Skip to content

Commit 96cf8c8

Browse files
csulitcsulit
committed
first commit
0 parents  commit 96cf8c8

File tree

1 file changed

+209
-0
lines changed

1 file changed

+209
-0
lines changed

‎README.md‎

Lines changed: 209 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,209 @@
1+
# Optimize trading strategy using Freqtrade
2+
3+
Short demo on building, testing and optimizing a trading strategy using Freqtrade.
4+
5+
The DevBootstrap YouTube screencast supporting this repo is [here](https://www.youtube.com/watch?v=wq3uLSDJxUQ). Enjoy! :)
6+
7+
## Alias Docker-Compose Command
8+
9+
First, I recommend to alias `docker-compose` to `dc` and `docker-compose run --rm "$@"` to `dcr` to save of typing.
10+
11+
Put this in your `~/.bash_profile` file so that its always aliased like this!
12+
13+
```
14+
alias dc=docker-compose
15+
dcr() { docker-compose run --rm "$@"; }
16+
```
17+
18+
Now run `source ~/.bash_profile`.
19+
20+
## Installing Freqtrade
21+
22+
Install and run [via Docker](https://www.freqtrade.io/en/stable/docker_quickstart/).
23+
24+
Now install the necessary dependencies to run Freqtrade:
25+
26+
```
27+
mkdir ft_userdata
28+
cd ft_userdata/
29+
# Download the dc file from the repository
30+
curl https://raw.githubusercontent.com/freqtrade/freqtrade/stable/docker-compose.yml -o docker-compose.yml
31+
32+
# Pull the freqtrade image
33+
dc pull
34+
35+
# Create user directory structure
36+
dcr freqtrade create-userdir --userdir user_data
37+
38+
# Create configuration - Requires answering interactive questions
39+
dcr freqtrade new-config --config user_data/config.json
40+
```
41+
42+
**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.)
43+
44+
## Config Bot
45+
46+
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.
47+
48+
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).
49+
50+
```
51+
"stake_amount" : "unlimited",
52+
"tradable_balance_ratio": 0.99,
53+
```
54+
55+
The above are used for Dry Runs and is the ['Dynamic Stake Amount'](https://www.freqtrade.io/en/stable/configuration/#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!)
56+
57+
```
58+
"tradable_balance_ratio": 0.2,
59+
"cancel_open_orders_on_exit": true
60+
```
61+
62+
For details of all available parameters, please refer to [the configuration parameters docs](https://www.freqtrade.io/en/stable/configuration/#configuration-parameters).
63+
64+
## Create a Strategy
65+
66+
So I've created a 'BBRSINaiveStrategy' based on [RSI](https://www.investopedia.com/terms/r/rsi.asp) and B[ollenger Bands](https://www.investopedia.com/terms/b/bollingerbands.asp). Take a look at the file [`bbrsi_naive_strategy.py`](ft_userdata/user_data/strategies/bbrsi_naive_strategy.py) file for details.
67+
68+
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`
69+
70+
For more details on Strategy Customization, please refer to the [Freqtrade Docs](https://www.freqtrade.io/en/stable/strategy-customization/)
71+
72+
## Remove past trade data
73+
74+
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`.
75+
76+
## Sandbox / Dry Run
77+
78+
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!).
79+
80+
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:
81+
82+
```
83+
dc up -d
84+
dc ps
85+
dc logs -f
86+
```
87+
88+
## Setup a pairs file
89+
90+
We will use Binance so we create a data directory for binance and copy our `pairs.json` file into that directory:
91+
92+
```
93+
mkdir -p user_data/data/binance
94+
cp pairs.json user_data/data/binance/.
95+
```
96+
97+
Now put whatever pairs you are interested to download into the `pairs.json` file. Take a look at the [pairs.json](ft_userdata/user_data/data/binance/pairs.json) file included in this repo.
98+
99+
## Download Data
100+
101+
Now that we have our pairs file in place, lets download the OHLCV data for backtesting our strategy.
102+
103+
```
104+
dcr freqtrade download-data --exchange binance -t 15m
105+
```
106+
107+
List the available data using the `list-data` sub-command:
108+
109+
```
110+
dcr freqtrade list-data --exchange binance
111+
```
112+
113+
Manually inspect the json files to examine the data is as expected (i.e. that it contains the expected `OHLCV` data requested).
114+
115+
## List the available data for backtesting
116+
117+
Note to list the available data you need to pass the `--data-format-ohlcv jsongz` flag as below:
118+
119+
```
120+
dcr freqtrade list-data --exchange binance
121+
```
122+
123+
## Backtest
124+
125+
Now we have the data for 1h and 4h OHLCV data for our pairs lets Backtest this strategy:
126+
127+
```
128+
dcr freqtrade backtesting --datadir user_data/data/binance --export trades --stake-amount 100 -s BBRSINaiveStrategy -i 15m
129+
```
130+
131+
For details on interpreting the result, refer to ['Understading the backtesting result'](https://www.freqtrade.io/en/stable/backtesting/#understand-the-backtesting-result)
132+
133+
## Plotting
134+
135+
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.
136+
137+
Note that the `plot_config` that is contained in the strategy will be applied to the chart.
138+
139+
```
140+
dcr freqtrade plot-dataframe --strategy BBRSINaiveStrategy -p ALGO/USDT -i 15m
141+
```
142+
143+
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.
144+
145+
## Optimize
146+
147+
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:
148+
149+
```
150+
dcr freqtrade new-hyperopt --hyperopt BBRSIHyperopt
151+
```
152+
153+
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:
154+
155+
```
156+
dcr freqtrade hyperopt --hyperopt BBRSIHyperopt --hyperopt-loss SharpeHyperOptLoss --strategy BBRSINaiveStrategy -i 15m
157+
```
158+
159+
## Update Strategy
160+
161+
Apply the suggested optimized results from the Hyperopt to the strategy. Either replace the current strategy or create a new 'optimized' strategy.
162+
163+
## Backtest
164+
165+
Now we have updated our strategy based on the result from the hyperopt lets run a backtest again:
166+
167+
```
168+
dcr freqtrade backtesting --datadir user_data/data/binance --export trades --stake-amount 100 -s BBRSIOptimizedStrategy -i 15m
169+
```
170+
171+
## Sandbox / Dry Run
172+
173+
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!
174+
175+
```
176+
"max_open_trades": 10,
177+
"stake_currency": "USDT",
178+
"stake_amount" : "unlimited",
179+
"tradable_balance_ratio": 0.99,
180+
"fiat_display_currency": "USD",
181+
"timeframe": "15min",
182+
"dry_run": true,
183+
"dry_run_wallet": 1000,
184+
```
185+
186+
## View Dry Run via Freq UI
187+
188+
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:
189+
190+
```
191+
...
192+
193+
"api_server": {
194+
"enabled": true,
195+
"listen_ip_address": "0.0.0.0",
196+
"username": "Freqtrader",
197+
"password": "secretpass!",
198+
P
199+
...
200+
```
201+
202+
In the docker-compose.yml file also map the ports like so:
203+
204+
```
205+
ports:
206+
- "127.0.0.1:8080:8080"
207+
```
208+
209+
Then you can access the Freq UI via a browser at [http://127.0.0.1:8080/](http://127.0.0.1:8080/). You can also access and control the bot via a REST API too!

0 commit comments

Comments
 (0)