Spaces:
Runtime error
Runtime error
File size: 9,488 Bytes
e664a97 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 |
import pandas as pd
import numpy as np
import streamlit as st
import plotly.graph_objects as go
from plotly import tools
import plotly.offline as py
import plotly.express as px
import requests,json
from datetime import datetime, time,timedelta
import plotly.express as px
from json_loader import json_load
from returnsDf import agg_df
import warnings
warnings.filterwarnings('ignore')
pd.options.display.float_format = '${:,.2f}'.format
one_day=timedelta(days=1)
st.set_page_config(page_title="SquareoffbotsPerformance",layout='wide')
st.markdown("""<style> div[role="listbox"] ul {
background-color: rgb(229, 236, 122);
}
div[role="listbox"],option{
float:right;
}
#MainMenu {display:none;}
footer {display:none !important}
</style>
"""
,unsafe_allow_html=True)
pnl_url=r'http://performance.squareoffbots.com/assets/json/sqbots_allData_21052021.json'
cap_url=r'http://performance.squareoffbots.com/assets/json/newCAp21052021.json'
charges_url=r'http://performance.squareoffbots.com/assets/json/charges.json'
query_params = st.experimental_get_query_params()
@st.cache(ttl=20*60*60)
def getResources1():
charges_dic=requests.get(charges_url).json()
# charges_dic=json_load('charges.json')
pnl_data=requests.get(pnl_url).json()
cap_data=requests.get(cap_url).json()
pnl_df_t=pd.DataFrame.from_dict(pnl_data)
cap_df_t=pd.DataFrame.from_dict(cap_data)
pnl_df=pnl_df_t.T
cap_df=cap_df_t.T
return charges_dic,pnl_data,cap_data,pnl_df,cap_df,query_params
charges_dic,pnl_data,cap_data,pnl_df,cap_df,query_params=getResources1()
# charges_dic=requests.get(charges_url).json()
botNameDic={"orb":"ORB","rsi":"RSI","it":"Intraday Trend","sh":"StopHunt","grb":"GRB","orb2pm":"ORB2pm","pcr":"NiftyOptionSelling","lapp":"Learnapp","bss":"BNF Straddle","nss":"Nifty Straddle","bos":"BNFOptionSelling","grbo":"GRB Options","bssr":"BNF Strangle","mlb":"ML Bot","bnfmon":"BNF ORB","mss":"1% Short Straddle (BNF)","mssn":"1% Short Straddle(NF)","dts":"Double Top"}
botCapitalDic={"orb":50000,"rsi":50000,"it":50000,"sh":50000,"grb":300000,"orb2pm":300000,"pcr":300000,"lapp":300000,"bss":300000,"nss":300000,"bos":300000,"grbo":150000,"bssr":300000,"bnfmon":150000,"mlb":400000,"mss":300000,"mssn":300000,"dts":150000}
curBots=['orb','rsi','it','grb','bss','grbo','bssr','bnfmon','mlb','mss','mssn','dts']
botName = query_params["bot"][0] if "bot" in query_params else None
botsList=list(botNameDic.keys())
if botName not in botsList:
botName='bss'
botsList.remove('bss')
botsList=['bss']+botsList
if not botName:
botName = st.selectbox('Select a Strategy',tuple(botsList))
@st.cache(ttl=20*60*60,suppress_st_warning=True,allow_output_mutation=True)
def strategy_statistics():
ret_dic={}
for botName in curBots:
eq_bots=["orb","rsi","sh","it"]
botFullName=botNameDic[botName]
botCapital=botCapitalDic[botName]
strat_pnl_Df=pnl_df[[botFullName]]
strat_pnl_Df.dropna(inplace=True)
strat_cap_df=cap_df[[botFullName]]
#returns calculation
strat_df=agg_df(strat_pnl_Df,strat_cap_df)
##PNL plot
strat_df['pdTime']=pd.to_datetime(strat_df.index,format="%Y-%m-%d")
strat_df.sort_values('pdTime',inplace=True)
strat_df[botFullName+'_adj_PnL']=(botCapital/100)*strat_df[botFullName+' Returns'].astype(float)
strat_df[botFullName+'_gross_PnL']=strat_df[botFullName+'_adj_PnL']
charges_types=['Brokerage','TransactionCharges','ClearingCharges','STT','GST','SEBI','StampDuty','TotalCharges']
def getCharges(x):
notGotCharges=True
original_date=x
cDayObj=datetime.strptime(x,'%Y-%m-%d')
direction=-1
direction_counter=0
while notGotCharges:
try:
return charges_dic[cDayObj.strftime('%Y%m%d')][botName.upper()+'_'+ct]
except Exception as e:
direction_counter+=1
if direction_counter==10:
direction*=-1
cDayObj+=direction*one_day
continue
for ct in charges_types:
strat_df[ct]=strat_df.index.to_series().apply(lambda x:getCharges(x))
i_fields=['PNL','Brokerage','TransactionCharges','ClearingCharges','STT','GST','SEBI','StampDuty','TotalCharges','net_PNL']
##xxx
if botName in eq_bots:
i_fields=["PNL"]
strat_df['net_PNL']=strat_df[botFullName+'_gross_PnL']
else:
strat_df['net_PNL']=strat_df[botFullName+'_gross_PnL']-strat_df['TotalCharges']
strat_df['net_rets']=100*strat_df['net_PNL']/botCapital
strat_df["Time"]=strat_df.index
strat_df['PNL']=strat_df[botFullName+'_adj_PnL']
strat_df['cum_pnl']=strat_df[botFullName+'_adj_PnL'].cumsum()
##DRAWDOWN
drawdown_df=strat_df.copy()
drawdown_df.reset_index(drop=True,inplace=True)
drawdown_df['max_value_so_far']=drawdown_df['cum_pnl'].cummax()
drawdown_df['drawdown']=drawdown_df['cum_pnl']-drawdown_df['max_value_so_far']
max_drawdown=drawdown_df['drawdown'].min()
##Strategy statistics .
stats_Df=pd.DataFrame(columns=["Total Days","Winning Days","Losing Days","Winning Accuracy(%)","Max Profit","Max Loss","Max Drawdown","Average Profit on Win Days","Average Profit on loss days","Average Profit Per day","Gross Profit","Charges","Net profit","Returns (%)","net Returns (%)"])
total_days=len(strat_df)
win_df=strat_df[strat_df[botFullName+'_adj_PnL'].astype('float')>0]
lose_df=strat_df[strat_df[botFullName+'_adj_PnL'].astype('float')<0]
noTrade_df=strat_df[strat_df[botFullName+'_adj_PnL'].astype('float')==0]
win_days=len(win_df)
lose_days=len(lose_df)
win_ratio=win_days*100.0/(lose_days+win_days)
max_profit=strat_df[botFullName+'_adj_PnL'].max()
max_loss=strat_df[botFullName+'_adj_PnL'].min()
# max_drawdown=0
win_average_profit=win_df[botFullName+'_adj_PnL'].sum()/win_days
loss_average_profit=lose_df[botFullName+'_adj_PnL'].sum()/lose_days
total_profit=strat_df[botFullName+'_adj_PnL'].sum()
if botName not in eq_bots:
total_charges=strat_df['TotalCharges'].sum()
else:
total_charges=0
net_profit=total_profit-total_charges
average_profit=total_profit/total_days
gross_returns=strat_df[botFullName+' Returns'].sum()
net_returns=strat_df['net_rets'].sum()
results_row=[total_days,win_days,lose_days,win_ratio,max_profit,max_loss,max_drawdown,win_average_profit,loss_average_profit,average_profit,total_profit,total_charges,net_profit,gross_returns,net_returns]
results_row=[results_row[i] if i<3 else round(results_row[i],2) for i in range(len(results_row)) ]
stats_Df.loc[0,:]=results_row
t_stats_Df=stats_Df.T
t_stats_Df.rename(columns={0:''},inplace=True)
fig=px.line(strat_df, x="Time", y='cum_pnl', title=botFullName+' PNL',width=800, height=400)
dd_fig=px.line(drawdown_df,x="Time",y="drawdown", title=botFullName+' PNL',width=800, height=400)
strat_df['month']=strat_df['pdTime'].apply(lambda x:x.strftime('%b,%Y'))
month_groups=strat_df.groupby('month',sort=False)[i_fields].sum()
##last 30 days pnl
strat_df=strat_df.reindex(strat_df.index[::-1])
if botName in eq_bots:
capital_used_appendum=''
else:
capital_used_appendum=' per Lot'
ret_dic[botName]=[botCapital,capital_used_appendum,results_row,t_stats_Df,fig,dd_fig,month_groups,strat_df,i_fields,botFullName]
return ret_dic
ret_dic=strategy_statistics()
botCapital,capital_used_appendum,results_row,t_stats_Df,fig,dd_fig,month_groups,strat_df,i_fields,botFullName=ret_dic[botName]
title_text="<h1 style='text-align: center; color: rgb(21, 86, 112);'>**♟**SQUAREOFF BOTS PERFORMANCE**♟**</h1><br><div style='text-align: center; color: rgb(21, 86, 112);'>**LIVE PERFORMANCE OF "+botFullName+"****[Capital used is "+str(botCapital)+capital_used_appendum+"]** </div>"
st.markdown(title_text, unsafe_allow_html=True)
if botCapital>50000 and botName!='mlb':
# col1.write("**(Capital used before July 2021 is "+str(int(botCapital/1.5))+capital_used_appendum+")**")
st.markdown("<div style='text-align: center; color: rgb(21, 86, 112);'>**(Capital used before July 2021 is "+str(int(botCapital/1.5))+capital_used_appendum+")** </div>", unsafe_allow_html=True)
col1, col2 = st.columns(2)
col2.markdown('##')
col1.write("Net ROI : "+str(results_row[-1])+"%")
col1.write("**Statistics**")
col1.table(t_stats_Df)
col2.write("**PNL Curve**")
col2.plotly_chart(fig)
col2.write("**Drawdown Curve**")
col2.plotly_chart(dd_fig)
st.write("**Month-wise PNL**")
st.table(month_groups)
st.write("**Date-wise PNL (Last 30 Days)**")
st.table(strat_df[i_fields][:30])
##TABLE
# fig2=go.Figure(data=[go.Table(
# header=dict(values=['Date','BSS'],
# fill_color='white',
# line_color='black',
# align='left'),
# cells=dict(values=[strat_df.index,strat_df[botFullName+'_adj_PnL']],
# fill_color='white',
# align='left'))])
# st.plotly_chart(fig2) |