I have been running this bot for about a week now. I have been messing around with the algorithms and fine-tuning them. The initial algorithm seems to vote too often and the bot would run out of voting power quickly. I have settled on a new algorithm that would vote for high quality post as the community sees it.
The question is how do you define high quality post? Embedded in the reward pool algorithm is one definition of high quality:
The posts with the highest rewards are of the highest quality.
In order to make an efficient curation bot you must then try to predict which post will earn the most rewards. With these assumptions in place I created a very rough model for predicting what post will do well and this model is embedded in my algorithm:
After creating the algorithm the next step is to run the bot and see if it does indeed do a good job of predicting high reward post. When you run the bot you are trying to make predictions so you have to wait for the payout data to come in. As you can, the algorithm has seven variables and the coefficients are linear. The minimum and maximum values specify the range at which the variable is included in the model. I know for certain that my model is not perfect because my model was built on my assumptions and I have not collected data to test my assumptions.
There are also many other variables that could be added to the model and if you collect enough data you might be able to use regressional analysis to build a model that could be used to direct your bot. Of course, you can also add variables with negative coefficients in order to improve how well your model functions. The only problem is that even if you could predict what post would be successful it would still be difficult for you to earn curation rewards when you are a minnow. This is the current state of my experiment and my results so far. Please comment with your thoughts.