The CapWize Simulation Model
The philosophy behind our model is to acknowledge that if two teams were to play many different times, there would be many different outcomes, even if all of the conditions of the game are the same. So, instead of directly modeling a point spread or over/under, we have developed a model that allows us to take the conditions of a game and produce realistic simulations of the outcome. Doing this enough times paints a clear picture of how likely certain game outcomes are, which helps inform us as to where the good value lies for placing bets.
This is the most effective way to attempt to find inefficiencies in the lines and to make intelligent sports picks vs relying heavily on ATS and O/U historical trends which are great at showing correlation but not the best at predicting the outcomes of future games.
For example, let’s take the NFL 2020 opener, KC -9.5, and an over/under of 54.5. Using this information, as well as the past performances of these teams, we can use the CapWize NFL Model to simulate this game as many times as we want. How about…a million times!
Across the million simulations, the average (median) score we saw was 31-19 in favor of the Chiefs. The model has the Chiefs covering the spread in 57.8% of simulations and the game going under in 61.9% of simulations.
An important thing to note is that this does NOT mean that the Chiefs actually have a 57.8% chance of covering the spread or that there is a 61.9% chance of the game going under. In backtesting, bets in these ranges hit at 52%, which is right about break-even when you take the vig into account. That being said, we have identified many model outcomes that are historically both profitable and abundant.
So what can you expect to see going forward? The subjective human element will always play a huge role in sports betting. Our model seeks to augment this with a purely objective perspective as well. We’re still figuring out the best way to synthesize all of this information.