Summer in the NBA means preparation time. As soon as we turn the calendar over, free agency and summer league will start. It’s also a good time to prepare for the next season of NBA betting and if you are looking to set up your own mathematical model or simply to rank teams by their strength, you need to know what to look at. Various websites provided all sorts of different statistics, basic or advanced, so it’s not all that easy to separate important from the noise. Add the individual player performances, fantasy box scores and you have yourself a chaos.
So, what should you look at? Can some of those stats be completely ignored and if so which ones? Some statistics must be more important than the others, but how to grade them?! Once again, you can have it all served up here, but let’s not waste time and go straight to the heart of matter.
Determining important stats
Those who claim that statistics hides as much as it reveals have never actually learned how to use it. Still, not everyone can be a wizard with the numbers and fortunately, this is not essential to be a successful punter. Yes, the NBA stats databases are used and yes, there’s this thing called entropy that can help us find out how much of a correlation does a certain statistical category has with winning. And guess what, there are fewer stats with strong correlation than you can imagine.
The most obvious one is the point differential. It’s a remainder from a simple subtraction of how many points can a NBA team score and how many points they allow. Advanced technique with this statistical category is that the point differential is adjusted for the strength of the opposition, so that winning by 10 against the Warriors and the Suns is weighed differently. It is by far the most powerful NBA team stat for the against the spread betting, but obviously doesn’t tell much about the expected total points.
Other team stats with high correlation are the field goal percentage and the three point attempts. Some of the essential basketball statistics have very little impact of the actual outcome of the game – like the rebounding. The statistics naysayers are instantly going to jump in here and use it as a proof that once again, the math is failing to see the obvious, so I’d better explain it.
The rebounds are strongly related to the shooting percentage. If one team shoots 55 percent and the other one is grasping for air with 35 percent, the rebounding stats will show that the better-shooting team is actually winning the battle on the glass, no matter how poor of a rebounding this team might be.
There are improved stats that adjust for the availability of the rebounds, but those can find a better use with judging how good of a rebounder a single player is. Not all rebounds are of an equal difficulty either – take away Westbrook’s rebounds after the missed free throws and his statistics will take a serious dive. Assists are also not important, although they have slightly better correlation with the final outcome.
The projected totals also use points, scored and allowed, but you have to be more careful with it. Simple addition may lead you to the wrong side of a bet, as points are much more influenced by the type of the game teams have played, especially recently. The tempo of the game can determine the total points more than field goals percentage and some sources calculate it for you.
That’s the so-called pace, which practically represents how many possessions the teams had. To convert it to the basic team stats, you should look at the only ways a possession can end – it’s either a field goal attempt, a turnover or foul shots. Adding it together (you can use 2 free throws per trip to the charity stripe as the and-ones are now almost as frequent as three freebies) and you’ll get the number of possessions per team.
Besides the pace adjustment, it’s also important to monitor the number of overtime games as unusually large number of extra periods can skew the total points data. The solution is to observe larger samples of games as well as the current trend. Most of the websites will track the recent streaks – don’t let that misguide you, always track at least 50 last games as well. If a team didn’t changed much, the long term numbers should look to balance out and recent trend will fade over time.
Avoiding pitfalls
Before you even decide what to do with the stats mentioned above, it’s crucial that you understand that no matter how accurate, mathematical models are merely a tool to help you in your decision process. The bookmakers will have the upper hand on you with their PhD statistician crews and almost unlimited sources. If your math shows a too big of a discrepancy with the actual offer, it’s wrong. That’s the main reason you should deal with it during the summer, when there’s no everyday NBA games rhythm and you can have enough time to test things out.
Let’s take the most basic example. If the Spurs are winning by 8 points on average and the Heat are losing by 1 point, the Spurs should ideally be a 9 points jolly on the neutral court, or about 11.5 points at home. Think of this as a base number that would be further adjusted to account for many other things.
For the purpose of this practice, I’ll just keep in there. What does 11.5 points mean? It means that the most of the games played under these circumstances are expected to end close to that spread, with both sides equally distributed. It’s rather obvious, right – but there’s a big pitfall hiding in further conclusion: if the Spurs aren’t really expected to win by exactly 11 or 12 points, 8 points is close enough – if the books are offering a -8 handicap on the Spurs, everything is normal. It is not. If the Spurs are an 8 points favorite, and you are on 11.5, or if the total line is 210 and you have a 194, something is seriously wrong with your number. Do not think that such a large values ever occur. Check or adjust your math.
So how to implement the NBA team stats in your calculations?
Always start with the simple relations and build on them. There are many different mathematical models and the best of them are the most expensive intellectual property you can imagine, reaching six figures with ease. You can’t even buy them, but it does show that it’s well worth investing a little more than just a couple of hours in building one for yourself.
My advice is to keep a good balance between the season long stats and the last 5-to-10 games. Use point differentials of both periods and adjust for the level of competition. Add small influence to the home/away points. This should be a good foundation for further implementation (or fine tuning if you want) with the field goal percentages and three point attempts for setting up the spread. Similarly, use points scored and adjust with the mentioned stats for the projection of the total points line.
Once you get the desired look – TEST. Testing is incredibly important. Even if you aren’t building a predictive model, always leave a good portion of available data unenclosed – this will serve as your testing sample and can be used to define the model’s precision. Use the closing lines at some respected books for comparison and measure the error on different sizes of sample. Try minimizing it if it’s too large.
Finally, allow the time for the live testing. Once the NBA season begins, use virtual betting to keep track on the model’s efficiency for at least couple of months before you use it in deciding which wagers to place. How much weight should you add to the numbers your model gives you? The answer to this question is too complex to write down in just a couple of rows, so I’ll leave it for some of my future articles, but it should not outweigh human knowledge and your understanding of the basketball game. After all, it can never be as complex as the human thinking. Not in foreseeable future at least.
