RakambdaOrg/ChannelPointsMiner

Improve mostTrusted outcome picker

Rakambda opened this issue · 2 comments

Rakambda commented

Description

Follow up of #259

  • Filter users with a too high ROI when performing calculations (configurable)
  • Keep only top n predictors to perform choose side to bet on (configurable)
FirstThread commented

The problem we currently have is that we currently get worse results with a higher amount of data. Sometimes I also had someone who had an ROI of 9000 or so, whose entry I then had to delete manually. We don't make a difference with ROI if someone has only bet once or has "experience". I have implemented a weight function for this, which I am currently testing. Currently I have a winrate of around 60-70%, for comparison before I had 40%.

  • 👍1
FirstThread commented

In the future, I may also try neural networks to make betting decisions. However, I'm not 100% sure how to implement this with the current state of the data, or if I should create a new table in the database.