Skip to content

Analyzing the satistics & results of the Canadian Premier League - Dashboards and Stat Tables - MNB - Random Forest

Notifications You must be signed in to change notification settings

tamccullough/nutmeg

Repository files navigation

CANPL

CANPL Heroku App LINK

WOW. That's Different

It may not seem like much to you. But the main.py file and the accompanying cpl_main.py file have both gone through some incredible clean ups. Why? Because I've gotten better. You learn and you progress, and you should always go back and clean up bad code.

Canpl

Nutmeg is a CanPL Fan Site designed to provide stat tables and charts based on user input

Currently using MultinomialNB(), RandomForestRegressor() and RandomForestClassifier() to make the match predictions between teams as games are listed in the schedule.

How Match Prediction is Deployed

This app is using the following models for predictions;

After a lot of retooling. It has all changed.

Likehood table is created before a match with the results being fed into

  • MultinomialNB()

Using the predictions from that, those results are then fed into

  • RandomForestRegressor() to generate a score prediction using a 14 men roster and other features

To which those predictions are fed into

  • RandomForestClassifier() to generate a match prediction using both 14 men rosters and other features

The final output predicts the home teams probability and result label

Comparison Results of Classifiers

results knn adaboost bagged random forest voting classifier
count 10 tests 10 tests 10 tests 10 tests 10 tests
mean 0.344000 0.404000 0.430000 0.448000 0.440000
std 0.074117 0.100133 0.062004 0.070048 0.077172
min 0.240000 0.260000 0.300000 0.360000 0.300000
25% 0.275000 0.345000 0.405000 0.380000 0.395000
50% 0.350000 0.400000 0.450000 0.470000 0.460000
75% 0.395000 0.470000 0.475000 0.515000 0.480000
max 0.460000 0.580000 0.500000 0.520000 0.540000

Comparison Results of Regressors

results voting regressor decision tree random forest
count 10 tests 10 tests 10 tests
mean 0.334000 0.278000 0.310000
std 0.088969 0.064944 0.041366
min 0.180000 0.200000 0.260000
25% 0.280000 0.240000 0.280000
50% 0.350000 0.250000 0.300000
75% 0.390000 0.340000 0.355000
max 0.460000 0.380000 0.360000

The League

canpl.ca

Wikipedia Article

The Beautiful Game in Canada

I hope this becomes a great resource for the fans of the CanPL. I hope to incorporate Machine Learning and provide a great resource for Canadian fans of the beautiful game.

About

Analyzing the satistics & results of the Canadian Premier League - Dashboards and Stat Tables - MNB - Random Forest

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published