• 

  Homepage

Use this shorter version:

CapCraft AI

Description of the Project

CapCraft AI is a basketball front office dashboard which allows users to experiment with building five-player lineups considering cap constraints, player value, on-court contributions and roles.

Motivation

Our motivation was the complexity NBA teams face when it comes to balancing between having stars and being able to work within their salary caps. While adding a valuable player to your roster could bring benefits, it might also cause some salary constraints or conflicts with other ball-handlers on the team.

Technologies

CapCraft AI was made using Streamlit, Pandas, NumPy and Plotly libraries. For our salary calculations we used public salaries from CSV file as well as backup dataset to ensure data quality.

In order to overcome unstable nature of live API, we developed deterministic proxies for BPM and Usage Rate. Then we calculated player value, creation overlap, payroll concentration and roster risk based on those values.

Challenges and Learning

This was mainly the difficulty of making the application both analytical and consistent. We managed to learn how to make stable data back-ups, construct transparent proxy measures, and demonstrate the complicated decision-making regarding rosters through dynamic charting and simulation.

What It Does

A user may select a team, compare the rosters, set cap limits, run simulations, compare players, and observe how all those affect payroll and production.

Built With

• github
• numpy
• pandas
• plotpy
• python
• streamlit