Artificial Intelligence is a broad term that applies to any situation in which a machine or computer is built to make decisions and/or react based on information presented to it.

Machine Learning is one type of Artificial Intelligence.

Machine learning is when software is provided data and through statistical modeling and analysis teaches itself varying patterns, calculations, and groupings in order to further understand and evaluate the data it was given.

This allows large amounts of data (thousands or even millions of data points) to be evaluated, processed and learnt from by just one statistical model.

We use Machine Learning to build some of our statistical models. These processes allow for complex statistical analyses to analyze large amounts of data in short periods of time and spit out precise predictions.

For example, our NBA model is given game by game statistics on advanced shooting metrics, turnovers, home court, assists, blocks and much more to produce predicted outcomes for each team each game. The models first evaluate large amounts of historical data to ascertain which statistics lead to important outputs, such as scoring and points allowed. Further, the models will not only evaluate which data points are important to predict these outcomes, but to what extent do individual statistics help predict outcomes.

After using past data to learn patterns, it will apply this knowledge to future predictions.

When one looks to analyze any sports game there is an over abundance of information available to pull from.

For example, let's say one is analyzing a basketball game. If a good three point shooting team plays a good three point defense who has the advantage?

Maybe one team is good at offensive rebounding and the other defensive rebounding? One may notice a team gets to the free throw line often and shoots well in the paint, while the other has a healthy assist to turnover ratio and shoots threes well. How do we balance which of these is more likely to lead to a win? Trying to thoroughly analyze all of these data points and synthesize them into a win loss prediction can prove fairly difficult.

Algorithms however, help fix this problem. An algorithm or statistical model will look at and analyze all information it is given, even up to millions of lines in a data set.

So while a human analyzing a game maybe can look at a few dozen statistics and find a few important patterns, a computer can look at millions of statistics and use sophisticated math to locate a multitude of patterns.

Additionally, the statistical models take away our human bias. Maybe we have a bias for our hometown team, or an outdated philosophy on the importance of mid range shooting. It is possible we might overrate a formerly great player or care too much about what happened in the last game. The statistical models will only base their predictions on the information given to them, factual statistics, while humans will have favorite players, teams, hated rivals, among a variety of other biases.