A Bayesian Approach to Predict the Number of Goals in Hockey

Round 1

Reviewer 1 Report

(1) Please go through the paper to fix typos and some bad sentences.  For example, page 2, "in the from of " --> "in the form of " etc.

(2) A simulation section should be added to investigate the proposed method.

(3) Section 4, page 3, the authors make two different assumptions. Please justify it. 

Author Response

Dear reviewer

Thank you so much for your useful suggestions. I applied your comments in the manuscript in Green color.

  Here are my responses:

1-Done.

2-I added a subsection (sec. 5.2 Simulation study)

3-This paper proposed the posterior predictive density estimation for two well-known cont models (Po distribution (assumption I) and COM-P distribution (assumption II)). We applied the proposed methods

to the hockey dataset. Sections 5.1 and 5.2 (recently added) evaluate the posterior density estimators under these two assumptions.

Thank you.

Reviewer 2 Report

The authors develop a bayesian prediction model for the hockey scores, based on the poisson and com-p distributions. The performance of the model is evaluated in the cases of two hockey teams.

The major issues are:

1.       Evaluations are based on a small sample size. Given the abundant information available on the subject matter, it would be more convincing if the authors can conduct a larger scale evaluation and compare their methods with others.

2.       Using of expert opinions in prediction could be arbitrary. It would be interesting to know how the model preformed without  using expert opinions.

Author Response

Dear Reviewer,

Thank you for your useful comments. Here are my responses to your comments:

 1- Here we applied Bayesian predictive modeling to predict the upcoming outcome results team A (here Edmonton Oilers) vs team B (Here Arizona Coyotes). We are working on different project, to predict the outcome, using (multivariate) Poisson regression, in that project we have considered many covariates to predict the game outcome (a large sample size along with many regressors). But at this manuscript, we are using past records which Oilers played home versus Coyotes (Away)

and it is very limited since each team play on average 2-3 times (specified to home and away) each season.

 2- I mentioned in the paper if we need to predict just based on data without considering the specialists' opinion we need to take d=0 (page 4). But that is right. It is more interesting to see the results which I added to manuscripts (in orange color)

Thank you.

Round 2

Reviewer 1 Report

The revision is OK.