Python 3 + 2 fixes

.gitignore

@@ -0,0 +1,19 @@
+*.py[cod]
+__pycache__/
+__pycache__
+
+# OSX
+.DS_Store
+
+# Eclipse
+.project
+.pydevproject
+/.settings
+
+# Do not commit data files
+/data/*.csv
+/models/*.pkl
+/models/*.npy
+
+# generally do not commit node_modules (controversial...)
+/node_modules

README.md

@@ -12,7 +12,8 @@ The Armchair Analysis data is *not free*. It costs $49 (one-time) to gain access
 data from the 2000-2014 NFL seasons. There is also a professional package that will provide
 weekly updates. Without the Armchair Analysis data, you will not be able to use much of this code.
 
-This code currently requires Python 2.7 and is not Python 3 compliant to our knowledge. Questions about the Python code can be directed to [Trey Causey](mailto:[email protected]).
+This code currently requires Python 2.7 and or Python 3.4+. 
+Questions about the Python code can be directed to [Trey Causey](mailto:[email protected]).
 
 Please note that none of the file operations are supported on Windows.
 
@@ -27,6 +28,10 @@ NOTE: If you are unable to purchase the Armchair Analysis data, [Ben Dilday](htt
 - pandas
 - scikit-learn
 
+## Node.js package requirements
+
+- underscore
+
 ## Usage
 
 Unzip the play-by-play data into a directory. Run the following code from the directory

bot.py

@@ -8,6 +8,12 @@
 
 import winprob as wp
 
+try:
+    raw_input
+except NameError:
+    raw_input = input # python3
+
+import sys
 
 def load_data():
     click.echo('Loading data and setting up model.')
@@ -27,9 +33,14 @@ def load_data():
     return data, model
 
 def fg_make_prob(situation):
-    args = ' '.join("--%s=%r" % (key,val) for (key,val) in situation.iteritems())
+    if sys.version_info[0] >= 3:
+        args = ' '.join("--%s=%r" % (key,val) for (key,val) in situation.items())        
+    else:
+        args = ' '.join("--%s=%r" % (key,val) for (key,val) in situation.iteritems())
     model_fg = muterun_js('model-fg/model-fg.js', args)
-    return model_fg.stdout.split()[-1]
+    stdoutResults = model_fg.stdout
+    stdoutSplit = stdoutResults.split()
+    return stdoutSplit[-1]
 
 @click.command()
 def run_bot():

data_prep.py

@@ -1,7 +1,7 @@
 from __future__ import division, print_function
 
 import os
-
+import sys
 import click
 import numpy as np
 import pandas as pd
@@ -68,8 +68,9 @@ def load_pbp(pbp_data_fname, games, remove_knees=False):
     pbp = pbp[pbp.qtr <= 4]
 
     # pid 183134 should have a value of 0 for min, but has "0:00"
-    pbp['min'] = pbp['min'].replace({'0:00': 0})
-    pbp['min'] = pbp['min'].astype(np.int64)
+    if sys.version_info[0] == 2:
+        pbp['min'] = pbp['min'].replace({'0:00': 0})
+        pbp['min'] = pbp['min'].astype(np.int64)
 
     # Restrict to regular season games after 2000
     pbp = pbp[pbp.gid.isin(games.index)]

winprob.py

@@ -4,6 +4,8 @@
 import random
 import sys
 
+from sklearn.utils.validation import NotFittedError
+
 from collections import OrderedDict
 
 import plays as p
@@ -126,7 +128,12 @@ def generate_win_probabilities(situation, scenarios, model, data, **kwargs):
     # Note there is more information in situation than just model features.
 
     feature_vec = [val for key, val in situation.items() if key in features]
-    feature_vec = data['scaler'].transform(feature_vec)
+    try:
+        feature_vec = data['scaler'].transform(feature_vec)
+    except NotFittedError:
+        raise Exception("Sklearn reports that the instance is not yet fitted. " + 
+                        "This usually means that the version of python used to train " +
+                        "the model is different from the version you are currently running.")
 
     probs['pre_play_wp'] = model.predict_proba(feature_vec)[0][1]