There is nothing in the life of a sports bettor that is as wild as these next 4 days. Live bullets are flying, Gus Johnson is screaming about vegetables and Greg from Sales is telling you about how he has Oregon beating Wisconsin as his big upset while you peer over his shoulder at the TV sweating out your Baylor/Syracuse 1h over that you placed because Baylor had a high-powered football offense 4 years ago.
The point is, for as much fun as these next few days are, it can also be extremely stressful. With a new game tipping off every couple minutes, there is a lot to take in, especially when you are wagering your hard-earned money on the games. I am here to help. With my background in Predictive Analytics, I put together a model that predicts, based on 10 years of historical NCAA data what the spread of each game ‘should be’. The idea is that this can be another source of information when you are deciding to place bets on the games and hopefully bring some objectivity into your decisions.
A FEW NOTES
- I took 10 years of historical data (2008-2018) from the NCAA Tourney first round – every team’s advanced metrics and what the spread was when the two teams played on another
- The model was then trained on this data to produce an equation that will predict what the spread ‘should’ be based on these advanced rating metrics (Column D)
- To put it even simpler:
- advanced metrics = independent variables
- spread = dependent variable
- For any other numbers guys out there the R Squared of the model was .9556. Meaning 95.56% of all variance in spreads can be accounted for in these metrics. What this model does is find the 4.5%
How to use:
- This is not meant to be verbatim picks. This is another source of information for you this weekend when placing your bets
- This is strictly based on the numbers, the only manual adjustment made was to the Kansas State game for Dean Wade (3 points)
- The rating system is:
- 0-1 point predicted vs. actual differential “Low”
- 1-2 point predicted vs. actual differential “Medium”
- 2-3 point predicted vs. actual differential “Strong”
- >3 point predicted vs. actual differential “Very Strong”
- The small amount of ‘Very Strong’ games is a good thing. It means the model works
- You can update Column C if the spreads change – everything else will populate
- I have personally bet all three ‘Very Strong’ games and Virginia Tech (VT is a ‘Strong” play before even factoring in the return of Justin Robinson)
Below is a link to download the model. If the spread changes, you can change that number in Column ‘C’. I will try to send out an updated version after the play-in games occur tonight (3/20) or tomorrow (3/21) prior to tip.
Follow me on Twitter if you want to see the updates: @IPlayedD1
There you have it, from my model, I will be betting on:
Va Tech -10.5
St. Mary’s +5
Abilene Christian +22.5
Good luck out there folks!