Research Question

[million2one]

I am building a model that will predict the final score of an NBA game.
Would it be more prudent to look a teams overall stats and then adjust the final number to correct for home court? Or should I just look at a teams stats when they are at home vs. when they are on the road?

What about a teams stats during their last 5 games, how relevant is this?
Is it okay to average the stats from the last 5 games with the stats for a teams performance at home vs. on the road?

Also any advise about how to determine the total number of FGA for a given team.
Right now I am taking a teams avg. number of possessions and subtracting their given TO rate per avg. number of possessions for the given team and using this number as the total number of FGA, am I missing something?

Sorry if this is confusing, I guess I have lots of questions, and I know this is the place to turn.

Any and all help is appreciated.

[tto827]

http://www.basketball-reference.com/about/glossary.html

This should help a bit.

How you vary last 5 games in with the rest of the schedule, and home vs. road weight in your formula are some of the differences that make it successful vs. a failure. There is no correct answer, try out different weighting, and backtesting should give you your answer.

[million2one]

Thanks tto,

I figured as much and is what I am struggling with right now.
Guess it will be a lot of trial and error until I can find the right scale factor for each.

When backtesting should I compare my predicted score with the final score of the game or the game line?

One other question:
When comparing a random variable, say a FG% for each team, is it safe to average team A's FG% with team B's FG% allowed, or do you think these need to be weighted as well?
I am thinking that what the team can do is more important then how the other team defends.
i.e. Team A's FG% should be weighted more heavily then team B's FG% allowed.