We already looked at a price filter for the S&P500 index here.

Let us try it for some other common indices and see how it works on them.

List of indices to test

Dow Jones - United States
Russell 2000 - England
Nifty 50 - India
S&P500 - United States
Dax - Germany
Nikkei 225 - Japan
Hangseng - HongKong
Moex - Russia
SSE - China
Asx 200 - Australia
Euronext100 - Europe
Tsx Composite - Canada

I have downloaded the data from Yahoo from the start of 2000 and applied the moving average price filter for each of the indices. The get data function does all the data stuff. Finally, I group by each stock exchange and the price filter and take the daily mean of the returns.

In [15]:

import pandas as pd
import pandas_datareader as web
import seaborn as sns
sns.set()
sns.set(rc={'figure.figsize':(12,6)})

In [16]:

# parameters
ma = 60

# tickers are specified as name versus symbol in yahoo finance
tickers = {
    'dow': '^DJI',
    'russell': '^RUT',
    'nifty': '^NSEI',
    'snp500': '^GSPC',
    'dax': '^GDAXI',
    'nikkei': '^N225',
    'hangseng': '^HSI',
    'moex': 'IMOEX.me',
    'sse': '000001.SS',
    'asx': '^AXAT',
    'euronext': '^N100',
    'tsx comp': '^GSPTSE'
}

In [17]:

def get_data(index, name):
    """
    Get the data for the 
    """
    df = web.DataReader(index, 'yahoo', start='2000-01-01').rename(columns = lambda x: x.lower())
    df = df.sort_index()
    df['year'] = df.index.year
    df['ret'] = df.close.pct_change()
    df['ma_price'] = df.close.rolling(ma).median().shift(1)
    df['is_price'] = df.eval('close > ma_price')+0
    df['is_price'] = df.is_price.shift(1) # Shifting price since we use the signal only next day
    df['name'] = name
    return df

In [18]:

collect = []
for k,v in tickers.items():
    try:
        tmp = get_data(v, name=k)
        collect.append(tmp)
    except Exception as e:
        print(e)
df = pd.concat(collect)

In [19]:

mean_returns = df.groupby(['name', 'is_price']).ret.mean().reset_index()
sns.barplot(data=mean_returns, x='name', y='ret', hue='is_price')

Out[19]:

<AxesSubplot:xlabel='name', ylabel='ret'>

Looks like there is no common pattern among the indices.

And all except hangseng and see provide positive returns for both the filters.

On a second look, snp500, dow, russell,euronext does well with a filter of 0 while nifty,nikkei,sse,tsx does well with a filter of 1.

May be

the american and european markets exhibit different behaviour than the asian counterparts (but TSX is from Canada)
the older exchanges may exhibit a different behaviour than the new one
the timezone effect
or just plain random

I cannot come up with any conclusive explanation. This is a good topic to explore further with a whole lot of other indices

There is no substantial edge to practically trade this filter. Also, we have not run any statistical test to validate significance

In [20]:

# Starting date from which data is computed
df.reset_index().groupby('name').Date.min().sort_values()

Out[20]:

name
dow        1999-12-31
euronext   1999-12-31
russell    1999-12-31
snp500     1999-12-31
tsx comp   1999-12-31
dax        2000-01-03
hangseng   2000-01-03
nikkei     2000-01-04
sse        2000-01-04
asx        2007-06-15
nifty      2007-09-17
moex       2013-03-05
Name: Date, dtype: datetime64[ns]

Price filter major indices

Price filter for major indices¶