Hands-on Elixir & OTP: Cryptocurrency trading bot

$10+
$10+
Kamil Skowron
6 ratings

Want to learn Elixir & OTP by creating a real-world project?

With "Hands-on Elixir & OTP: Cryptocurrency trading bot", you will gain hands-on experience by writing an exciting software project from scratch. We will explore all the key abstractions and essential principles through iterative implementation improvements.

We will start by creating a new umbrella application, subscribing to WebSocket streams, implementing a basic trading flow, and focusing on improving it by expanding on the topics like supervision trees, resiliency, refactoring using macros, utilising the Registry, testing and others.

This book is 90% complete - chapters 1-20 are finished, and I'll add more content soon. It's also a loosely written representation of the "Hands-on Elixir & OTP: Cryptocurrency trading bot" video course released on YouTube.

Praise from the readers

"The most practical source of knowledge about the OTP-related topics I've ever seen! I've enjoyed watching the course and then jumped into reading the book, and whoah! I fully recommend reading, watching and following Kamil's content! It's just amazing if you still have problems with GenServers, Supervisors, etc. - thanks to this book, you can fully understand the pragmatic, practical usage of these mechanisms, which you can face in the Elixir language. Another thing that is not obvious is that you can easily get HOW to model and develop a loosely coupled application from this book. That's a rare and so important thing."
- Patryk Woziński @patrykwozinski

What this book covers

This book is an ongoing project, and at present, it contains the following chapters:

Chapter 1 - Stream live cryptocurrency prices from the Binance WSS

Stream live cryptocurrency prices (trade events) from the Binance exchange. We will create a new umbrella project and a streamer application inside it starting grounds up. The streamer application will use a Websocket client called WebSockex to connect with the Binance API and receive a live feed. After receiving the event as a JSON string, we will decode it using the jason library and convert it to our data struct. We will see decoded trade events logged to the terminal by the end of the chapter.

Chapter 2 - Create a naive trading strategy - single trader without supervision

In this chapter, we will create our first naive trading strategy. We will generate another application inside our umbrella called naive. We will put data streamed to our streamer application to good use by sending it over to the naive application. We will start with a very basic solution consisting of a single process called trader that will utilise the GenServer behaviour. It will allow us to go through the complete trading cycle and give us something that "works".

Chapter 3 - Introduce PubSub as a communication method

To allow our trading strategy to scale to multiple parallel traders, we need to find a way to distribute the latest prices (trade events) to those multiple traders. We will introduce PubSub to broadcast messages from the streamer(s) to the trader(s). PubSub will allow us to break hardcoded references between applications in our umbrella and become a pattern that we will utilise moving forward.

Chapter 4 - Mock the Binance API

Besides historical prices (trade events), to perform backtesting, we need to be able to mock placing orders and get trade events back as they are filled. In this chapter, we will focus on developing the solution that will allow our traders to "trade" without contacting the Binance exchange(for people without Binance accounts). This will also allow us to backtest our trading strategy.

Chapter 5 - Enable parallel trading on multiple symbols

Our basic strategy implementation from the last chapter is definitely too basic to be used in a "production environment" - it can't be neither scaled nor it is fault-tolerant. In this chapter, we will upgrade our naive strategy to be more resilient. This will require a supervision tree to be created and will allow us to see different supervision strategies in action and understand the motivation behind using and stacking them.

Chapter 6 - Introduce a buy_down_interval to make a single trader more profitable

At this moment our Naive.Trader implementation will blindly place a buy order at the price of the last trade event. Whenever the Naive.Trader process will finish trade, a new Naive.Trader process will be started and it will end up placing a buy order at the same price as the price of the previous sell order. This will cost us double the fee without gaining any advantage and would cause further complications down the line, so we will introduce a buy_down_interval which will allow the Naive.Trader processes to place a buy order below the current trade event's price.

Chapter 7 - Introduce a trader budget and calculating the quantity

Since the second chapter, our Naive.Trader processes are placing orders with a hardcoded quantity of 100. In this chapter, we will introduce a budget that will be evenly split between the Naive.Trader processes using chunks. We will utilize that budget to calculate quantity (to be able to do that we need to fetch further step_size information from the Binance API).

Chapter 8 - Add support for multiple transactions per order

Our Naive.Trader implementation assumes that our orders will be filled within a single transaction, but this isn't always the case. In this chapter, we will discuss how could we implement the support for multiple transactions per order and race conditions that could occur between the bot and the Binance API.

Chapter 9 - Run multiple traders in parallel

With PubSub, supervision tree, buy down and budget in place we can progress with scaling the number of traders. This will require further improvements to our trading strategy like introducing a rebuy_interval. At the end of this chapter, our trading strategy will be able to start and run multiple traders in parallel.

Chapter 10 - Fine-tune trading strategy per symbol

Currently, the naive strategy works based on settings hardcoded in the leader module. To allow for fine-tuning the naive trading strategy per symbol we will introduce a new database together with the table that will store trading settings.

Chapter 11 - Supervise and autostart streaming

In the last chapter, we introduced a new database inside the naive application to store default settings, in this chapter, we will do the same for the streamer application. Inside the settings, there will be a status flag that will allow us to implement the autostarting functionality on initialization using Task abstraction.

Chapter 12 - Start, stop, shutdown, and autostart trading

To follow up after autostarting streaming we will apply the same trick to the trading supervision tree using Task abstraction. We will need to introduce a new supervision level to achieve the correct supervision strategy.

Chapter 13 - Abstract duplicated supervision code

As both the naive and the streamer applications contain almost the same copy-pasted code that allows us to start, stop and autostart workers. We will look into how could we abstract the common parts of that implementation into a single module. We will venture into utilizing the __using__ macro to get rid of the boilerplate.

Chapter 14 - Store trade events and orders inside the database

To be able to backtest the trading strategy, we need to have historical prices (trade events) and a list of orders that were placed stored in the database, which will be the focus of this chapter. At this moment, the latest prices (trade events) are broadcasted to PubSub topic and traders are subscribing to it. We will create a new application called data_warehouse inside our umbrella that will be responsible for subscribing to the same PubSub topics and storing incoming prices (trade events) in the Postgres database. We will update the Naive.Trader module to broadcast orders as traders will place them.

Then we will move on to adding supervision similar to the one from the naive and the streamer applications but this time we will show how we could avoid using both common module and macros by utilizing the Registry module.

Chapter 15 - Backtest trading strategy

In this chapter, we will be backtesting our trading strategy by developing a publisher inside the DataWarehouse application. It will stream trade events from the database to broadcast them to the TRADE_EVENTS:#{symbol} PubSub topic. It will use the same topic as data would be streamed directly from the Binance. From the trader's perspective, it won't any difference and will cause normal trading activity that will be stored inside the database to be analyzed later.

Chapter 16 - End-to-end testing

We've reached the stage where we have a decent solution in place, and to ensure that it's still working correctly after any future refactoring, we will add tests. We will start with the "integration"/"end-to-end"(E2E) test, which will confirm that the whole "trading" works. To perform tests at this level, we will need to orchestrate databases together with processes and broadcast trade events from within the test to cause our trading strategy to place orders. We will be able to confirm the right behaviour by checking the database after running the test.

Chapter 17 - Mox rocks

In the previous chapter, we’ve implemented the end-to-end test that required a lot of prep work, and we were able to see the downsides of this type of tests clearly. This chapter will focus on implementing a more granular test that will utilize the mox package to mock out the dependencies of the Naive.Trader. We will look into how the Mox works and how we will need to modify our code to use it.

Chapter 18 - Functional Elixir

In this chapter, we will venture into the functional world by looking at how could we improve our code to push side effects to the edge. We will revise the Naive.Trader code to abstract away our strategy code into a new module called Naive.Strategy. From this place, we will reorganise the code to maximise the amount of easily testable pure functions. Finally, we will explore hypothetical implementations that will allow us to inject data into function or even manage effects to aid testability. We will compare those to solutions built into Elixir like the with statement.

Chapter 19 - Idiomatic OTP

In the last chapter, we were looking into how we could reorganise the code to maximise the amount of pure code. In this chapter, we will look into different ways of implementing the OHLC(open-high-low-close) aggregator, considering similar optimisation but expanding to limit the number of processes to aid testability and maintainability.

Chapter 20 - Idiomatic trading strategy

We will use the knowledge gained in the last chapter to revise our Naive trading strategy so we will minimise the number of processes required to trade. We will move the functionalities provided by the Naive.Leader and the Naive.SymbolSupervisor into our strategy, taking care to put as much of it as possible into the pure part. In the end, our Naive.Trader will be able to manage multiple positions(trade cycles), and the vast majority of code previously scattered across multiple modules/processes will become easily testable pure functions inside our Naive.Strategy.

-Quantity 1+
  • The PDF/ePub versions of the book containing the first 20 finished chapters.

  • Length
    285 pages
  • Chapters
    20
  • The PDF/ePub versions of the book containing the first 20 finished chapters.
  • Length285 pages
  • Chapters20

Ratings

5.0
(6 ratings)
5 stars
100%
4 stars
0%
3 stars
0%
2 stars
0%
1 star
0%
Powered by

Checkout

Hands-on Elixir & OTP: Cryptocurrency trading bot

Enter your info to complete your purchase

Card

or pay with
You'll be charged US$10.