These projections have been sorted in descending order based off one of two prediction models (noted as ‘GBR’ and ‘LR’ predictions). Because each prediction model uses a different algorithm, they produce values that can vary by a substantial amount. The reasoning behind sorting by one model over the other is due to their algorithmic properties – how they come up with their predictions.
Without getting too technical, GBR, or gradient boosting regression, uses decision trees to predict values. Think of this like a game of Heads Up where the algorithm asks thousands of questions to determine the answer. After a while it ‘learns to ask’ the right questions in order to give the best prediction.
Logistic regression (LR) takes a slightly different approach in that it uses more statistical probability with its determinations. If we keep the Heads Up example, this would be similar to getting a hint of the answer before making a prediction.
Full disclosure: I am not a professional data scientist! I am a master’s student with a focus in big data analytics, and an interest particularly in machine learning. Namely because I can use it with fantasy football research. The projections are the result from a self-directed project. These projections are meant to be used as a baseline for further analysis, and not a complete player-to-number projection.