Backtest Your Trading Strategy with Only 3 Lines of Python

By Lorenzo Ampil

Lorenzo Ampil
Image for post
Photo by Luke Chesser on Unsplash

Ever since I started investing back in college, I was exposed to the different ways of analyzing stocks — technical analysis and fundamental analysis. I’ve even read books and countless articles about these techniques.

In a nutshell, technical analysis argues that you can identify the right time to buy and sell a stock using technical indicators that are based on the stock’s historical price and volume movements. On the other hand, fundamental analysis argues that you can measure the actual intrinsic value of a stock based on the fundamental information found in a company’s financial statements.

Both types of analyses made sense to me and I was eager to use them to inform my trades; however, I was always frustrated about one main thing:

There are many possible strategies to take, but no systematic way to choose one. In practice, most trades still end up as “gut feel” decisions that are not driven by data.

So how can we possibly assess these strategies? We can do this by comparing the expected return on investment (ROI) that we can get from each approach. The best way to do this, is with a method called backtesting — where a strategy is assessed by simulating how it would have performed had you used it in the past.

Now, there are already quite a few backtesting frameworks out there, but most of them require advanced knowledge of coding. It’s typical for a simple hello world implementation to require as much as ~30 lines of code.

To fill this gap, I decided to create fastquant, with the goal of bringing backtesting to the mainstream by making it as simple as possible. With fastquant, we can backtest trading strategies with as few as 3 lines of code!

fastquant is essentially a wrapper for the popular backtrader framework that allows us to significantly simplify the process of backtesting from requiring at least 30 lines of code on backtrader, to as few as 3 lines of code on fastquant.

For the rest of this article, I will walk you through how to backtest a simple moving average crossover (SMAC) strategy through the historical data of Jollibee Food Corp. (JFC).

Let’s get started!

Backtest our first strategy

It’s as simple as using pip install!

# Run this on your terminalpip install fastquant# Alternatively, you can run this from jupyter this way

!pip install fastquant

Get stock data

Import the get_stock_data function from fastquant and use it to pull the stock data of Jollibee Food Corp. (JFC) from January 1, 2018 to January 1, 2019. Notice that we have columns corresponding to the date (dt), and closing price (close).

from fastquant import get_stock_datajfc = get_stock_data("JFC", "2018-01-01", "2019-01-01")print(df.head())# dt close# 2019-01-01 293.0# 2019-01-02 292.0# 2019-01-03 309.0# 2019-01-06 323.0

# 2019-01-07 321.0

Backtest your trading strategy

Backtest a simple moving average crossover (SMAC) strategy through the historical stock data of Jollibee Food Corp. (JFC) using the backtest function of fastquant.

In an SMAC strategy, fast period (fast_period) refers to the period used for the fast moving average, while slow period (slow_period) refers to the period used for the slow moving average. When the fast moving average crosses over the slow moving average from below to go above, this is considered a “buy” signal, while if it crosses over from above to go below, this is considered a “sell” signal. For more information on how this works, please check out the explanation in one of my previous articles.

To start out, let’s initialize the fast_period and slow_period as 15, and 40, respectively.

You should see the final portfolio value below at the bottom of the logs. This value can be interpreted as how much money your portfolio would have been worth at the end of the backtesting period (in this case January 1, 2019). If you get the difference between your “Final Portfolio Value” and your “Starting Portfolio Value”, this will be your expected earnings for that same period based on your backtest (in this case PHP 411.83).

from fastquant import backtest
backtest('smac', jfc, fast_period=15, slow_period=40)
# Starting Portfolio Value: 100000.00
# Final Portfolio Value: 100411.83

Bringing it all together — backtesting in 3 lines of Python

The code below shows how we can perform all the steps above in just 3 lines of python:

from fastquant import backtest, get_stock_datajfc = get_stock_data("JFC", "2018-01-01", "2019-01-01")

backtest('smac', jfc, fast_period=15, slow_period=40)

# Starting Portfolio Value: 100000.00
# Final Portfolio Value: 100411.83
Image for post
Simple Moving Average Crossover (15 day MA vs 40 day MA)

Improve our SMAC strategy

This shows how small changes can quickly turn a winning strategy into a losing one. Our final portfolio value went down from PHP 100,412 to PHP 83,947 (PHP 16,465 decrease), after increasing both fast_period, and slow_period to 30, and 50, respectively.

backtest('smac', jfc, fast_period=30, slow_period=50)# Starting Portfolio Value: 100000.00
# Final Portfolio Value: 83946.83

Decrease the slow period while keeping the fast period the same

In this case, the performance of our strategy actually improved! Our final portfolio value went up from PHP 100,412 to PHP 102,273 (PHP 1,861 increase), after decreasing the slow period to 35, and keeping the fast period the same at 15.

backtest('smac', jfc, fast_period=15, slow_period=35)# Starting Portfolio Value: 100000.00
# Final Portfolio Value: 102272.90
Image for post
Simple Moving Average Crossover (15 day MA vs 35 day MA)

The table below compares the performance of our 3 SMAC strategies:

Image for post
Strategy 3 (highlighted in yellow) was the most accurate SMAC strategy based on our backtest!

Overcome the limitations of backtesting

Now, does this mean we should go ahead and trade JFC using the best performing SMAC strategy? Maybe not just yet.

Backtesting has quite a few limitations and overcoming them will often require additional steps to increase our confidence in the reliability of our backtest’s results & recommendations.

Below are two of backtesting’s limitations followed by safeguards to overcome them:


This refers to the situation where the “optimal parameters” that you derived were fit too much to the patterns of a previous time period. This means that the expected profitability of your strategy will not translate to actual profitability in the future when you decide to use it.

One safeguard for this would be to test your strategies out-of-sample, which is similar to using a “test set” in machine learning. The idea is that you hold out some data, that you only use once later when you want to assess the profitability of your trading strategy. This way, it’s harder to overfit your parameters since you’re not optimizing your strategy based on that dataset.

Look ahead bias

This is the bias that results from utilizing information during your backtest that would not have been available during the time period being tested. For example, you could be testing the effectiveness of a strategy on JFC that assumes that you would have known about its financial performance (e.g. net income) a month before it was actually made available publicly. This would give you unreliable confidence in your strategy that could lose you a lot of money later.

In this case, one of the best things you can do to avoid this bias is to thoroughly validate the assumptions that you make when you’re backtesting your strategy. It pays to rigorously assess your strategy, and the information that has to be available for the strategy to be properly executed.