Thursday, September 22, 2016

21st Century NBA Basketball Prediction Program - Chance or Skill?

  After spending much of the summer in a fruitless attempt to get my chess game in shape for the tournaments in Duluth and Jackson I turned my attention to the upcoming professional basketball season and specifically my attempts to resuscitate my 30 year old basketball prediction program for the 21st century. I worked on this project last summer and thought I had a workable formula that predicted winners above the magical 52.4% break-even mark for the 2013-2014 and 2014-2015 seasons. After a promising start to the 2015-2016 season the formula stopped being successful and I finished the season making picks on my own and finished the season just short of the break-even line, losing $600 mythical dollars in the process.

  I started this year’s effort by creating a data table with different formulas combining the 16 variables I captured using the three seasons of NBA data I’ve assembled. I then created and ran a stored procedure to create the predictions for each season for each formula and save the results in my database. In addition to predicting winners, my software comes up with its own point spread which I compare to the published point spread. I summarize my results by how accurate my predictions are against the published line and how different my point spread is from the published spread. In a perfect world the bigger the difference between the two spreads the more correct my predictions will be.

  While running a query against my result summary I discovered that out of the hundreds of formulas, three gave me over 55% correct predictions on games with a 5 point difference between my spread and the published spread for each of the three seasons used. Better yet, when the point spread difference was 7 points or more the success rate jumped to over 60%. I have 6 weeks before the start of the 2016-17 season and decided to use the time to verify these new formulas. I loaded the 2012-13 scores into my database and am entering in the point spreads from I have 700 of the 1229 games loaded into the database and so far the results are very promising. Over the next week or so I’ll enter the rest of the season and find out for sure if my 3 ‘success stories’ are for real.

  It seems absurd on face value that any computer program could beat the odds makers. In reality it makes perfect sense. Point spreads are not set in order to predict the actual score. They are set in order to guarantee that there is an equal amount of money bet on either side to allow the gambling facilitators to make money (on the 5-10% extra paid by the losing bettors) no matter who wins the contest. Any false assumption by the betting public (like betting blindly on the Warriors or Cavaliers) or any team with a fan base that always bets on or against then will cause the point spreads to be skewed to attract an equality of wagers on both teams.

  Real or not, I will soon be loading the 2016-2017 schedule into my database, I resolved this season to get an online gambling account on and make my predictions on my NBA point spread predictions blog with real and not mythical money. By using real money on an established gaming site I hope to print my predictions on my blog the day after the games using screen prints as verification and offer my predictions the day of the games on a subscription basis.

  I plan to start my Bovada account with $500 and bet $10 on each game. I’m not expecting to go 45 games under .500 at any point in the season (which would exhaust my funds) but I’ll deal with that situation if it happens. The hardest part of this exercise will be to have the discipline to stick with my formula and not make picks every day just to be making picks. There are going to be plenty of days where my formula doesn’t differ from the published point spreads and those are the days that will test my patience.

  I hope I manage to make some actual money picking NBA games this season because the thought is occurring to me that wagering on the outcome of sporting events is going the way of the dinosaur and I want to be successful before the practice goes extinct. The new way of gambling on sporting events is not betting on the winners and losers but by playing ‘Fantasy Sports’ where the gambler picks a roster of players and roots for his ‘squad’ to outscore the competing squads by having better statistics. Fantasy sports is at least 40 years old, having gotten its start as ‘Rotisserie Baseball’ and graduated to the big time as Fantasy Football. Originally these activities were done by small groups but have gone global with the arrival of Internet sites that allow players from all over the world to play in the same ‘leagues’ for huge amounts of money.

  The biggest companies in the fantasy sports business are ‘’ and ‘’. Both companies advertise heavily on sports broadcasts and podcasts and like most of the successful drug dealers I’ve known will give new customers a free taste of the action.

  While sports betting is only legal in Nevada (Delaware has state sponsored sports parlay betting), playing fantasy sports for money is allowed in over 35 of the 50 states. New York famously banned online fantasy sports in 2015 as gambling but did a turnaround after passing regulations that included taxing the gaming sites to the tune of 4 million dollars a year.

  It’s amusing to see fantasy sports and betting on the outcomes of sport events decried as gambling by the same entities that run multi-million dollar lotteries that only pay out 50% of the take. Is sports betting gambling? For the unskilled, absolutely! But for the skilled it is not gambling at all. The Wall Street Journal reported that 91% of the prize money in the 2015 baseball fantasy season was won by 1.3% of the participants. I don’t know if the same ratio holds true for picking against the point spread but if it does I hope I’m a one percenter.