NBA All-Star Weekend, which was actually very entertaining (for a change), provided a much-needed respite for players. But now it’s back to business on the pro hardwood.

NBA teams are gearing up and making a final push for the playoffs (or the No. 1 draft pick). Likewise, bettors are trying to get in a few more winning bets before the season is over.

If you look at what worked in the first half of the season, then you might assume that San Antonio, Golden State and Orlando would be good bets, as they had the highest cover percentage in the first half of the season. But is that true?

Are teams that are good bets in the first half of the season also good bets in the second half of the season? If they aren’t, can we try to predict which teams will be good bets in the second half?

First, let us clarify what we mean by the first and second parts of the season. Ideally, we would like to look at games before and after the All-Star break, but as that date changes every year it is a little tough to code into the data, which is necessary as we will try to look at a lot of seasons.

Therefore, for simplicity, we will consider the first part of the season the first 52 games, and the second part all the games after that. As for what years we look at, for this question we looked at every season since 1999 (the lockout-shortened season was the year before) and we also threw out the 2011 season, the other lockout-shortened season.

Now, we can easily get a team’s winning percentage against the spread for the first part of the season and compare it to their winning percentage ATS in the second part. Plotting these two values, we get the chart below.

 
There doesn’t appear to be much relation, but there are also a ton of data points so it is hard to tell. To try to see whether there is some relation, we can do a regression on the data, trying to predict the winning percentage in the second half by using a team’s winning percentage in the first half.

Doing a regression, we find that as we expected, winning percentage ATS in the first part is not a significant predictor - the p-value associated with the variable was 0.107. Although this is nearly double the level usually required for statistical significance (0.05), it was close enough that we added two new terms into the model: winning percentage SU and average line for the first half of the season.

Once we controlled for these two variables, we found that our model was actually finding some statistically significant predictors. Here is the R output from that linear regression (R is just the statistical program that I used for our analysis):

The important thing here is the table of coefficients. The first row simply relates to the intercept of the model, which isn’t worth discussing. The second row relates to the effect of winning percentage ATS in the first part of the season. Two things are important here:

First, the estimate of the coefficient is positive, meaning that teams that win more games ATS in the first part of the season win a higher percentage of games ATS in the second part. Second, the p-value associated with it is 0.03, below 0.05, so it is statistically significant.

Looking at the other two variables, we can see that teams that win more games SU in the first part of the season win less ATS in the second half (this is nearly significant, at a 0.07 level). And finally, teams that have a higher spread (in this case, higher refers to the amount of points a team was giving, so if a team was favored by 10 points they would be -10) are less likely to win games ATS in the second part. So the ideal team to bet on would be a team that wins a lot ATS but loses SU and was projected to win their games by a large margin.

This ideal team is unlikely to exist but it leaves us asking, which NBA teams are the best bets by this model? We can very easily use the model to predict a team’s ATS win percentage in the second half of the season, as for all the teams we already know the relevant variables from the first half (win percentage ATS, win percentage SU, and average line).

Doing this, the top three teams to bet on, along with the percentage of games they are expected to win ATS are below.

Team	Expected Win Percentage ATS
San Antonio Spurs	53.89%
Golden State Warriors	52.26%
Boston Celtics	51.58%

And below are the bottom three:

Team	Expected Win Percentage ATS
Los Angeles Lakers	47.69%
Philadelphia 76ers	48.41%
Chicago Bulls	48.58%

Most of these aren’t a huge surprise, but two of them are rather interesting. For the top teams, the fact that San Antonio and Golden State are there shouldn’t be shocking, as both are clearly a notch above the competition and have been winning, by a large margin, against the spread this season.

Boston might be a little more shocking to see. They are fifth in win percentage ATS but at the same time haven’t won that many games. As Celtics fans (remember, we’re at Harvard), it makes us glad to see them on the list.

Meanwhile, in the bottom we have two teams that are just putrid (Lakers and 76ers). While they haven’t done awful ATS this year (as in they aren’t in the bottom five), they are both teams that you wouldn’t really to bet on.

Meanwhile, the Bulls are the third worst team against the spread but they have won more than half their games and are a playoff team in the East. Although it may tempting to bet on them because they are a playoff contender, it seems Chicago might not be the smartest bet.

One last thing to notice, although the predictors we found were statistically significant, the percentages we are assigning to each team aren’t exactly practically significant. If the Spurs win 53.8 percent of their games ATS, is that a high enough percentage to win money?

Once you account for vig, it will be close. And the Spurs are the team with the largest discrepancy from 50 percent. All of the other teams are less than 3 percent away from 50/50. This happens because we have a very large sample size - 446 observations in all - so we are able to detect small trends that aren’t obvious at first.

Be warned: although these teams are the sides expected to do the best by our model, and the variables used to predict this have been shown to be statistically significant, the actual value of this model may not be that large. The NBA still has a lot of basketball to play.

"); h = d.getElementById(tid); h.parentNode.insertBefore(s, h); })(document);

Top 10 NO deposit Bonus offers @

Read More... [Source: Covers.com: NBA News and Stories]