What I Learned from Testing a Bot

I read this post by @camb and it explained step by step how to deploy a bot. I highly recommend trying it out although I would like to include a warning:

If you want to try running your own bot you will have to provide a credit card number even if you are choosing the free package.

As I explained in my introductory post I am here to learn about A.I. and Machine Learning. I do not intend to keep running this bot as is. I intend to constantly change the algorithm and upgrade the bot to see if it has value. I am not gonna run the bot and let it just coast and collect money. My intention is to post the algorithm and post the results, so that I can learn and so that others can learn. I don't think bots would be all that effective in earning money without you buying Steem Power for them anyways. The bot might not be great for earning money, but it is great for learning about algorithms and how to deploy bots.

The instructions were pretty user-friendly. I chose to follow these instructions because these instructions were for a bot that could run on the cloud. This means that I am not running the bot on my computer, so my computer doesn't have to be on for the bot to run.

After following the default instructions the bot did not run. I end up going to the GitHub page and discovering that the default settings for the bot were designed for a time with considerably fewer users. After adjusting "MAX_POST_TO_READ" to a value of 1200 in the configuration settings the bot was able to run. I modified the configuration, but I did not modify the algorithm section.

The first time I ran the bot after adjusting the settings I thought that it failed because Horuku (cloud platform) stated that the bot failed, but it actually did not fail. I ended up running the bot again, but deploying it based on a timer rather than manually starting the bot and the second time it succeeded again.

On the first run, the bot read 241 post and upvoted 18 of the post before quitting because the robot fell below 70% voting power. On the second run the bot went through 699 post and upvoted 16 post before falling below 50% voting power.

Here is the default algorithm used:

In the first run the robot was allowed to deplete the voting power from 100% to 70% and in the second run the bot was allowed to deplete the voting power from 70% to 50%.

The bot was set to process a maximum of 1200 post, but the bot did not process that many post. I suspect that the bot will read a post and make a voting decision based on the score that it assigns and the binding constraint ends up being voting power.

What I will adjust the next time I run the bot is to reduce the voting power so the bot doesn't run out of voting power before checking all of the 1200 post. I also want the bot to be able to spread the vote a little more. I will remove the "post_has_english_language_use". I don't want the bot to be racist. Please let me know if you want me to test other algorithms. I want to build a bot based on what people would like to see tested. I also intend to read the manual and try to understand what each line of the algorithm does and how the algorithm assigns a score.

To see how the robot votes you can go to steemworld.org/@alphazero I currently have the bot scheduled to run at the 30 minute mark every hour.