Implementation of Trading Algorithm Betting Against Beta based on Andrea Frazzini and Lasse H.Pedersen's research paper. Regression on the CAPM equation is used to get Beta.
The project's aim is to generate an equally weighted portfolio.
- Apply learnings from the EPAT course, and
- Create Proof points of applicability and relevance of this theory in the Indian stock markets
There are two types of risks
| Idiosyncratic risk | Systematic risk |
|---|---|
Specific to firm |
Market wide |
Doesn't affect the whole market |
Affect all the firm |
Diversified away |
Cannot be diversified away |
The Systematic risk can be captured through beta.
| Beta | Interpretation |
|---|---|
| Greater than 1 | More volatile than the market |
| Close to 1 | As volatile as the market |
| Between 0 and 1 | Less volatile than the market |
| 0 | Not correlated to the market |
| Less than 0 | Negatively correlated to the market |
For example, if a stock's beta is 1.2, it's theoretically 20% more volatile than the market.
import statsmodels.api as sm
def calc_beta(stock_series, bechmark_series):
model = sm.OLS(stock_series, bechmark_series)
results = model.fit()
return results.params[0]- People prefer assets with higher expected returns per unit of risk
- Mutual funds and retail investor are constrained in the leverage
- They tilt allocation to high beta assets to improve returns resulting in over pricing
- Read the Nifty component stock list
nifty_list = pd.read_csv('ind_nifty50list.csv')- Read the price data of the stocks
data = pd.read_csv('nifty_stocks_prices.csv',index_col=0)
data.index = pd.to_datetime(data.index)- Calculate Daily percentage change
data_pct_change = data.pct_change().dropna
- Calculate the beta on train data (2013 - 2017, to avoid look ahead bias) and plotting the data
beta = pd.DataFrame(index=[0])
for ticker in data_pct_change.columns:
beta[ticker] = calc_beta(data_pct_change.loc[:'2017',ticker],data_pct_change.loc[:'2017','^NSEI'])
beta = beta.T
beta.columns = ['beta']
bar_p = beta.plot.bar(figsize=(18,7))
plt.show()
- Generate signals and calculate strategy returns Creating portfolio with low beta stocks
low_beta = beta[beta.values < 0.7].index
low_beta
def plot_performance(stock_list, strategy_name):
stk_returns = data_pct_change.loc['2018':, stock_list]
(stk_returns+1).cumprod().plot(figsize=(15,7),legend="left")
plt.title(strategy_name)
plt.show()
nifty = data_pct_change.loc['2018':, '^NSEI'] #Benchmark
portfolio = stk_returns.mean(axis=1)
plt.title(strategy_name + ' Portfolio Performance')
(portfolio+1).cumprod().plot(figsize=(15,7),label=strategy_name, color='purple')
(nifty+1).cumprod().plot(figsize=(15,7),label='Nifty', color='blue')
plt.legend()
plt.show()
return portfolio
Portfolio performanced well compared to Nifty index,even during the downfall and the recovery was even better.
Our strategy has an advantage since it falls less when the market falls and rises more when the market rises, gives edge to low beta portfolio.
We're going to add one more filter to our strategy, which is Return on Equity. For all stocks, we're using ROE figures till 2017(to avoid look ahead bias).
high_roe = nifty_list[nifty_list.ROE>18].Symbol.values
Combining Alpha
filtered_stocks = low_beta & high_roe
- Generate the signal based on beta and ROE
portfolio = plot_performance(filtered_stocks, 'Low Beta & High ROE')
- Returns analysis using pyfolio
import pyfolio as pf
pf.create_simple_tear_sheet(portfolio,benchmark_rets=data_pct_change.loc['2018':, '^NSEI'])
