Optimize Trading Strategies Using Freqtrade

Overview

Optimize trading strategy using Freqtrade

Short demo on building, testing and optimizing a trading strategy using Freqtrade.

The DevBootstrap YouTube screencast supporting this repo is here. Enjoy! :)

Alias Docker-Compose Command

First, I recommend to alias docker-compose to dc and docker-compose run --rm "$@" to dcr to save of typing.

Put this in your ~/.bash_profile file so that its always aliased like this!

alias dc=docker-compose
dcr() { docker-compose run --rm "$@"; }

Now run source ~/.bash_profile.

Installing Freqtrade

Install and run via Docker.

Now install the necessary dependencies to run Freqtrade:

mkdir ft_userdata
cd ft_userdata/
# Download the dc file from the repository
curl https://raw.githubusercontent.com/freqtrade/freqtrade/stable/docker-compose.yml -o docker-compose.yml

# Pull the freqtrade image
dc pull

# Create user directory structure
dcr freqtrade create-userdir --userdir user_data

# Create configuration - Requires answering interactive questions
dcr freqtrade new-config --config user_data/config.json

NOTE: Any freqtrade commands are available by running dcr freqtrade <command> <optional arguments>. So the only difference to run the command via docker-compose is to prefix the command with our new alias dcr (which runs docker-compose run --rm "$@" ... see above for details.)

Config Bot

If you used the new-config sub-command (see above) when installing the bot, the installation script should have already created the default configuration file (config.json) for you.

The params that we will set to note are (from config.json). This allows all the available balance to be distrubuted accross all possible trades. So in dry run mode we have a default paper money balance of 1000 (can be changed using dry_run_wallet param) and if we set to have a max of 10 trades then Freqtrade would distribute the funds accrosss all 10 trades aprroximatly equally (1000 / 10 = 100 / trade).

"stake_amount" : "unlimited",
"tradable_balance_ratio": 0.99,

The above are used for Dry Runs and is the 'Dynamic Stake Amount'. For live trading you might want to change this. For example, only allow bot to trade 20% of excahnge account funds and cancel open orders on exit (if market goes crazy!)

"tradable_balance_ratio": 0.2,
"cancel_open_orders_on_exit": true

For details of all available parameters, please refer to the configuration parameters docs.

Create a Strategy

So I've created a 'BBRSINaiveStrategy' based on RSI and Bollenger Bands. Take a look at the file bbrsi_naive_strategy.py file for details.

To tell your instance of Freqtrade about this strategy, open your docker-compose.yml file and update the strategy flag (last flag of the command) to --strategy BBRSINaiveStrategy

For more details on Strategy Customization, please refer to the Freqtrade Docs

Remove past trade data

If you have run the bot already, you will need to clear out any existing dry run trades from the database. The easiest way to do this is to delete the sqlite database by running the command rm user_data/tradesv3.sqlite.

Sandbox / Dry Run

As a quick sanity check, you can now immediately start the bot in a sandbox mode and it will start trading (with paper money - not real money!).

To start trading in sandbox mode, simply start the service as a daemon using Docker Compose, like so and follow the log trail as follows:

dc up -d
dc ps
dc logs -f

Setup a pairs file

We will use Binance so we create a data directory for binance and copy our pairs.json file into that directory:

mkdir -p user_data/data/binance
cp pairs.json user_data/data/binance/.

Now put whatever pairs you are interested to download into the pairs.json file. Take a look at the pairs.json file included in this repo.

Download Data

Now that we have our pairs file in place, lets download the OHLCV data for backtesting our strategy.

dcr freqtrade download-data --exchange binance -t 15m

List the available data using the list-data sub-command:

dcr freqtrade list-data --exchange binance

Manually inspect the json files to examine the data is as expected (i.e. that it contains the expected OHLCV data requested).

List the available data for backtesting

Note to list the available data you need to pass the --data-format-ohlcv jsongz flag as below:

dcr freqtrade list-data --exchange binance

Backtest

Now we have the data for 1h and 4h OHLCV data for our pairs lets Backtest this strategy:

dcr freqtrade backtesting --datadir user_data/data/binance --export trades  --stake-amount 100 -s BBRSINaiveStrategy -i 15m

For details on interpreting the result, refer to 'Understading the backtesting result'

Plotting

Plot the results to see how the bot entered and exited trades over time. Remember to change the Docker image being referenced in the docker-compose file to freqtradeorg/freqtrade:develop_plot before running the below command.

Note that the plot_config that is contained in the strategy will be applied to the chart.

dcr freqtrade plot-dataframe --strategy BBRSINaiveStrategy -p ALGO/USDT -i 15m

Once the plot is ready you will see the message Stored plot as /freqtrade/user_data/plot/freqtrade-plot-ALGO_USDT-15m.html which you can open in a browser window.

Optimize

To optimize the strategy we will use the Hyperopt module of freqtrade. First up we need to create a new hyperopt file from a template:

dcr freqtrade new-hyperopt --hyperopt BBRSIHyperopt

Now add desired definitions for buy/sell guards and triggers to the Hyperopt file. Then run the optimization like so (NOTE: set the time interval and the number of epochs to test using the -i and -e flags:

dcr freqtrade hyperopt --hyperopt BBRSIHyperopt --hyperopt-loss SharpeHyperOptLoss --strategy BBRSINaiveStrategy -i 15m

Update Strategy

Apply the suggested optimized results from the Hyperopt to the strategy. Either replace the current strategy or create a new 'optimized' strategy.

Backtest

Now we have updated our strategy based on the result from the hyperopt lets run a backtest again:

dcr freqtrade backtesting --datadir user_data/data/binance --export trades --stake-amount 100 -s BBRSIOptimizedStrategy -i 15m

Sandbox / Dry Run

Before you run the Dry Run, don't forget to check your local config.json file is configured. Particularly the dry_run is true, the dry_run_wallet is set to something reasonable (like 1000 USDT) and that the timeframe is set to the same that you have used when building and optimizing your strategy!

"max_open_trades": 10,
"stake_currency": "USDT",
"stake_amount" : "unlimited",
"tradable_balance_ratio": 0.99,
"fiat_display_currency": "USD",
"timeframe": "15min",
"dry_run": true,
"dry_run_wallet": 1000,

View Dry Run via Freq UI

For use with docker you will need to enable the api server in the Freqtrade config and set listen_ip_address to "0.0.0.0", and also set the username & password so that you can login like so:

...

"api_server": {
  "enabled": true,
  "listen_ip_address": "0.0.0.0",
  "username": "Freqtrader",
  "password": "secretpass!",
P
...

In the docker-compose.yml file also map the ports like so:

ports:
  - "127.0.0.1:8080:8080"

Then you can access the Freq UI via a browser at http://127.0.0.1:8080/. You can also access and control the bot via a REST API too!

Owner
DevBootstrap
Full Stack Blockchain Developer Tutorials
DevBootstrap
A platform for intelligent agent learning based on a 3D open-world FPS game developed by Inspir.AI.

Wilderness Scavenger: 3D Open-World FPS Game AI Challenge This is a platform for intelligent agent learning based on a 3D open-world FPS game develope

46 Nov 24, 2022
3.8% and 18.3% on CIFAR-10 and CIFAR-100

Wide Residual Networks This code was used for experiments with Wide Residual Networks (BMVC 2016) http://arxiv.org/abs/1605.07146 by Sergey Zagoruyko

Sergey Zagoruyko 1.2k Dec 29, 2022
Transformer Tracking (CVPR2021)

TransT - Transformer Tracking [CVPR2021] Official implementation of the TransT (CVPR2021) , including training code and trained models. We are revisin

chenxin 465 Jan 06, 2023
Point Cloud Denoising input segmentation output raw point-cloud valid/clear fog rain de-noised Abstract Lidar sensors are frequently used in environme

Point Cloud Denoising input segmentation output raw point-cloud valid/clear fog rain de-noised Abstract Lidar sensors are frequently used in environme

75 Nov 24, 2022
🗣️ Microsoft Edge TTS for Home Assistant, no need for app_key

Microsoft Edge TTS for Home Assistant This component is based on the TTS service of Microsoft Edge browser, no need to apply for app_key. Install Down

152 Dec 31, 2022
Metric learning algorithms in Python

metric-learn: Metric Learning in Python metric-learn contains efficient Python implementations of several popular supervised and weakly-supervised met

1.3k Jan 02, 2023
The code used for the free [email protected] Webinar series on Reinforcement Learning in Finance

Reinforcement Learning in Finance [email protected] Webinar This repository provides the code f

Yves Hilpisch 62 Dec 22, 2022
MBPO (paper: When to trust your model: Model-based policy optimization) in offline RL settings

offline-MBPO This repository contains the code of a version of model-based RL algorithm MBPO, which is modified to perform in offline RL settings Pape

LxzGordon 1 Oct 24, 2021
PAWS 🐾 Predicting View-Assignments with Support Samples

This repo provides a PyTorch implementation of PAWS (predicting view assignments with support samples), as described in the paper Semi-Supervised Learning of Visual Features by Non-Parametrically Pre

Facebook Research 437 Dec 23, 2022
Fast and simple implementation of RL algorithms, designed to run fully on GPU.

RSL RL Fast and simple implementation of RL algorithms, designed to run fully on GPU. This code is an evolution of rl-pytorch provided with NVIDIA's I

Robotic Systems Lab - Legged Robotics at ETH Zürich 68 Dec 29, 2022
Supervised 3D Pre-training on Large-scale 2D Natural Image Datasets for 3D Medical Image Analysis

Introduction This is an implementation of our paper Supervised 3D Pre-training on Large-scale 2D Natural Image Datasets for 3D Medical Image Analysis.

24 Dec 06, 2022
python 93% acc. CNN Dogs Vs Cats ( Pytorch )

English | 简体中文(测试中...敬请期待) Cnn-Classification-Dog-Vs-Cat 猫狗辨别 (pytorch版本) CNN Resnet18 的猫狗分类器,基于ResNet及其变体网路系列,对于一般的图像识别任务表现优异,模型精准度高达93%(小型样本)。 项目制作于

apple ye 1 May 22, 2022
NER for Indian languages

CL-NERIL: A Cross-Lingual Model for NER in Indian Languages Code for the paper - https://arxiv.org/abs/2111.11815 Setup Setup a virtual environment Th

Akshara P 0 Nov 24, 2021
Collection of generative models in Pytorch version.

pytorch-generative-model-collections Original : [Tensorflow version] Pytorch implementation of various GANs. This repository was re-implemented with r

Hyeonwoo Kang 2.4k Dec 31, 2022
Audio Domain Adaptation for Acoustic Scene Classification using Disentanglement Learning

Audio Domain Adaptation for Acoustic Scene Classification using Disentanglement Learning Reference Abeßer, J. & Müller, M. Towards Audio Domain Adapt

Jakob Abeßer 2 Jul 06, 2022
TOOD: Task-aligned One-stage Object Detection, ICCV2021 Oral

One-stage object detection is commonly implemented by optimizing two sub-tasks: object classification and localization, using heads with two parallel branches, which might lead to a certain level of

264 Jan 09, 2023
Multi-Glimpse Network With Python

Multi-Glimpse Network Our code requires Python ≥ 3.8 Installation For example, venv + pip: $ python3 -m venv env $ source env/bin/activate (env) $ pyt

9 May 10, 2022
PerfFuzz: Automatically Generate Pathological Inputs for C/C++ programs

PerfFuzz Performance problems in software can arise unexpectedly when programs are provided with inputs that exhibit pathological behavior. But how ca

Caroline Lemieux 125 Nov 18, 2022
Kaggle Feedback Prize - Evaluating Student Writing 15th solution

Kaggle Feedback Prize - Evaluating Student Writing 15th solution First of all, I would like to thank the excellent notebooks and discussions from http

Lingyuan Zhang 6 Mar 24, 2022