Around 10 months ago, I released a series of Instagram posts where I attempted to predict NBA team records. I did my predicitons on 14 teams who I thought were interesting. Those predictions were based on simply adding up Win Shares from the previous season. When I first created those series of posts my objective was to just see how accurate these predictions were. I knew that Win Shares were a great stat to measure a player’s impact on a team in previous seasons- this is because if one adds the total amount of player Win Shares on a team, the sum is the amount of wins a team had in that season. As a result, I decided to predict this season’s wins by adding each team’s total Win-Shares. Below are my results:
2019-20 W% Predictions Based On WS:
Looking at those results, I was quite excited with my results because they were near great. To verify that I did a T-test on the diffrence between the actual records and my predicted records which gave me a P-value of 0.6 (meaning my predictions are valid), in addition to a one-mean-95%-confidence interval. The confidence interval indicated that the predictions had an estimated error of 2% which translates to an error of 1.5 games across the season and there was a margin of error of arround 7% which translates to a 6 game margin of error of my predictions.
WS Based W% Prediction Errors:
Delighted by the results of the WS based predictions, I thought I had a chance at having my estimates beat Caesar’s Over and Unders. Little did naive me know, Caesar’s Over and Unders are set near perfectly. I calculated the diffrence between Caesar’s predicted W% and the actual W%. The results were literally perfect. Caesar had a median error of 0.5% which translates to around 0.41 games, contrary to my 2% error which makes up 1.5 games. Nonetheless, given those results, I thought it would be worth “simulating” what would have happened had I bet.
Like the last TRILLION times I tried to outsmart Vegas, I ended up right where I started. In this situation I tried to “bet” $700 and I ended up with a net profit of -$50. I am still determined to come up with a formula that can beat Vegas odd-setters (I will be applying one of those formulas once the NBA returns and I am optimistic about that formula) ; however, I need to respect my genius opponents who are the best in the game which is why they set the odds so well.
HoopsData 