sansbacon / nbapr Goto Github PK

nbapr is a python-based player rater for nba fantasy. It uses simulation rather than the typical z-score approach used by ESPN and other sites. The output is the average amount of fantasy points a player contributes to a team (or in a specific category). This information is much more useful than a sum of z-scores which are not tied to any real values. It also accounts for the fact that a player's effect on a team is capped by the bounds of being first or last in a category. Z-scores know nothing about fantasy scoring rules, so they continue to penalize or reward a player in a category beyond the practical effect the player could have on team results.

8-Category Player Rater: https://sansbacon.github.io/nbapr/player-rater-8cat/

9-Category Player Rater: https://sansbacon.github.io/nbapr/player-rater-9cat/

Documentation: https://sansbacon.github.io/nbapr/

Source Code: https://github.com/sansbacon/nbapr

The key nbapr features are:

• Fast: takes advantage of pandas and numpy to run 50,000 simulations in less than 30 second.
• Interpretable results: nbapr ties stats to fantasy points by simulating numerous leagues of players. This is a more useful and comprehensible metric than a sum of z-scores across categories.
• Better results: Z-score based player raters are very sensitive to outliers and the initial selection of the player pool and tend to assign too much weight to players who dominate or tank a single category.
• Positional value: future iterations of nbapr will enforce position constraints, and thus give more insight than z-scores into relative position value.
• Pythonic: library is easy to use and extend as long as you are familiar with data analysis in python (pandas and numpy).

• Python 3.8+
• pandas 1.0+
• numpy 1.19+
• requests 2.0+

$ pip install nbapr

# example python code

This project is licensed under the terms of the MIT license.