Crypto Trading Bot with Gekko on Binance — Day 1 Testing & Reflections
In this post I will be discussing my experiences testing out (live trading) an open source crypto trading bot “Gekko” on Binance Exchange. Profit/Loss, strategy source code and tips/tricks are included within this writeup.
as you are all aware I have been building my own crypto trading bot in Python for use on the Binance Exchange for a couple of months, with all coding available through my YouTube Channel.
While this project initially seemed quite simple and straightforward, the complexity of the endeavor soon took over the project. I was able to code 4 very usable bots, however, further development of user experience, AI, and trading functionality requires significant software development. Despite these challenges, many of these python scripts are still under development within private discord group, including a GUI for use with the bot.
I followed the installation instructions for Gekko (which were quite simple), and implemented pre-designed bot strategies (Such as a Neural Network-based strategy) by downloading, unzipping and placing files within proper folders. After this set up (approximately 15 minutes for installation, 3–4 more hours further research, and 15 minutes to install Neural Network strategy) I was recording data and trading algorithmicly through Binance.
I was amazed at the simplicity of installation, resources available for gekko, and powerful abilities of the gekko crypto trading bot platform.
Installation of Gekko — User Interface of Gekko Trading Bot
The creator of Gekko has created an instructional video which shows installation of Gekko, as well as the User Interface in the following video:
After watching the above videos, you should feel confident in starting a gekko bot on your own computer, collecting data, implementing back testing, and Trading successfully on a cryptocurrency exchange.
Required Reading: Neural Networks
Neural Networks are attempts to replicate Neural Learning with code. A fantastic introduction to this topic is available from Natalia Pattarone where she describes the code and theory behind a neural network based off of white-blood cells and bacteria. Natalia’s article references further reading describing how to build a Neural Network and the technical detail behind this new technology.
An additional required reading is a post by Nan Li describing how scientific thought has become too reliant upon AI/Machine Learning. While this may seem to be a strange recommendation on the required reading list, I have placed it on here due to the fact that we can not treat the Neural Network algorithm as a simple black box in a blind search for profit, and my own personal beliefs that we should treat this new ‘tool’ like a tool for discovery without compromising real-world applicability (as is covered in Nan’s post) and we must approach this method in as scientifically as possible.
Additional research in this area includes applying ‘Genetic Algorithms’ to Financial Markets, including the cryptocurrency and bitcoin markets. Discussion concerning the downside and areas of concern regarding Genetic Algorithms is available through stackexchange, and considering these concerns, we will be implementing GA in the future using Gekko.
With this research, I was curious to apply this technology to the cryptocurrency / bitcoin trading bot from Gekko.
Results from implementation of Neural Network Crypto Trading Strategy
A Neural Network-based crypto trading strategy file for Gekko was found within the xFFFF github repository of Gekko Trading Strategies, which implements the DeepQLearning algorithm produced by Blake Milner and Jeff Soldate. After implementing a 1-minute candle and increased trading frequency by reducing the absolute values of trading thresholds (to 0.5 and -0.5), I implemented the bot on BNB/BTC because I had some BNB in my account from previous botting attempts, and BNB was in a severe bear market.
Note: In the Gekko suite, green dots are where you purchased, and red dots are where you sell. Ideally the green dots will be below the red dots in order to profit as a trading bot.
After running the bots for just under 8 hours (using 1-minute candles), the bot had made a number of buys and sells based on its own signals, and automatically placed an order for these trades within the Gekko Crypto Trading Bot.
The above graph is after running the Neural Network for 21.5 hours on the same setting. As you can see, BNB is in a bear market, but the Neural network seems to buy/sell at quite good times, in particular is the Buying of bottom and selling of top near the middle of the graph. To me the Neural Network is working very effectively and I will continue research on this topic, while also attempting to push the bot to trade (buy/sell) more often.
Neural Network Crypto Trading Strategy on VEN/BNB
Profit report — 8 Hours
Start balance: 0.07979
Current balance: 0.08024
Market: -0.00001 BTC (-0.58588 %)
Profit: 0.00045 BTC (0.56378 %)
Alpha: 0.58633 BTC
This strategy is still running and future updates will be adding to the Medium Blog here.
Closing Thoughts on Gekko Crypto Trading Bot Software
After spending a few days researching Gekko, and a full day testing and implementing the software which has been developed for this open source crypto trading bot project: I am very impressed by the level of development and support available for the Gekko project, the amount of functionality built in, and ability to test/iterate/change strategies among a larger community of programmers is a very large selling point to me.
In conclusion, The Gekko Crypto Trading Bot Open Source software project is a fantastic resource for anyone interested in trading cryptocurrencies due to its immensely powerful software, ease of installation all at a cost of ‘free’.
I am currently running a bot on this platform and will continue researching and implementing new trading strategies, because the Gekko software has taken care of all of that part already.
As a trader, I simply connect to the user-friendly web-page based AI, determine what data to collect, and apply back testing using a variety of community-created trading strategies, which I can then test and modify for use in live trading.