Predicting the outcome of NBA playoffs using the

Naïve Bayes Algorithms

Na Wei
Department of Biomedical Engineering, College of Engineering, University of South Florida,
Tampa, FL 33620, USA
Abstract
In this paper we examine the use of naïve Bayes model as a tool for predicting the success of
basketball teams in the National Basketball Association (NBA). Statistics for 2005-2011 NBA
games were collected and used to examine and predict the outcome of NBA playoffs. The home-
road effect in the results of NBA playoffs is also considered, and compared with the
unconsidering. Results obtained demonstrate that it is effective on predicting the outcome of
NBA playoffs using naïve Bayes algorithm. The home-road effect does not have significant
influence on the outcome and increasing the accuracy of prediction. Further, according to the
prediction, we infer that the final championship of 2010-2011 NBA playoffs is Chicago Bulls.
Keywords: naïve Bayes algorithm, NBA playoffs, home-road effect, prediction

1. Introduction

Since the winter of 1891, following the day Dr. James Naismith nailed a peach basket into a gym
wall, basketball has evolved into a true American game (NMHOF, 2008). Nearly 270,000 people,
each game day, attend basketball arenas around the country to watch the best of the best sweat,
hustle, and entertain (Ibisworld, 2008). Along with watching the games, millions of fans are
involved with the ever growing arena of fantasy basketball leagues and other gambling
alternatives. Involvement in these leagues and gambling precipitates the desire to know which
team will win before they even participate in a game. And use algebra to predict the result just
offer much fun to people, also supply some benefits to the assuming champion team because
people will take more attention on the strong team, thus the focused team will receive more
income and fame.

According to the rules of the NBA, we know that each team will play at most 7 games with his
opponent in the every round of the playoffs, and the team which wins 4 games firstly will stay
and go on, on the other side, the team which loses 4 games firstly will lost the chance to go to
next round and go home. So in these 7 games, one team which has more winning rate and higher
ranks in the regular season will have 4 games in home and 3 away in the playoffs. Thus it is
always acknowledged that the team which obtains 4 home advantages will be more likely to win
and go to next round. In this paper, we examine the use of naïve Bayes model as tool for
predicting the success of basketball teams in the National Basketball Association (NBA). We
will also investigate whether the home advantages take effect in playoffs and whether it will
promote the assuming accuracy with considering this factor. Finally, we will predict the final
champion of 2010-2011 NBA season.

2. Materials and methods

2.1 Data

We obtained our data set form three sources: NBA ESPN, NBA WIKIPEDIA and HOOPS
STATS. We collect the data covering five seasons from 2005-2006, 2006-2007, 2007-2008,
2008-2009, 2009-2010 and 2010-2011. Figure 1 displays the typical box score downloaded.

Fig 1 Layout of typical box score from 2009-2010

Information on the team totals in the game, as well as their home/road situation, were the only
features used to conduct the Naïve Bayes analysis. For each season we transfer the information
into the form like below which is preferred to our model:
 Table 1 The total won rate of east top 8 teams in regular season 2009-2010

Eastern team Total winning rate Home winning rate Road winning rate
1.Clevealand 0.744 0.853 0.634
2.Orlando 0.720 0.829 0.609
3.Atlanta 0.646 0.829 0.463
4.Boston 0.610 0.585 0.609
5.Miami 0.573 0.585 0.56
6.MIlwaukee 0.561 0.682 0.439
7Charlotte 0.537 0.756 0.317
8Chicago 0.5 0.585 0.414

Table 2 The record of east top 8 playing with each other in regular season 2009-2010

road Clevelan Orland Atlant Bosto Miam Milwauke Charlott Chicag

home d o a n i e e o

Cleveland 1-1 2-0 1-1 1-0 2-0 1-1 1-1

Orlando 1-1 2-0 1-1 1-1 2-0 1-1 1-0
Atlanta 1-1 1-1 2-0 1-1 2-0 2-0 2-0
Boston 1-1 2-0 0-2 1-0 1-1 2-0 1-1
Miami 2-0 1-1 2-0 0-2 0-2 1-1 2-0
Milwaukee 1-1 0-1 1-1 1-1 1-1 2-0 2-0
Charlotte 2-0 0-2 2-0 0-1 2-0 2-0 1-1
Chicago 1-1 1-1 1-1 1-1 0-1 1-1 2-0

The database used in this research from NBA games played in the other seasons presented in the
database appendix.

2.2 Naïve Bayes model

We use Naïve Bayes model algorithm to calculate the probability that team i will beat its
opponent team j when they meet in the playoffs routine according to their regular season record.
Through this way, we predict the results which team will stay and go to next round even to the
final champion. For reference the equation is as follow:
P (player i winning given i and j are playing) =

3. Result and Discussion

3.1 Without considering the home-road effect in the results of NBA playoffs

Without considering the home-road effect in the results of NBA playoffs, we calculate the
probability which team will beat his opponent when we know the schedule of games. Then we
predict which team will win and go to the next round. Finally compare the prediction of each
round between two teams with the real results to indicate the accuracy. The result of 2005-2006
playoffs of eastern term as an example is given in Table 3 and Figure 2 and others are presented
in database appendix.

Table 3 Statistics and their Abbreviations for 2005-2006 playoffs

Eastern team Total winning rate Home winning rate Road winning rate
Detroit .78 .902 .659
Miami .634 .756 .512
New Jersey .598 .707 .488
Cleveland .610 .756 .452
Washington .512 .659 .366
Indiana .5 .659 .341
Chicago .5 .512 .488
Milwaukee .488 .61 .366
 Fig 2 The prediction of each round about 2005-2006 playoffs

Following this initial assessment, the accuracy of predicting the outcome of 2005-2006 eastern
team playoffs is 0.571; and the accuracy of predicting the outcome of 2005-2006 all team games
is 0.667 (shown in Table 4 below). The result is not very good, and the reason is possible that the
smaller the differences between the total winning rates of two teams is, the more the wrong
probability appears. So, this maybe needs to use other models and algorithms to increase the
overall accuracy of the prediction model. However, the model is very easy compare to others,
such as the Bradley-Terry-Lute model, the contest success function. And Results for the
accuracy analysis of 2005-2010 NBA playoffs are shown in Table 4.

Table 4 The accuracy of Predicting the outcome of NBA playoffs without considering the home-road effect

years right wrong accuracy

2009-2010 11 4 0.733
2008-2009 11 4 0.733
2007-2008 13 2 0.867
2006-2007 11 4 0.733
2005-2006 10 5 0.667

From the analysis of the accuracy of predicting the outcome of 2005-2010 NBA playoffs, it
demonstrates that the predictions are effective by Naïve Bayes model algorithm. Mostly, the
right probability is 0.733; of course there are also some special things (better or worse). And the
average accuracy is 0.747.
3.2 Account the effect of home-road on the results of NBA playoffs

We account the effect of home-road effect on the winning rate to measure the accuracy to
demonstrate whether it will decrease the error rate of prediction. And Results for the accuracy
analysis are shown below in Table 5.

Table 5 The accuracy of predicting the outcome of NBA playoffs without considering the home-road effect

years right wrong accuracy

2009-2010 11 4 0.733
2008-2009 10 5 0.667
2007-2008 13 2 0.867
2006-2007 10 5 0.667
2005-2006 11 4 0.733

It is shown that the average accuracy of predicting the outcome of 2005-2010 NBA playoffs
accounting the effect of home-road is 0.733, it is less than without considering that. The further
analysis will be shown in section 3.3.

3.3 Comparison of the accuracy of two methods

We compare the accuracy of two methods to analyze whether the home-road effects have
influences on the result of the prediction. The results are given in the figure 3.

considered unconsidered

1
0.9
0.8
0.7
accuracy

0.6
0.5
0.4
0.3
0.2
0.1
0
2009-2010
1 2008-2009
2 2007-2008
3 2006-2007
4 2005-2006
5
year

Fig 3 The comparison with the accuracy of two methods to predict the outcome of NBA playoffs
From the bar diagram above, we conclude that the home-road effect do not have enough
influence to promote the accuracy of prediction when using Naïve Bayes model, and sometime it
has negative effect on the accuracy, such as 2008-2009data, 2006-2007data. In another word, it
is not necessary to take account of the home-away effect as a feather in Naïve Bayes model to
predict the results of playoffs. Thus, we will predict the outcome of the latest data set from NBA
2010-2011 playoffs without considering the home-road effect in the following test.

3.4 Prediction

According to the latest data set-NBA 2010-2011 regular season record, we use Naïve Bayes
model to illustrate the playoffs process and predict which team will win the NBA final
championship without considering the home-road effect. Statistics are presented in Table 6 and
the prediction of 2010-2011 playoffs is given in Figure 4. Other raw data is presented in database
appendix.

Table 6 Statistics and their Abbreviations

Eastern team Total winning rate Home winning rate Road winning rate
Chicago 0.756 0.878 0.634
Miami 0.707 0.732 0.683
Boston 0.683 0.805 0.561
Orlando 0.634 0.707 0.561
Atlanta 0.537 0.585 0.488
New York 0.512 0.561 0.463
Philadelphia 0.500 0.634 0.366
Indian 0.451 0.585 0.317

western team Total winning rate Home winning rate Road winning rate
San Antonio 0.744 0.878 0.610
LA Lakers 0.695 0.732 0.659
Dallas 0.695 0.707 0.683
Oklahoma City 0.671 0.732 0.610
Denver 0.610 0.805 0.415
Portland 0.585 0.732 0.439
New Orleans 0.561 0.683 0.439
Memphis 0.561 0.732 0.390
 Fig 4 The prediction of each round about 2010-2011 NBA playoffs

So far NBA games have been carried out for all games except the finals. Compared to our
prediction of each round, we have successfully predicted 12 games and failed to 2 games. In fact,
we are not able to predict the game’s winner played between Dallas and LA Lakers, because the
total winning rates we calculated between them are equivalent. According to the prediction, we
infer that the final championship of 2010-2011 NBA playoffs is Chicago Bulls, and let time to
prove whether the final champion it is. Even if the answer is wrong, the right probability of
prediction has a high accuracy.

4. Conclusion

When we use Naïve Bayes model to predict the NBA results, we only need winning rate as
feather. On this condition, the algorithm is effective and we are able to predict which team has
more ability to go to next round in playoffs. However, it is not available to predict the result of
single game. In the further, what we have to do to increase the accuracy of prediction is using
Neural Networks as model, then construct and discover new features that can more accurately
capture the effects of real game situation to the winning/losing margin of games.

References

[1]Ralf Herbrich, Tom Minka, Thore Graepel. A Bayesian Skill Rating System, in Advances in
Neural Information Processing Systems 20, MIT Press, January 2007
[2]Bernard Loeffelholz, Earl Bdenar, Kenneth W. Bauer. Predicting NBA Games Using Neural
Networks, Journal of Quantitative Analysis in Sports, Vol. 5 [2009], Iss. 1, Art. 7
[3]Kenneth J. Koehler, Harold Ridpath. An Application of a Biased Version of the Bradley-
Terry-Luce Model to Professional Basketball Results. Journal of Mathematical Psychology 25,
187-205 (1982)
[4] Jia Hao. An Empirical Study of Contest Success Functions: Evidence from the NBA.
Department of Economics, University of California, Irvine, CA 92697-5100, USA. October 17,
2007