just a conceptual (just methods + one fit)

fitnpredict.py

#Sidharth Rajaram, 2017, part of NBA-Predict
from sklearn import svm
import numpy as np

def combineData(values1, values2):
    combined_stats = []
    for a in range(values1.size):
        result_component = [values1[a],values2[a]]
        combined_stats.append(result_component)
    combined_stats = np.array(combined_stats)
    return combined_stats

classifier = svm.SVC() #rbf kernel by default
# X represents dataset, Y represents labels 
scoring_ability = combineData(X[:,3], X[:,45])
def offense(player):
	result = 0.0
	classifier.fit(scoring_ability, Y)
	result += classifier.predict([[ player[1][3], player[0][23] ]])[0]
	return result

#just like this, many other factors are predicted and the cumulative sum represents the composite STAR rating