reversion-algo

Heroku worker for reversion algorithm

Credit: https://github.com/sturlese/pylivetrader/tree/mean_reversion/examples

Setup

First you will need to set up a heroku app and install a few buildpacks. Navigate to https://dashboard.heroku.com/apps/ and create a new application using the python buildpack.

Once you have the app created click on the settings tab and go to the Buildpacks area to add the two extra build packs we need. We will be adding a talib installer as well as heroku-buildpack-run to execute bash commands during deployment.

In the buildpacks area add the links:

• https://github.com/numrut/heroku-buildpack-python-talib
• https://github.com/weibeld/heroku-buildpack-run

The order is important

Make sure you have python followed by ta-lib followed by heroku-buildpack-run in the buildpacks area. If they are out of order the algorithim may not deploy properly.

Next set your environment variables in heroku. Navigate to https://dashboard.heroku.com/apps/

From there go to settings and Reveal Config Vars. In this section you will set your base_url, key, and secret for alpaca. use the string before the = for the KEY and the string after as the VALUE.

export APCA_API_BASE_URL=bbb
export APCA_API_KEY_ID=xxx
export APCA_API_SECRET_KEY=yyy

Now you can choose to deploy from the heroku tab. Navigate to Deploy and link your github repo. Choose to deploy from the master branch and click deploy.

After deployment navigate over to the Resources tab. Here you will see a line that states worker pylivetrader run -f reversion.py --data-frequency minute -b paper followed by a switch, dollar amount, and pen. Click the pen and move the slider. Click confirm to activate the worker.

Verify

You can verify the app is running by navigating to the logs in your dashboard. To do this click the More drop down and select logs. You should see output similar to:

2019-04-09T06:20:58.132013+00:00 app[api]: Scaled to worker@1:Free by user khays@csumb.edu
2019-04-09T06:21:10.019054+00:00 heroku[worker.1]: Starting process with command `pylivetrader run -f reversion.py --data-frequency minute -b paper`
2019-04-09T06:21:10.841619+00:00 heroku[worker.1]: State changed from starting to up
2019-04-09T06:21:13.341974+00:00 app[worker.1]: [heroku-exec] Starting
2019-04-09T06:21:24.059524+00:00 app[worker.1]: [2019-04-09 06:21:24.059296] INFO: Algorithm: livetrader start running with backend = paper data-frequency = minute
2019-04-09T06:21:25.793038+00:00 app[worker.1]: [2019-04-09 06:21:25.792790] DEBUG: reversion: Initializing...
2019-04-09T06:21:25.805833+00:00 app[worker.1]: [2019-04-09 06:21:25.805718] DEBUG: reversion: Done initializing.

You can also verify the logs through an ssh tunnel. You will need the Heroku CLI at https://devcenter.heroku.com/articles/heroku-cli#download-and-install

launch a powershell and enter the following commands.

heroku login
heroku logs -a <app_name>

You will see a prompt

heroku: Press any key to open up the browser to login or q to exit:

press any key besides q when prompted.

Maintenance

All trading algorithms need routine maintenance. Most algorithms need to run continuously without interruption to save variables accrued on the heap during its runtime. We must make sure the algorithm is running cleanly to ensure we don't have to re-deploy (restart) the algorithm and lose the heap variables.

After first trading day:

• verify successful run
• verify program did not exit

weekly:

• check for web_socket errors
• verify successful run
• verify program is trading at proper time intervals