For the past few seasons, I’ve been manually tracking Game Projections from different models. Now, I want to wholly automate the tracking process. Let me know if you have a projection that you want included in the tracking process.
You can contact me on Twitter and via email: lars.sunesen.skytte@gmail.com
To be part of the tracking project you will have to upload a CSV-file, a Google Sheets document or something similar with date, team and win probability.
Specific Game Projections:
I will publish the model data in two categories:
- Specific Game Projections for each game.
- Model Evaluations.
For each day you will be able to see Game projections from the models.
Let me know if you don’t want to have the specific game projections published – Then I will remove the model from the Game Projections Tab.
If projections are uploaded early, it will allow the user to see the model projections before game start. You can upload early game projections and then afterwards change the probabilities to game time projections. I trust that everyone only uploads projections based on pre-game information.
Model Evaluations:
The other category I will publish is Model Evaluations. Here I will calculate log loss and betting results (based on calculations using the Kelly Criterion). This will allow the user to get information about the model’s performance.
Daily Routine:
- I will refresh betting lines (odds) and model projections multiple times during the day.
- Around an hour before the first game starts I will my projections on Twitter. This is when betting lines and my projections are locked.
- Model evaluations will be posted daily on Twitter.
- Game projections can always be found via the visualization.
Hockey-Statistics 