WhitmanTrading

Position Correlation Calculator

Position correlation measures how much your holdings move together. At zero they are separate bets and the risk grows with the square root of the count; at one they are a single trade in several accounts. Correlation, not the number of positions, sets your real exposure.

What your positions actually diversify

Enter the number of positions, the risk on each, and how much they move together.

Effective risk 2.90%
Equivalent independent positions 1.07
Risk if they were independent 1.73%
Naive sum 3.00%

Effective risk is r × √(n + n(n−1)ρ). The equivalent-independent figure is the diversification ratio squared, and it answers the question people think they are asking when they count positions.

Runs entirely in your browser. Nothing you type is sent anywhere or stored.

How the number is built

A candlestick chart with two positions moving together.
How much two positions move together. Illustrative chart - not real market data.

Correlation runs from zero to one for this purpose. Combining positions is not addition, because positions that move independently offset each other some of the time and positions that move together never do.

Effective risk = r × √(n + n(n−1)ρ)

The first half of the price series with identical positions.
At one they are the same trade in two accounts. Illustrative chart - not real market data.

At ρ = 1 that collapses to r × n, the naive sum — because identical positions are one position held several times.

The second half of the price series with independent positions.
At zero they are genuinely separate bets. Illustrative chart - not real market data.

At ρ = 0 it becomes r × √n, which for three positions is 1.73% rather than 3%. That reduction is the entire mathematical benefit of diversification.

A worked example

Take the defaults: three positions, 1% each, average correlation 0.9.

The naive sum is 3.00%.

The effective risk is 1 × √(3 + 6 × 0.9) = √8.4 = 2.90%.

If they were independent it would be 1.73%.

And the equivalent-independent figure is 1.07. Three separate trades, three separate charts, three separate decisions — carrying the diversification of one position and a very small fraction of another.

A section of the price series showing combined exposure.
So correlation decides your real risk, not the count. Illustrative chart - not real market data.

That last output is the one worth carrying. “How many positions do I have” is the question people ask; “how many independent positions do I have” is the question they mean, and the two answers can be three and one.

Where the correlation comes from

Positions correlate for reasons that are not visible on the ticker. Two miners are one commodity bet. A bank and a homebuilder are one interest-rate bet. Two different indices in the same region are close to the same instrument.

And direction compounds it. Six long positions in different names are more correlated than three long and three short in the same names, because the market factor is common to all six.

A useful shortcut when you have no data: if the same news would move all of them the same way, treat the correlation as high and check whether the effective risk figure is one you accepted.

A window of price bars over a chosen measurement period.
The period you measure over changes the answer. Illustrative chart - not real market data.

Estimating it without a data feed

Most people running this calculation do not have a correlation matrix, and the honest answer is that a rough estimate applied consistently beats a precise one never applied.

Start at 0.5 for anything in the same market. Two shares on the same exchange share the market factor, and that alone puts them well above zero.

Add to it for a shared sector. Two banks, two miners or two housebuilders move on the same news before they move on their own. 0.7 is a defensible working figure.

Add again for shared direction. All-long positions are more correlated than a mixed book, because a general fall reaches every one of them.

And go to 0.9 or above when the same headline would move all of them. That is the practical test, and it needs no data at all — if you can name one event that would hit every position, the correlation is high whatever a historical calculation says.

Then run the calculator twice: once at your estimate and once at 1.0. The gap between those two answers is the size of the assumption you are making, and if the second number is unacceptable the estimate was never the thing protecting you.

Why the measurement is unstable

A sideways, range-bound candlestick series.
And it rises towards one exactly when it matters. Illustrative chart - not real market data.

Correlation is not a property of two assets; it is a property of a period. Measured over a calm year it can be low, and the same pair in a shock can move together almost exactly.

A long-horizon candlestick view averaging over a long period.
A long window hides the week that hurt you. Illustrative chart - not real market data.

A long measurement window averages away the episodes you are protecting against. Two assets with a 0.2 correlation over five years may have spent one week at 0.95, and that week is the one your risk figure needed to describe.

Which means the honest way to use this calculator is twice. Once with your measured correlation, to see the ordinary case. Once with correlation set to 1, to see what a shock does. If the second number would end you, the position count is wrong regardless of what the first one says.

The original data

Of the 24,971 unique videos in research/search-study-corpus.jsonl, zero have an instruction-shaped title about position correlation. Correlation generally appears in 5 instruction-shaped titles at a median of 1,162 views across 5 channels, and position sizing in 195 at 1,738. The counts come from site/rank_tools2.py.

A candlestick chart with a volume histogram beneath it.
Participation is not part of the calculation. Illustrative chart - not real market data.

Five videos on correlation at a median of 1,162 views is the lowest demand measured for any subject on this site. That is worth stating plainly rather than dressing up: almost nobody searches for this. It is here because the arithmetic is load-bearing, not because the audience is asking — and it is the one page in this section built on principle rather than on measured demand.

Price bars cut short at a diversification decision.
Three positions, all 0.9 correlated. Diversified? Illustrative chart - not real market data.

The answer to the question on that chart is no, and the figure is 1.07. Three positions at 0.9 carry the diversification of one — so the account is exposed as though it held a single trade at nearly three times the intended size, arrived at through three individually reasonable decisions.

When it fails

A candlestick series with several gaps, the largest marked.
In a shock everything correlates and nothing diversifies. Illustrative chart - not real market data.

The number you enter is historical and the number that matters is the future one. In a genuine shock, correlations across almost every risk asset move towards one at the same moment, so a portfolio measured at 0.3 behaves like one at 0.9 on the day it counts. This is not a flaw in the arithmetic — it is the reason the second run with correlation set to 1 is the one that should size the account.

The second failure is treating different tickers as independent. The ticker is not the exposure.

A third is a single average correlation across a mixed book. One pair at 0.95 inside an otherwise uncorrelated set is not visible in an average.

A fourth is a long measurement window. It averages away the episodes the calculation exists for.

A fifth is ignoring direction. All-long positions share the market factor whatever else differs.

A candlestick chart annotated with the round-trip cost.
And unwinding several at once costs 2% of a bar each. Illustrative chart - not real market data.

And a sixth is forgetting the exit. Correlated positions need closing at the same moment, and each one costs a round trip — 2% of a median bar’s range on this site’s shared series, paid simultaneously in conditions where the market is least willing to take the other side.

Correlation covers the measure itself and why it is unstable. Portfolio building is where these decisions sit as an allocation. And risk management is the wider set of controls this one feeds.

What I actually do

Different tickers is the illusion I fell for longest. Five names, five charts, five separate decisions — and all five were the same bet on one interest-rate expectation. They went down together on the same afternoon, which is the only test that was ever going to reveal it.

— Michael Whitman

This page is educational, not financial advice. Test every idea on your own charts before risking money.