luctivud / codeforces-future-rating-predictor

Predict your future ratings from your past performance.

The working of this model will be updated here, in markdown, the work is in progress. Till then you can enjoy the app.

Link to Web App: https://cffrpredict.netlify.app/

Link to the codeforces post: https://codeforces.com/blog/entry/92017

Codeforces-Future-Rating-Predictor

Currently, it works on a statistical approach with a few randomisation techniques and outlier management technique. A modified form of K-means clustering has been used to detect the upper bound of rating and a free-form mean is used to get the lower bound of the rating.

Currently just two clusters are used and the data is modified based on medians and random seeds. The actual rating prediction mechanism is based on the values provided by the seeds.

K-Means employed rating is given at least 2x more preference than the normal mean employed rating, and the final rating is calculated.

Currently, the model needs the following input from the user:

• max_rating
  
• current_rating
  
• ratings_change1
  
• ratings_change2
  
• ratings_change3
  
• lucky_number
  

Future Development:

To make it work under a Machine Learning model with the help of Codeforces API.