How to read the correlation table
A correlation coefficient runs from −1 to +1 and tells you whether two currency pairs usually moved in the same direction or in opposite directions over the candles you chose. It measures how far the risk in two trades overlaps; it is not a buy or sell signal.
- Near +1: the two pairs usually move together. Buying both EURUSD and GBPUSD while the coefficient is high is close to one trade at double the size: if the US dollar strengthens, both trades lose.
- Near −1: the two pairs usually move in opposite directions. Buying EURUSD and buying USDCHF at the same time means the two trades mostly cancel each other out, so in practice you hold almost no position.
- Near 0: the two pairs had little to do with each other over that stretch.
Correlation changes over time. 20 candles reflect what happened recently but turn over quickly; 100 candles are steadier but slower to catch up. The 1-hour timeframe shows how the pairs have been moving together over the last few days, which suits trades you hold intraday; the daily timeframe gives you the picture over several months. Before you open another trade in a pair whose coefficient with an open trade is high, work out your total risk again with the position size tool.
How it is calculated
- The price of each candle: on the 1-hour and 4-hour timeframes it is the close of the MT5 candle on the CXMDirect-Live server, closed candles only; on the daily timeframe it is the European Central Bank reference rate published each business day, and a pair without EUR is crossed through EUR.
- The log return of one candle: r = ln(this candle price ÷ the previous candle price).
- The Pearson coefficient over the N most recent returns both pairs have figures for: the covariance of the two series divided by the product of their standard deviations.
The two sources take their prices at different moments. The ECB takes its rates at one point each day, which is not a market close; MT5 candles are built from the CXM Bid price, and another broker may sit slightly off it. So the coefficients here can differ a little from the ones you calculate on another platform.