Is there any time in Forex that repeats?

From the archive, written in 2018. Libraries and versions have moved on since.

CHAPTER 5

via GIPHY

Let’s leave the deep learning models for a while and try some simply statistics to create our strategy.

Is there any time during the week that the next candle will be most likely bullish or bearish? Let’s find out

Import Libraries

import matplotlib
import numpy as np
import pandas as pd
import time
import datetime

from matplotlib import pyplot as plt
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker
from mpl_finance import candlestick_ohlc
import seaborn as sns

print('Numpy version: ' + np.__version__)
print('Pandas version: ' + pd.__version__)
print('Matplotlib version: ' + matplotlib.__version__)

Numpy version: 1.16.4
Pandas version: 0.24.2
Matplotlib version: 3.1.0

float_data = pd.read_csv('Hour/EURUSD.csv')
print(len(float_data))

101160

float_data.head(10)

def transform_to_heatmap(dataframe):
    dataframe.loc[dataframe['Close']>dataframe['Open'], 'color'] = 1
    dataframe.loc[dataframe['Close']<dataframe['Open'], 'color'] = -1
    del dataframe['Volume']
    del dataframe['Open']
    del dataframe['High']
    del dataframe['Low']
    del dataframe['Close']
    return dataframe

float_data=transform_to_heatmap(float_data)

float_data.head(10)

def get_week_activity(raw_predictions,date_values,Filter=False):
    Monday_ls = [0] * 24
    Tuesday_ls = [0] * 24
    Wendsday_ls = [0] * 24
    Thursday_ls = [0] * 24
    Friday_ls = [0] * 24

    counter = 0

    for a in raw_predictions:
        raw_date_converted = datetime.datetime.strptime(date_values[counter], '%d.%m.%Y %H:%M:%S.000')
        week_day = raw_date_converted.weekday()
        hour = raw_date_converted.hour

        candle_check =[0,0]
        if a == 1: candle_check =[1,0]
        if a == -1: candle_check =[0,1]


        if week_day==0: Monday_ls[hour-1]=Monday_ls[hour-1]+np.array(candle_check)
        if week_day==1: Tuesday_ls[hour-1]=Tuesday_ls[hour-1]+np.array(candle_check)
        if week_day==2: Wendsday_ls[hour-1]=Wendsday_ls[hour-1]+np.array(candle_check)
        if week_day==3: Thursday_ls[hour-1]=Thursday_ls[hour-1]+np.array(candle_check)
        if week_day==4: Friday_ls[hour-1]=Friday_ls[hour-1]+np.array(candle_check)
        counter=counter+1


    final_list = [Monday_ls,Tuesday_ls,Wendsday_ls,Thursday_ls,Friday_ls]
    weekdays_count = 0
    for weekdays in final_list:
        hours_count = 0
        for hours in weekdays:
            if type(hours) != int:
                stats = round((hours[0]*100)/(hours[1]+hours[0]),0)
                if Filter==True:
                    if stats <= 55 and stats >= 45: stats=0
                    if stats < 45: stats= -stats

                final_list[weekdays_count][hours_count]=stats
            hours_count=hours_count+1
        weekdays_count=weekdays_count+1

    return [Monday_ls,Tuesday_ls,Wendsday_ls,Thursday_ls,Friday_ls]

test=get_week_activity(float_data['color'].values,float_data['Gmt time'].values)

def make_heat_map(dataset):


    weekdays=['Monday','Tuesday','Wendsday','Thursday','Friday']
    plt.figure(figsize=(18,5))
    ax = sns.heatmap(dataset,annot=True,annot_kws={"size": 10},
                        linewidth=0.5, yticklabels=weekdays,cmap="Blues")
    plt.title('Trade activity during the week')
    ax.set_ylabel('Week Days')
    ax.set_xlabel('Hours')

    plt.show

EUR/USD 2003-2019 1h candles data

make_heat_map(test)

GBP/USD 2003-2019 1h candles data

GBPUSD_dataset = pd.read_csv('Hour/GBPUSD.csv')
float_data=transform_to_heatmap(GBPUSD_dataset)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values)
make_heat_map(stats_list)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values,Filter=True)
make_heat_map(stats_list)

USD/CAD 2003-2019 1h candles data

USDCAD_dataset = pd.read_csv('Hour/USDCAD.csv')
float_data=transform_to_heatmap(USDCAD_dataset)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values)
make_heat_map(stats_list)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values,Filter=True)
make_heat_map(stats_list)

NZD/USD 2003-2019 1h candles data

NZDUSD_dataset = pd.read_csv('Hour/NZDUSD.csv')
float_data=transform_to_heatmap(NZDUSD_dataset)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values)
make_heat_map(stats_list)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values,Filter=True)
make_heat_map(stats_list)

USD/JPY 2003-2019 1h candles data

USDJPY_dataset = pd.read_csv('Hour/USDJPY.csv')
float_data=transform_to_heatmap(USDJPY_dataset)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values)
make_heat_map(stats_list)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values,Filter=True)
make_heat_map(stats_list)

AUD/USD 2003-2019 1h candles data

AUDUSD_dataset = pd.read_csv('Hour/AUDUSD.csv')
float_data=transform_to_heatmap(AUDUSD_dataset)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values)
make_heat_map(stats_list)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values,Filter=True)
make_heat_map(stats_list)

USD/CHF 2003-2019 1h candles data

USDCHF_dataset = pd.read_csv('Hour/USDCHF.csv')
float_data=transform_to_heatmap(USDCHF_dataset)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values)
make_heat_map(stats_list)
stats_list=get_week_activity(float_data['color'].values,float_data['Gmt time'].values,Filter=True)
make_heat_map(stats_list)