Portfolio Backtesting & Deployment Framework¶
In this section, we investigate the framework that goes behind researching and developing investment strategies. It ties the previous sections together,
There are several elements that are considered in academia and industry practice, which we discuss below.
Securities Universe¶
The securities universe is a set of securities based on defined parameters.To apply this procedure, we make use of a StockScreener object.
The process typically starts with first picking stocks that are trading in a specific region (i.e. USA, Canada, Asia…), exchange (i.e. NASDAQ, NYSE…), index (i.e. Dow Jones, S&P 500…), sector, or industry.
The process continues with conducting an initial filtering based on concerns like liquidity (i.e. market capitalization, volume traded). Finally, narrowing down based on types of analysis such as technical, fundamental, and quantitative, which we detail in their respective modules.
We use the metrics above (either as raw, mean, or growth values), then compare against either a number, another metric, or percentile against competitors.
Market Timing¶
It is important that these conditions are all re-evaluated when we rebalance the portfolio, to avoid succumbing to mistakes like survivorship bias.
Rebalancing Frequency Trigger Based: Define another StockScreener object
Portfolio Allocation¶
Other Considerations¶
Reinvesting dividends
Including capital gain taxes
Historical Simulation¶
Broker Deployment¶
from datetime import datetime from functools import partial from pprint import pprint
- from fundamental_analysis.equity_valuation_models.equity_valuation_models import valuation_wrapper,
absolute_valuation_two_stage_model
from fundamental_analysis.financial_statements_entries import net_income from portfolio_management.Portfolio import Portfolio from portfolio_management.portfolio_optimization import EquallyWeightedPortfolio from portfolio_management.portfolio_simulator import Strategy from portfolio_management.stock_screener import StockScreener import config import fundamental_analysis.accounting_ratios as ratios import fundamental_analysis.supporting_metrics as me import pandas as pd from quantitative_analysis.risk_factor_modeling import FamaFrench_ThreeFactorModel from quantitative_analysis.risk_factor_modeling.asset_pricing_model import FactorModels
- pprint(me.market_price(stock=’BA’,
date=[datetime.now(), datetime(2019, 1, 1)]))
- pprint(net_income(stock=[‘AMGN’, ‘AXP’],
date=datetime.now(), period=’FY’))
- pprint(ratios.price_to_book_ratio(stock=[‘BA’, ‘AXP’, ‘AAPL’],
date=[datetime.now(), datetime(2019, 9, 1)], tangible_book_value=True, period=’FY’))
- pprint(valuation_wrapper(model_type=partial(absolute_valuation_two_stage_model),
model_metric=me.dividend_per_share, stock=’AAPL’, period=’FY’))
ff3 = FamaFrench_ThreeFactorModel(frequency=’Monthly’, to_date=datetime.today()) reg = ff3.regress_factor_loadings(portfolio=Portfolio(assets=[‘MSFT’])) pprint(reg.params)
stock_screener = StockScreener(securities_universe=config.MarketIndices.DOW_JONES) stock_screener.filter_by_comparison_to_number(partial(ratios.price_to_earnings, period=’FY’), ‘>’, 5) stock_screener.filter_by_sector(sector=config.GICS_Sectors.INFORMATION_TECHNOLOGY) stock_screener.run()
# TODO percentile_against_macro
lower_bounds = pd.Series(data=[40], index=[‘Alpha’]) upper_bounds = pd.Series(data=[80], index=[‘MKT’]) stock_screener.filter_by_exposure_from_factor_model(factor_model=FactorModels.CAPM,
lower_bounds=lower_bounds, upper_bounds=upper_bounds)
stock_screener.run(date=datetime(2018, 1, 1)) print(stock_screener.stocks)
- class Alainps(Strategy):
- def is_time_to_reschedule(self, current_date, last_rebalancing_day):
return (current_date - last_rebalancing_day).days > config.RebalancingFrequency.Quarterly.value
- strategy = Alainps(starting_date=datetime(2019, 1, 1), ending_date=datetime(2020, 12, 1),
starting_capital=50000, stock_screener=stock_screener, max_stocks_count_in_portfolio=12, net_exposure=(100, 0), portfolio_allocation=EquallyWeightedPortfolio)
strategy.historical_simulation()