Interests in Quant investing (or trading) are growing fast every day. Many people try to create profitable strategies and test it using past data (It is called 'backetest' by quants). It is good phenomenon in that it can prevent people from wasting their money with unverified strategies, or just news) and enables investing with more statistical approaches. However, some people started to cling to just higher CAGR (compound annual growth rate) and total returns and ignore other statistical performance to ensure robustness of strategy. CAGR is just one factor to evaluate strategy, and if you don't care other evaluations, your investing can be in danger in the future. See below pictures.

'Strategy B' has underperformed 'Strategy A' before 2014, but it outperformed 'Strategy A' at the end in the perspective of total returns and CAGR. But as you know, we cannot say 'Strategy B' is more robust than 'Strategy A'

Rofa is abbreviation for 'Robust Factor'. This module helps you to check factor's robustness. All you do is just prepare data with certain formats, and ask rofa to simulate, and evaluate, and make summarized graph of it. Now you had powerful quant tool 'Rofa'.

pip install rofa

pip install jupyter

To start it

jupyter notebook

• Import Rofa

First of all, import rofa and QuantileSimulator from rofa

import rofa
from rofa import QuantileSimulator

• Registering daily returns data

Unfortunately, rofa does not have any data with it. So, in order to run simulation, you need to **register daily returns data (pandas DataFrame). If you register returns data once, rofa will find returns data later without re-registering (You can change this option using save=False).

import pandas as pd

# Read Close Data
returns= pd.read_pickle('../data/returns.pkl')  # Your returns data
rofa.register_returns(returns)

• Prepare data

Data (pandas Dataframe) must have formats where columns are asset symbols (or code) and index is date such as returns. You can download data used in this example in here.

# Read Close Data
close = pd.read_pickle('../data/close.pkl')  # Your data

• QuantileSimulator

quan_sim = QuantileSimulator(close, rebalance_freq=10, bins=5)

In QuantileSimulator, first argument accepts factor data (close here). Additionaly, you can set rebalance_freq, bins, tax_rate, weight_model and etc.

• Run simulation

Just run the simulation. Simulation logics are all done by rofa

quan_sim.run()

• Plot portfolio returns

Simulation classes has plotter plugin inside it, which makes it possible to visuallize the simulation result.

quan_sim.plotter.plot_portfolio_returns()

From portfolio returns graph, we can compare overall performances and drawdowns of each portfolio.

• Plot performance metrics (CAGR, MDD, Sharpe, Calmar)

From this graph, we can check performance metrics and check if there is strong relationship between factor and performance.

• Plot rolling performance.

• Wait, we can plot all at once

You might have though about how come I can memorize all plot methods. Here's a method for you. plot_all plots all above. Super simple!

quan_sim.plotter.plot_all()

• Add more performance indicators
• Optimize code for efficiency. There are some points to make code inefficient
• Create LongShortPlotter and make all methods used in QuantilePlotter
• Add statistical analysis plugin such as Linear Regression, t-test, and ANOVA
• Create NakedSimulator and add plotter plugins
• Create Evaluator Plugin Later
• Use numba or cython to improve performance
• Better documentation!