Get help with your bracket pool

Settle a debate with your friends

Learn about predictive analytics

How do I Interpret Predicted Winner and Probability?

The predicted winner is the team who our model believes has a greater than 50% chance of winning in the given match.

It’s important to note here that we can never be 100% sure when we’re trying to predict the future. The model is based upon a variety of historical factors, and you’ll find it to be fairly accurate, especially when the model probability was confident. We would define confident as having a win probability greater than 90 percent.

When the win probability is nearer the 50% line, it’s going to be a good game and has high upset potential. The model is predicting that these teams are actually fairly evenly matched. When factors are fairly equal going into a match, the unpredictable human nature of sports increases in importance and comes into play.

What is the Evolytics College Basketball Game Predictor?

Put simply, the College Basketball Game Predictor is a tool that uses predictive analytics to forecast which team would win in a matchup.

This collaborative, cross-functional project was created by a team of developers, data engineers, statisticians, and sports fanatics. We created four machine learning models and trained them with past regular season and tournament data in order to help others with their 2019 bracket selections.

If you really dig data science, or want to see which statistical model most agrees with your gametime intuition, you can compare all four of our models above game for game.

Why the Elo Rating System?

Originally created by Physics Professor Arpad Elo to predict the likely outcome of chess games, the Elo rating system has evolved to calculate the relative skill of competitors, therefore forecasting a likely winner, in all sorts of zero sum games.

We developed our own mathematical model and continually worked to increase the accuracy of the basic Elo foundation. We would like to credit Nate Silver of  FiveThirtyEight, whose NBA Elo calculations were an inspiration for this particular model.

What Went into the Other Predictive Models?

The Logistic Regression, Gradient Boost, and Multilayer Perceptron rely on the historical win percentage, turnover average, effective field goal percentage, free throw rate, and offensive efficiency for each team.


Select two teams in the fields above and find out the likelihood one team would win in a matchup on a neutral court.

Everyone has a different strategy when filling out their tournament bracket. Some people just pick their alma mater, while others make their picks according to their favorite mascots or color schemes. Evolytics relies on data.

While nothing in March is guaranteed, the win probability figures presented by our predictor will give you a good idea of the likely outcome of each game. Evolytics lets you take the hassle out of picking that perfect 5/12 upset by trusting the numbers.

Yes. Predictive analytics will help you get a leg up on your office bracket pool by picking the likeliest winner in each tournament matchup.

Pair teams against their potential matchups throughout the tournament and select your cinderella picks before the first game tips off.

The odds of a perfect bracket are long. Really long.1-in-9,223,372,036,854,775,808-long. That's 9.2 quintillion.

While our tool won't guarantee a perfect bracket, it can certainly improve your odds of winning your bracket pool. All you need to do is pick a team, match them up against all possible opponents during the tournament, and make more informed decisions on your bracket. After that, it’s all up to the basketball gods.

Not at this time. Individual player stats and related items are in the queue for future iterations.

Drop us a line.