Why a 42% position can be 68% of your risk

A broker app tells you what you own: each position, its value, its share of the total. It does not tell you where your risk comes from, and the two are almost never the same. This guide shows the gap on one real portfolio — the demo portfolio, four invented companies with six years of generated prices — and explains the three numbers that measure it.

The figures here are the ones on the demo's Risk tab: today's weights, measured over the whole price history from January 2019 to December 2024.

Where the money is

The demo portfolio holds four companies:

Holding	Weight	Volatility on its own
Helix Semiconductors (HLX)	42.0%	50.8%
Keystone Payments (KYP)	25.1%	30.3%
Arbor Industrial (ARB)	24.0%	20.1%
Verdant Utilities (VRD)	8.8%	13.3%

A glance at the weights says: fairly concentrated in one name, but three real positions and a small fourth. That is the picture a broker app gives.

Where the risk comes from

Here is the same portfolio, measured by risk share — how much of the portfolio's volatility each holding is responsible for:

Holding	Weight	Risk share
HLX	42.0%	68.0%
KYP	25.1%	21.7%
ARB	24.0%	10.4%
VRD	8.8%	−0.1%

Helix is 42% of the money and 68% of the risk. Arbor, almost the same amount of money as Keystone, carries half its risk. And Verdant's share is slightly negative: it doesn't just add little risk, it takes a little away.

Why the two differ

Risk share comes from splitting the portfolio's volatility exactly among its holdings — Euler's decomposition. For each holding it multiplies three things: how much of it you own, how volatile it is, and how much it moves with the portfolio as a whole.

• RC_i — the risk share of holding i; the shares of all holdings add up to exactly 1
• w_i — the weight of holding i in the portfolio
• Σ — the covariance matrix of the holdings' returns: each one's variance, and how each pair moves together
• (Σ w)_i — the covariance of holding i with the whole portfolio
• σ_p² — the variance of the portfolio

Helix scores high on all three counts. It is the biggest position, by far the most volatile (50.8% a year against 20–30% for the others), and it moves closely with Keystone — their correlation is 0.79. When Helix falls, the second-largest holding tends to fall with it. The portfolio feels Helix twice.

Verdant is the opposite: small, calm (13.3%), and moving slightly against the two big holdings (correlations of −0.10 with Helix and −0.07 with Keystone). On the days Helix and Keystone drop, Verdant tends to hold or rise a little, which is why its contribution is below zero.

What one more point would do

The marginal contribution asks a practical question: if you moved one percentage point of the portfolio into a holding, how much would total volatility change?

For the demo portfolio, one point more of Helix adds about 0.49 points of portfolio volatility. One point more of Arbor adds 0.13. One point more of Verdant lowers it slightly. This is a description of how the portfolio is built, not an instruction — but it is the number to look at before adding to a position, because it measures the addition against what you already hold rather than on its own.

Four holdings, how many bets?

The last question is the one that surprises people most: how diversified is this portfolio, really?

Counting holdings says four. Counting money says a bit more than three — the portfolio's weights are equivalent to 3.3 equal positions (equivalent positions). But counting the independent directions the holdings actually move in, the answer is 1.05 (independent bets).

That is not a typo. Helix dominates the volatility, Keystone moves with Helix, and Arbor moves partly with both; almost all of the portfolio's movement is one shared movement. Owning four names spreads the money. It barely spreads the risk.

The diversification ratio puts a number on the same thing from another angle. If the four holdings moved in perfect lockstep, the portfolio's volatility would simply be the weighted average of theirs: 35.0%. It is actually 30.6%. The ratio of the two is 1.14 — diversification cuts volatility by about an eighth. A ratio of 1.00 would mean none at all.

The equal-risk ruler

One way to read all of this is to ask what the weights would be if every holding carried the same share of the risk. For the demo portfolio:

Holding	Actual weight	Equal-risk weight
HLX	42.0%	9.2%
KYP	25.1%	15.5%
ARB	24.0%	22.8%
VRD	8.8%	52.4%

This is a ruler, not a recommendation. There can be good reasons to hold far more of a volatile company than an equal-risk split would — conviction, taxes, what else you own elsewhere. What the ruler does is show how far the portfolio sits from spreading its risk evenly, in the units that matter: in an equal-risk split, Helix would be about a fifth of its current size — 9.2% against 42.0%.

What these numbers do not say

• They depend on the window. Volatilities and correlations are measured over 2019–2024. Correlations in particular rise in crises — for the demo portfolio, the average correlation between holdings was 0.29 normally and 0.48 during 2022. A risk share measured in calm years understates concentration in a bad one.
• They are not a forecast. They describe how the holdings behaved together, which is the best available guide to how they will, and not a guarantee.
• Negative is not free. Verdant's negative risk share is real, but it is small, and it comes from a correlation close to zero rather than strongly negative.

Seeing it on your own portfolio

The Risk tab of the dashboard computes these numbers for your own holdings: weight next to risk share, the correlation heat map, independent bets and the equal-risk weights. See them on the demo, or import a broker CSV and measure your own.