A trading strategy shared by kjmsb2 on this post.
I have been using this method for a while with pretty consistent results.
I use 3 pairs on the daily charts (chart time frame is actually irrelevant): EURJPY, EURUSD, USDJPY. This gives me a hedge between the 3 positions.
At the opening hour of Forex trading (5pm EST Sunday for me) I place both a pending buy-stop and sell-stop order 30 pips away from the opening price for each pair.
I place stop-loss at 50 pips and take-profit at 150 pips for each. I also instruct MT4 to use a trailing stop of 500 points (50 pips) once the position moves in to profit.
That's it.
I now wait for 1 to 3 orders to fill and either hit target or stop. If it hits stop I will re-enter a new pending order the same as the one just stopped.
If my targets are hit... GREAT! If not, at Friday 3pm EST I close all open positions and cancel any open pending orders.
Position size for me is 1% of account size on a 50 pip stop loss for each position.
Consistent results with literally only a few MINUTES per week of management.
This version doesn't implement the trailing stop.
Clone the repository
git clone git@github.com:26medias/bot-lazy-trader.git
Open a console, cd into the directory, then install the dependencies:
npm install
Create env vars to save your Oanda API key & your Oanda Account Number (or hard-code them in app.js line 18 & 19):
OANDA_KEY=13a5f4284fa8598ac37c4r5e0ef0fa3a-cr45169bceer7b1318c0cde3292o15e9
OANDA_ACC=101-001-3387465-001
(invalid keys & acc number, just for the example. Replace with your owns.)
Test locally:
node test
Unless it's sunday 5pm EST, no positions will be opened.
You can force the code to open positions as if it was time to trade, to ensure that your settings are correct. In app.js on line 15, change testMode from false to true.
Open app.js to edit the settings. You can edit:
- Test mode (to force taking positions no matter what day/time it is)
- Risk per position (default at %)
- The pairs to trade (Default: "EUR_USD","EUR_JPY","USD_JPY" but you can add more)
- The timezone to use (defaut is America/New_York)
- The day & time that are valid to open positions (default sunday at 5pm)
- The limit distance (default 30)
- The take-profit distance (default 150)
- The stop-loss distance (default 50)
On your AWS console, open CodeStar, and filter to only show the templates for Web Services in NodeJS, and select the "AWS Lambda" template.
Select a name, then connect your github account.
Click Next until the end. Leave everything as default.
CodeStar will create a new github repository on your github account. It might take a few minutes.
Copy the following files into your repository:
- app.js
- bot.js
- core.js
- index.js
- oanda.js
- package.json
- test.js
- .gitignore
- buildspec.yml
To deploy, just push to github. CodeStar will monitor your github repository and trigger a rebuild of your lambda function everytime there is a new commit to master.
git add --all && git commit -m "Base code" && git push
It takes a few minutes for your code to deploy & build on lambda.
You can follow the deployment and investigate any deployment issues by looking at your CodePipeline console. Open CodePipeline, locate your bot, and the UI is pretty self-explanatory:
Once deployed, you can setup & test your bot.
By default, your lambda function has a 3 seconds timeout & barely any memory allocated. That won't work for this.
Go to your AWS console, then open AWS Lambda.
Look for your bot in the list:
When you are on your bot's lambda settings, scroll down until you see Basic Settings:
Even though AWS will charge you more when you use more memory, you also get charged for the execution time. Counter-intuitively, it's usually cheaper to allocate more memory, as it speeds up the execution time quite significantly.
Here are my settings:
Now you'll want to setup the env vars, to provide the Oanda API key & account number without having to hard-code them in your code and risk accidentally pushing them to github.
Scroll back up and locate Environment variables:
Create the OANDA_KEY and OANDA_ACC env vars, along with their values:
There is a link on your OANDA fxTrade account profile page titled “Manage API Access” (My Account -> My Services -> Manage API Access). From there, you can generate a personal access token to use with the OANDA API, as well as revoke a token you may currently have.
Don't forget to save, and you're ready to test.
Scroll all the way up, locate the test UI and click "Select a test event":
Create a new test. Leave the parameters empty (we don't need any), and give it a neme, then click Create:
Now select & execute your test:
In your AWS console, open CloudWatch.
Go to Events > Rules, then click Create Rule:
Select schedule then Cron expression.
I asked Gogole to convert 5pm EST to UTC, ended up with 9pm UTC, so I setup the rule: Execute once every sunday at 9pm:
0 21 ? * SUN *
I'm not 100% of this, so we'll see if this executes at the right time. Pretty sure that won't work when daylight saving changes again in EST or UTC. Executing once per hour might be a safer bet to test. The bot won't trade if it's not the right time anyway and your usage will be covered by the AWS free tier if you have nothing else running:
0/60 * ? * SUN *
Check the official doc to understand the cron format.
If you want to execute your bot every hour Monday to Friday for example, you'd use:
0/60 * ? * MON-FRI *
On the right side, select your lambda function:
Leave everything as default and click Configure Details
On the next step, give your rule a name, a description, then click Create.
To check the execution logs of your bot and make sure everythign is working as expected, in your AWS console open CloudWatch.
On the left click Logs then look for the log starting with /aws/lambda/awscodestar- + the name of your bot.
If you don't find a log group for your bot, that means your bot hasn't been executed yet.
Wait for it to execute or execute it manually or via a test to start creating logs.
In your AWS console open API Gateway. Look for the name of your bot and go to Stages.
Click "Prod", and you'll find your endpoint's url:
Open that url in your browser to execute the bot.
The bot doesn't return anything as an HTTP response right now, you you should get {} as a response.