Give it to me straight, doc. How many games have we gotten right recently?
| % of winners picked correctly | |
| Vegas | BTG |
|---|---|
| 67.56% | 62.61% |
| Last 50 Games (BTG) | 60% |
| Last 100 Games (BTG) | 63% |
| Last 200 Games (BTG) | 66% |
Model Methodology
When it comes down to the basic methodology, logistic regression is used to check whether the home team is expected to win an upcoming game. The data input is previous performance for both the home and away teams. This type of model outputs a score between 0-1 (ie 0% to 100%) for each match, representing the probability of the event happening (in this case whether the home team will win). If the output for a match is equal to or greater than 50%, that can be interpreted as “the model thinks the home team will win”. If it’s below 50%, the model thinks the away team will win the match. There’s a fully automated data extraction, cleaning, and formatting (ETL) pipeline running every night behind the scenes feeding into the model to make sure all past games are included and accounted for.
Every day during the NBA season BetterThanGuessing.com will update to show the predicted outcome for upcoming games.