How to build a Sports Betting Model
The most common question I get asked from followers (and even friends) is 'how do you make a sports betting model?' This question isn't easy to answer, because the starting point for each individual varies. Some need to learn statistics first and others just need a few pointers. That said, I'm going to give a high-level overview of where I believe you should focus. I'm not sure how Vegas Dave makes his sports picks, or how any other prominent handicapper, but for me, it all comes down to the analytics below:
1) Learn the basics
There are no shortcuts. The shortest 'cut' would be to purchase my betting models and watch my videos, however for those that are on a budget (I hear ya!), don't expect to magically wake up with the ability to build a sports betting model. First, you should do the following:
- Learn basic mathematics and how statistics work. You can do this for free or via online courses (masterclass, udemy, LinkedIn Learning etc.). These are typically $10-$15
- Hone in on ONE sport of interest and dive deep. Learn the metrics that other statisticians use
- Spend time mastering excel. Buy a course from Udemy and learn basic excel formulas (web-scripting, index match, vlookup, regression, etc.).
I work in data analytics and have been studying this material for over 10+ years (outside of sports). It's a process, but building a model does not require years of experience. Spend time on those above areas, acclimate yourself to statistics and sports and you'll be fine.
2) Collect Relevant Data
Figure out what data and statistics you find most impactful to the sport of your choosing. For example, in basketball I have found that points and how a team performs home and away to be the most predictive statistics, and therefore I assign the most weight to them.
For baseball, I use saber-metrics instead of common metrics. Saber-metrics are fantastic and the gold standard used by GMs and statisticians, and better yet, they are calculated for you. The best websites for Saber-metrics are FanGraphs and Baseball Prospectus
3) Projected Scores and Outcomes
I've found that it's best to project scores and compare/contrast those projections against Vegas. For baseball, I take this a step further and run a regression analysis to determine win %. You can then convert your winning percentage into odds (-110, -120, etc.) and use Kelly Criterion to determine how much to bet. For more on this, check out this article by Pinnacle here.
So... how do you project scores?
This task isn't as daunting as it seems. The basis for projecting scores is assigning a rating, and then compare how these teams perform against that rating. Let me give a quick and easy example of using just one metric in a model (points).
Say you have four teams in the league, as shown below:
Team A: 60 Points/game (offense), 140 points/game (defense)
Team B: 80 Points/game (offense), 80 points/game (defense)
Team C: 120 Points/game (offense), 100 points/game (defense)
Team D: 140 Points/game (offense), 60 points/game (defense)
From this data you can conclude that the average points per game is 400. If team A were to play team D, the average person might think the score would be 60 to 140, however this couldn't be further from the end result. Kenpom (the most widely respected NCAAB statistician) discusses this in detail on his site, but the outcome of this game would be a much larger blowout than 60 to 140.
In this fictional league with four teams, we can assume that Team D is playing defenses that average 100 points/game, yet scoring 140. Similarly we can see that Team A is only scoring 60 points a game, facing an average defense that allows 100. In layman's terms: Team A is consistently facing worse teams than Team D, yet scoring less, and Team D is consistently facing better teams than Team A and scoring more.
What would the score be?
Using just points as a predictor, I'd project this score to be 20 to 180 with a true line of Team D -160 points. This was calculated by simply comparing each team to the league average of 100, and determining how much stronger they were than the league, and then comparing those figures (negative for A, positive for D) against each other.
I hope this article helps you build your own sports betting model. With a little effort, study, and experimentation you'll be there in no time. I am strong believer that all picks made without a sports betting model, or without projections, are just luck. Bookmakers like when people 'think' they have systems, but hate when they actually have systems. Systems need to be updated regularly, as bookmakers are getting more intuitive each and every year.
If you're interested in purchasing my sports analytics and betting models you can do so here.