简体   繁体   中英

Is there a way to vectorize this procedure in pandas?

I have a simple stock portfolio simulation I am trying to model, but despite some attempts, I cannot figure out a way to vectorize this. Maybe it's not possible, but I wanted to see if anyone out there had any thoughts.

My sticking point is that the shares on a given day are a function of the account value and stock price of two days previous. But the account value on a day is a function of the previous day's account value and today's number of shares and stock price change. So there is a back and forth relationship between shares and account value that I can't think of a way to vectorize, and thus my only solution below is the for loop below.

Thanks in advance!

import pandas as pd
import numpy as np
stats = pd.DataFrame(index = range(0,10))

stats['Acct Val'] = 0.0
stats['Shares'] = 0.0
stats['Stock Px'] = pd.Series([23,25,24,26,22,23,25,25,26,24],index=stats.index)
# Wgt is the percentage of the account value that should be invested in the stock on a given day
stats['Wgt'] = pd.Series([0.5,0.5,0.5,0.5,0.3,0.4,0.4,0.2,0.2,0.0,],index=stats.index)
stats['Daily PNL'] = 0.0
# Start the account value at $10,000.00
stats.ix[0:1, 'Acct Val'] = 10000.0
stats.ix[0:1, 'Wgt'] = 0
for date_loc in range(2, len(stats.index)):
    # Keep shares the same unless 'wgt' column changes
    if stats.at[date_loc,'Wgt'] != stats.at[date_loc-1,'Wgt']:
        # Rebalanced shares are based on the acct value and stock price two days before
        stats.at[date_loc,'Shares'] = stats.at[date_loc-2,'Acct Val'] * stats.at[date_loc,'Wgt'] / stats.at[date_loc-2,'Stock Px']
    else:
        stats.at[date_loc,'Shares'] = stats.at[date_loc-1,'Shares']
    # Daily PNL is simply the shares owned on a day times the change in stock price from the previous day to the next
    stats.at[date_loc,'Daily PNL'] = stats.at[date_loc,'Shares'] * (stats.at[date_loc,'Stock Px'] - stats.at[date_loc-1,'Stock Px'])
    # Acct value is yesterday's acct value plus today's PNL
    stats.at[date_loc,'Acct Val'] = stats.at[date_loc-1,'Acct Val'] + stats.at[date_loc,'Daily PNL']


In [44]: stats
Out[44]:
       Acct Val      Shares  Stock Px  Wgt   Daily PNL
0  10000.000000    0.000000        23  0.0    0.000000
1  10000.000000    0.000000        25  0.0    0.000000
2   9782.608696  217.391304        24  0.5 -217.391304
3  10217.391304  217.391304        26  0.5  434.782609
4   9728.260870  122.282609        22  0.3 -489.130435
5   9885.451505  157.190635        23  0.4  157.190635
6  10199.832776  157.190635        25  0.4  314.381271
7  10199.832776   85.960448        25  0.2    0.000000
8  10285.793224   85.960448        26  0.2   85.960448
9  10285.793224    0.000000        24  0.0   -0.000000

In [45]:

EDIT: 11:01 PM October 19, 2013:

I tried using foobarbecue's code but I couldn't get there:

import pandas as pd
import numpy as np
stats = pd.DataFrame(index = range(0,10))
stats['Acct Val'] = 10000.0
stats['Shares'] = 0.0
stats['Stock Px'] = pd.Series([23,25,24,26,22,23,25,25,26,24],index=stats.index)
# Wgt is the percentage of the account value that should be invested in the stock on a given day
stats['Wgt'] = pd.Series([0.5,0.5,0.5,0.5,0.3,0.4,0.4,0.2,0.2,0.0,],index=stats.index)
stats['Daily PNL'] = 0.0
# Start the account value at $10,000.00
#stats.ix[0:1, 'Acct Val'] = 10000.0
stats.ix[0:1, 'Wgt'] = 0

def function1(df_row):
    #[stuff you want to do when Wgt changed]
    df_row['Shares'] = df_row['Acct Val'] * df_row['Wgt2ahead'] / df_row['Stock Px']
    return df_row

def function2(df_row):
    #[stuff you want to do when Wgt did not change]
    df_row['Shares'] = df_row['SharesPrevious']
    return df_row

#Find where the Wgt column changes
stats['WgtChanged']=stats.Wgt.diff() <> 0 # changed ">" to "<>"
#Using boolean indexing, choose all rows where Wgt changed and apply a function
stats['Wgt2ahead'] = stats['Wgt'].shift(-2)
stats = stats.apply(lambda df_row: function1(df_row) if df_row['WgtChanged'] == True else df_row, axis=1)
stats['Shares'] = stats['Shares'].shift(2)
#Likewise, for rows where Wgt did not change
stats['SharesPrevious'] = stats['Shares'].shift(1)
stats = stats.apply(lambda df_row: function2(df_row) if df_row['WgtChanged'] == False else df_row, axis=1)
def function1(df_row):
    [stuff you want to do when Wgt changed]

def function2(df_row):
    [stuff you want to do when Wgt did not change]

#Find where the Wgt column changes
stats['WgtChanged']=stats.Wgt.diff() > 0
#Using boolean indexing, choose all rows where Wgt changed and apply a function
stats[stats['WgtChanged']].apply(function1, axis=1)
#Likewise, for rows where Wgt did not change
stats[~stats['WgtChanged']].apply(function2, axis=1)

The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM