Currency Pair Correlation Analysis for Strategic Trading

What Idea about Correlation Currency ?

The idea of correlation-based currency pair trading is to analyze the correlation between different currency pairs and utilize this information to make informed trading decisions. By identifying currency pairs with positive correlation, we can implement pair trading strategies or predict the potential movement of one currency based on the performance of another. In this example, we will analyze the correlation between the following currency pairs:

Import the necessary libraries such as yfinance, pandas, numpy, matplotlib, and seaborn.

import yfinance as yf
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

Download historical currency data from Yahoo Finance.

data = yf.download("EURUSD=X GBPUSD=X AUDUSD=X NZDUSD=X CADUSD=X CHFUSD=X FJDUSD=X GHSUSD=X JPYUSD=X KYDUSD=X SGDUSD=X THBUSD=X BYNUSD=X MYRUSD=X HUFUSD=X ARSUSD=X",start = '2021-01-01',end='2021-10-14',period='1d')

df= data['Adj Close']

df.head(2)

|  Date |  ARSUSD=X |  AUDUSD=X |  BYNUSD=X |  CADUSD=X |  CHFUSD=X |  EURUSD=X |  FJDUSD=X |  GBPUSD=X |  GHSUSD=X |  HUFUSD=X |  JPYUSD=X |  KYDUSD=X |  MYRUSD=X |  NZDUSD=X |  SGDUSD=X |  THBUSD=X |

| 2021-01-01  | 0.011895  | 0.770297  | 0.383408  | 0.809723  | 1.113462  | 1.218027  | 0.490196  | 1.367301  | 0.171233  | 0.003370  | 0.009687  | 1.224190  | 0.248756  | 0.718200  | 0.756659  | 0.033444  |

| 2021-01-04  | 0.011869  | 0.771230  | 0.381441  | 0.786034  | 1.132375  | 1.225070  | 0.492902  | 1.368420  | 0.170109  | 0.003383  | 0.009686  | 1.217919  | 0.248756  | 0.719839  | 0.757191  | 0.033385  |

we will see dataframe like this

df.stack().groupby(level=1).head(1)

 Date                
2021-01-01      ARSUSD=X    0.011895
                AUDUSD=X    0.770297
                BYNUSD=X    0.383408
                CADUSD=X    0.809723
                CHFUSD=X    1.113462
                EURUSD=X    1.218027
                FJDUSD=X    0.490196
                GBPUSD=X    1.367301
                GHSUSD=X    0.171233
                HUFUSD=X    0.003370
                JPYUSD=X    0.009687
                KYDUSD=X    1.224190
                MYRUSD=X    0.248756
                NZDUSD=X    0.718200
                SGDUSD=X    0.756659
                THBUSD=X    0.033444
dtype: float64

Clean the data by filling missing values using the forward-fill method. Calculate the daily percentage returns for each currency pair.

df = df.fillna(method='ffill')
df_return = df.pct_change()
df_ret = df_return.loc[~df_return.isnull().sum(1).astype(bool)]

Create a correlation matrix using the percentage returns.

Visualize the correlation matrix using Seaborn’s clustermap function to identify groups of correlated currency pairs.

g = sns.clustermap(df_ret.corr(), method='ward', center = 0.0, cmap="RdYlGn",
                   dendrogram_ratio=(.2, .2),
                   linewidths=.75, figsize=(12, 13),
                   annot = True)

Based on the visualization, create two groups of correlated currency pairs:

Group 1: GBPUSD, SGDUSD, AUDUSD, NZDUSD

Group 2: CHFUSD, EURUSD, HUFUSD

Calculate the cumulative geometric returns for each group.

group1 = df[['GBPUSD=X', 'SGDUSD=X', 'AUDUSD=X', 'NZDUSD=X']]
group2 = df[['CHFUSD=X', 'EURUSD=X', 'HUFUSD=X']]

create return value of group1

group1_geo_ret = (group1.pct_change()+1).cumprod()
group1_geo_ret
| Date       | GBPUSD=X | SGDUSD=X | AUDUSD=X | NZDUSD=X |
|------------|----------|----------|----------|----------|
| 2021-01-01 | NaN      | NaN      | NaN      | NaN      |
| 2021-01-04 | 1.000818 | 1.000704 | 1.001211 | 1.002282 |
| 2021-01-05 | 0.991772 | 1.000288 | 0.994645 | 0.997208 |

plot group 1 geo return

group1_geo_ret.plot(figsize=(10,6))

Now we can pair trade in this currency group or make decision GBP will going down like SGD AUD NZD

create return value of group2

group2_geo_ret = (group2.pct_change()+1).cumprod()

plot group 2 geo return

group2_geo_ret.plot(figsize=(10,6))

Now we can pair trade in this currency group or make decision HUF will going down like CHF EUR

the cumulative geometric returns for both groups to visualize their performance over time.

By analyzing the correlation between various currency pairs, we identified two groups of correlated pairs.

For Group 1 (GBPUSD, SGDUSD, AUDUSD, NZDUSD), we can now make informed decisions, such as predicting a decline in the value of GBP in line with SGD, AUD, and NZD. Similarly,

for Group 2 (CHFUSD, EURUSD, HUFUSD), we can anticipate a decrease in the value of HUF based on the performance of CHF and EUR.

Conclusion

In conclusion, analyzing the correlation between currency pairs can provide valuable insights for pair trading strategies or predicting the potential movement of one currency based on the performance of another. This approach can be extended to other asset classes, such as equities or cryptocurrencies, to identify potential trading opportunities.