Metrics

What is Correlation?

Reviewed by Karthikeyan Anbarasan (Founder, TradeLyser) · Updated 19 Oct 2026 · Editorial policy

DefinitionCorrelation measures how closely two return series move together, on a scale from −1 (opposite) through 0 (unrelated) to +1 (in step). An illustrative stock portfolio whose monthly returns have a correlation of 0.93 with Nifty 50 is, for risk purposes, mostly a bet on the index, however many stocks it holds.
Formula
Correlation (r) = Covariance (X, Y) ÷ (Standard deviation of X × Standard deviation of Y)
Example
Illustrative 12 months: 7.43 ÷ (3.22 × 2.48) = 0.93
12 months · portfolio return vs Nifty 50 return (illustrative)−4%−2%0%+2%+4%−4%−2%0%+2%+4%Nifty 50 monthly returnFitted lineBeta 1.0BETA (SLOPE)1.21CORRELATION0.93INTERCEPT+0.22% a monthEach dot is one month
Figure 1. Illustrative 12 monthly returns of a stock portfolio against Nifty 50. The dots sit close to a straight line, so the correlation is high at 0.93. The slope of that line is a separate number, the beta, here 1.21.

How correlation is calculated

  1. Convert both series to returns for the same periods, such as daily or monthly % changes. Never use raw prices.
  2. Work out the covariance: how far the two returns move together around their averages. In Figure 1 it is 7.43.
  3. Divide by the product of the two standard deviations: 3.22% for the portfolio and 2.48% for Nifty.
  4. Result: 7.43 ÷ (3.22 × 2.48) = 0.93. Squared, it is 0.87: about 87% of the variation in the portfolio’s monthly returns is explained by Nifty’s moves.
+0.7 to +1 · Strong positive
Move together; little diversification.
+0.3 to +0.7 · Moderate
Related but with independent moves.
−0.3 to +0.3 · Weak or none
Largely independent over this window.
Below −0.3 · Negative
Tend to move opposite; can offset each other.

Worked example: two positions, one risk

Illustrative: two open positions, each with a daily swing (standard deviation) of ₹5,000. How much the combined position swings depends entirely on their correlation: √(5,000² + 5,000² + 2 × r × 5,000 × 5,000).

CorrelationCombined daily swingWhat it means
+0.9₹9,747Almost one double-sized position
+0.5₹8,660Some offset, still mostly additive
0₹7,071Genuine diversification
−0.5₹5,000Each position partly hedges the other

A long Nifty futures position and a long Bank Nifty futures position usually sit near the top row. Holding both does not spread risk; it roughly doubles it, which is why open risk should be added up per theme, not per symbol. The same arithmetic applies to strategies. If an opening range breakout and a momentum setup both tend to win on trending days and lose on choppy ones, running both at full size doubles the bet on one kind of day. Correlating the daily P&L of each strategy tag over a few months shows whether they really spread risk or only look different on paper.

How traders use it

  • Portfolio heat: before a new entry, check whether it is highly correlated with something already open, and count the two as one risk.
  • Pairs trading: look for two stocks with high correlation, then trade the spread when it moves unusually far, accepting that the relationship can break.
  • Strategy mix: correlate daily P&L of your strategy tags. Two setups with a correlation of 0.8 are close to the same bet.
  • Hedging: a hedge only works if its correlation with the position is strongly negative over your holding period.

Limits and the beta comparison

Correlation

0.93
How tightly the dots follow a line. Says nothing about how big the moves are.
Range: −1 to +1

Beta

1.21
The slope of the line: size of the move for each 1% in the benchmark.
Range: any value
  • Correlation is not cause. Two series can move together because both respond to something else.
  • It changes with the window: 20-day and 250-day correlations between the same stocks can differ sharply.
  • It tends to rise in sharp sell-offs, so diversification measured in calm months can vanish when it is needed.
  • It captures straight-line relationships only, so an option position’s curved payoff is poorly described by it.
  • Short samples mislead. Twenty data points can show a correlation of 0.5 between two series that are unrelated over a longer period; use at least several months of daily returns before relying on a figure.
  • Price correlation is not return correlation. Two stocks that both trended up for a year will show near-perfect price correlation even if their daily moves were unrelated, which is why the calculation must use returns.
For Indian traders
Banks and financial services make up a large share of the Nifty 50, so Nifty and Bank Nifty positions often move closely together, and a portfolio of large-cap bank stocks behaves much like a Bank Nifty position. Check correlation over your actual holding period before treating them as separate risks.

Reference guide

ContextValueReading
Correlation measured on returns over a stated windowUsableRe-check it periodically; it drifts and often rises in sell-offs
Correlation measured on raw pricesMisleadingTwo rising price series can show a high correlation even if their daily moves are unrelated
Several open positions with high correlationOne big betAdd up their risk as if they were one trade

Questions

How is correlation calculated?

Divide the covariance of two return series by the product of their standard deviations. The result always falls between −1 and +1.

What is a good correlation for diversification?

Lower is better for spreading risk: near zero means largely independent moves, and negative means the positions tend to offset each other. Above about 0.7, treat the positions as one risk.

What is the difference between correlation and beta?

Correlation says how consistently two series move together; beta says how large one’s moves are relative to the other’s. A portfolio can have high correlation with Nifty and a beta below 1.

Are Nifty and Bank Nifty correlated?

Usually strongly and positively, since financial stocks are a large part of the Nifty 50. The exact figure varies with the period, so measure it on returns over your own holding window.

Related terms