How diversification is measured in portfolio management

How diversification actually gets measured

Two holdings can be given a single score for how closely they move together. That score is correlation, and it runs from minus one, meaning they move in exact opposite directions, through zero, meaning no relationship at all, to plus one, meaning they move together almost every time. Correlation between assets is the mechanism diversification actually depends on. A correlation near zero means two positions cushion each other when one falls. A correlation near plus one means they do not, no matter how different their names look.

This is not an abstract concern in crypto. Bitcoin and the total altcoin market carried a 30-day rolling price correlation of 0.92, a reading close to moving in lockstep, on Newhedge's tracker, read on 2026-09-10. A portfolio holding Bitcoin plus four other large-cap coins, at that kind of correlation, is not running five positions. Most of it is one position on Bitcoin's direction, split across five names, and correlation risk inside a crypto portfolio is exactly this gap between how spread out a portfolio looks and how it actually behaves.

Why a correlation matrix beats counting holdings

A single pairwise correlation is a start, not the answer, because a real portfolio holds more than two things. A correlation matrix is the same reading done across every pair at once, so a five-position portfolio produces ten pairwise numbers rather than one. Read across a full matrix, and clusters show up that a holdings count never would: three coins that each show a correlation above 0.8 to one another are one cluster wearing three tickers, even if nothing else links them on the surface. A holding with a low or negative correlation to the rest of the matrix is the one actually doing diversification's job, and it is rarely the fifth or sixth position added to an already-correlated portfolio.

This is the same insight modern portfolio theory formalized: combining holdings that do not move together can push a portfolio toward the efficient frontier, the best possible return for a given level of risk. Modern portfolio theory gave the idea its math. A correlation matrix is how that math gets applied to a real portfolio instead of staying theory.

Position sizing turns the matrix into a rule

A correlation matrix tells you what is true about a portfolio. Position sizing is what you do about it: setting a ceiling on how much of a portfolio any single holding, or any cluster of correlated holdings, is allowed to reach.

Take a portfolio of ₹1 lakh split five ways: ₹20,000 each into Bitcoin and four other large-cap coins. If the matrix shows those four coins move with Bitcoin at 0.85 or higher, the useful question is how much of the ₹1 lakh sits in that whole Bitcoin-linked cluster, not how much sits in any one of the five. Here, the answer is all five positions: the cluster is 100 percent of the portfolio regardless of the five-way split. A ceiling applied at the cluster level, not the position level, is what actually controls the damage.

Qatobit's QSI GEQ8 index applies the same discipline as a published rule rather than a judgment call made position by position: no single holding is allowed to cross 15 percent of the basket or fall below 3 percent, a cap set by the methodology and enforced at each rebalance rather than decided fresh each time. The same logic applies to a cluster of correlated crypto holdings: a written ceiling fixes in advance how much any one shared driver is allowed to cost.

Rebalancing bands: calendar versus threshold

A ceiling only holds if something enforces it as prices move, and that is what rebalancing does: selling down whatever grew past its target weight and adding to whatever fell below it, bringing the portfolio back to the ceiling rather than letting it drift past it.

There are two disciplines for deciding when to do this. Calendar rebalancing acts on a fixed date, monthly or quarterly, regardless of how far anything has drifted on that date. Threshold rebalancing, built on rebalancing bands, acts whenever a holding moves past a set distance from its target weight, regardless of the day. The difference between the two is what triggers the response: a date, or a distance.

Qatobit's QSI indices rebalance on a monthly calendar rather than a threshold. Within that monthly cycle, QSI Core and QSI Growth also carry a Gold allocation the methodology trims at highs and buys crypto at troughs, a counter-cyclical mechanic built into the rebalance rather than a call made in the moment. Neither method removes drift: a calendar can let a position run for a stretch between rebalances, and a tight band can trigger frequent adjustments that each add cost. The choice is which kind of drift a portfolio is built to tolerate.

What drift does to a portfolio nobody rebalances

Go back to the ₹1 lakh portfolio, simplified to two positions of ₹50,000 each. Left alone through a stretch where the first holding roughly triples and the second falls by half, the split is no longer even. It is now closer to ₹1,50,000 and ₹25,000, roughly 86 percent to 14 percent, without a single trade being made. The portfolio drifted into a concentrated position purely because nobody rebalanced it back.

That drift changes what a later fall in the winning position costs. At the original 50-50 split, a 40 percent fall in either holding costs 20 percent of the portfolio. At the drifted 86-14 split, the same 40 percent fall costs roughly 34 percent, because so much more of the portfolio now rides on that one position. Drawdown is not only a function of what markets do. It is a function of how much drift built up before markets tested it.

Every piece covered so far exists to manage this one outcome: maximum drawdown, how far a portfolio falls from its peak before it recovers. Ten correlated holdings can still produce a large maximum drawdown, because correlated positions fall together regardless of the count. A portfolio held to a sizing ceiling and rebalanced back to it regularly cannot let one holding, or one correlated cluster, do that much damage alone, because the ceiling was set before the fall happened. None of this changes whether a portfolio falls in a given period, only how much of that fall traces back to one holding that grew too large rather than to the market as a whole.

Where the measurement breaks down

Correlation is not a fixed property of two assets. It is a statistic computed over a lookback window, 30 days or 90 days, say, and it moves as that window moves. A matrix built in a calm stretch can understate how connected two holdings really are, because calm markets let unrelated stories dominate short-term price moves.

The failure case matters more than the general point. Correlations that sit lower in ordinary conditions tend to rise toward one during a sharp sell-off, because the same forces, a liquidity squeeze and a rush to cash, hit most risk assets at once. A portfolio whose matrix looked spread out a month earlier can behave like one large position on exactly the day it needed not to, a property of markets under stress rather than a flaw in the calculation, and no rebalancing band repeals it. A published methodology fixes the response in advance. It does not change how correlation behaves once the stress arrives.

What this means for measuring your own portfolio

Checking whether a portfolio is diversified is a measurement exercise, not an impression. Pull the correlation between every pair of holdings, not a glance at how many tickers appear on a statement. Size any correlated cluster to a stated ceiling before a fall tests it. Rebalance back to that ceiling on a calendar or a band, and expect the correlations behind it to shift exactly when the portfolio needs them not to.

What diversification in investment actually protects you from covers the concept this piece assumes: why diversification limits one holding's damage rather than lowering the odds of a loss. Four crypto indexes, one methodology walks through how QSI Core, Growth, VRION and GEQ8 apply position caps and rebalancing differently. And how much of my portfolio should be in crypto is the sizing question that comes before this one: how large a crypto allocation should be before its internal diversification is worth measuring at all.

Frequently asked questions

How do you measure diversification in a portfolio?

Diversification is measured with correlation, a score from minus one to plus one for how closely two holdings move together, read across every pair in a correlation matrix. A portfolio with many holdings that all carry a high correlation to each other is not well diversified, whatever the count on the statement says.

What is the difference between calendar and threshold rebalancing?

Calendar rebalancing acts on a fixed date, such as monthly, regardless of how far any holding has drifted. Threshold rebalancing acts whenever a holding moves past a set distance from its target weight, called a rebalancing band, regardless of the date. One is triggered by time, the other by distance.

What happens to a portfolio that is never rebalanced?

Its winners grow into a larger share of the total and its losers shrink, so the original weights drift without a single trade being made. A portfolio that started evenly split can end up concentrated in whichever holding grew the most, which raises how much a later fall in that holding costs the whole portfolio.

What is position sizing in a diversified portfolio?

Position sizing is setting a ceiling on how much of a portfolio any single holding, or any cluster of correlated holdings, is allowed to reach. The ceiling comes from what the correlation matrix shows, not from how many positions exist, since correlated holdings behave as one oversized position even when counted separately.

Does the correlation between assets stay constant over time?

No. Correlation is calculated over a specific window and shifts as conditions change. It tends to rise toward one during a sharp sell-off, when a liquidity squeeze pushes most risk assets down together, precisely when a portfolio most needs its holdings to move independently.

Crypto investments are subject to market risk. Not financial advice.