Understanding Variability in Sports Models

G’day, lads! Merry Christmas!

A common question I often get from those new to the community is: Why do the projected probabilities for different markets change, even when there’s no change in a selection, like a team adjustment?

The simple answer is that projected probabilities change because of the nature of simulations. In sports modelling, there isn’t a single formula to calculate a team’s exact chances of beating another. Sports are complex, with numerous factors influencing outcomes, such as player performance, strategies, and luck. To account for this, simulations are used to replicate the unpredictability of a single game. Each simulation models a possible outcome, and this process is repeated many times to build a distribution of results.

For instance, in the ‘Excel LADZ - NBA Team-Based Model’, I projected the upcoming Denver vs Cleveland game. Out of 2500 simulations, Cleveland won 1,507 against the Nuggets, giving them an expected Win % of 60.28%. Meanwhile, the Nuggets won the remaining 993 games, resulting in an expected Win % of 39.72%.

If I refresh the worksheet to run another batch of 2,500 simulations, the probabilities can change slightly. For example, Cleveland’s expected Win % might now be 60.48%, slightly higher than the previous result of 60.28%.

This variation occurs because simulations rely on randomness to replicate the unpredictability of real-world outcomes. Over a larger number of simulations, these fluctuations average out, converging on the true expected value. However, smaller sample sizes - like 2,500 simulations - are still subject to the effects of statistical variation and random luck.

You might wonder why I built this model to run only 2,500 simulations rather than a million. The reason is practicality. While more simulations would slightly reduce variability, they require a lot of computing power, which can make Excel slow and frustrating to use. A model that takes over 10 minutes to load for each game or lags constantly isn’t practical, no matter how precise it is. By running 2,500 simulations, the model strikes a balance between accuracy and usability, providing reliable probabilities without sacrificing performance.

Thanks for reading, lads! If you’d like to download the NBA Model seen in this article, click the link below.

Download ‘NBA Team-Based Model’: https://www.patreon.com/posts/excel-ladz-nba-116613408