 5 business days to maintain a minimum equity level. This is a regulator imposing a minimum account size for a strategy, and it exists because the strategy destroys small accounts.
Halts are designed to reopen. Individual shares are covered by volatility pause mechanisms that stop trading briefly and resume with wider bands. An exit is more likely to execute than in a market with daily locks, though the price is not guaranteed.
Account protection has a specific and limited meaning. The Securities Investor Protection Corporation covers customer assets if a member brokerage fails, up to stated limits. It protects against the failure of the intermediary. It does not protect against investment losses, and confusing the 2 is common.
Tax interacts with risk reduction. The wash sale rule disallows a loss for tax purposes when a substantially identical security is repurchased within 30 days before or after the sale. A de-risking rule that exits and re-enters inside a month has tax consequences the arithmetic did not include.
Where they differ, and what that tells you
The exit is a stronger promise in one market than the other, and that reorders your controls. American halts reopen and fractional shares allow exact sizing, so an exit rule is the primary protection there. Indian price bands can prevent any trade at all, so position size is the primary protection here and the exit is the backup. An Indian reader who learns the American ordering is protecting themselves in the wrong sequence, and will find out during the one event where it matters.
Leverage is available in different forms and the risk arrives differently. American margin is standardised, disclosed and applies to a portfolio that can usually be liquidated. Indian retail leverage sits mostly inside derivatives, in fixed lot sizes, with expiry dates. A derivatives position does not merely fall, it expires, which means time itself is a risk that no share position has.
Intermediary risk has different remedies. The United States has an industry-funded protection scheme for the failure of a brokerage, within limits. India's protection runs mainly through the depository system, in which securities are held in the investor's own name and can be verified directly with CDSL or NSDL, plus exchange investor protection funds. Neither system protects against market losses. Both require the investor to actually check their own statements, and in India the depository statement is the independent record. Reading it once a month costs nothing and is the only portfolio risk control in this article that requires no arithmetic at all.
Carry this
- Portfolio heat is the total you lose if every open position reaches its exit. Compute it. Write it down with a date.
- A position with no written exit risks its entire value, not zero.
- Group risk by driver, not by sector name. The largest group is your real exposure.
- Position size has an optimum. Being above it is worse than being below it, and your edge estimate is probably too high.
- Write the de-risking ladder and the re-entry condition before the drawdown. A rule invented during one will not be followed.
- Count your broker as a position. Read the depository statement, not the app.