Need your thoughts and assistance

[headcoach]

How many games equal a good sample to use for data purposes.
For example

Using closing numbers on hoops:
If a home dog wins 78% of the time when the line is 9.5
How many games would be considered a good sample for data purposes?

Interested in your thoughts for those that get into the data end of it all.
Thanks in advance ?

 

[G.MONEY]

This is just my opinion, and thought that you need at least 300 games, why did i pick that number not sure...But the higher the better if you are looking for true data...

 

[exploited17]

Closing number wouldn't mean anything anyway. I could be bought up or down. If it was something you could look at that meant anything it would be opening number. Second half of the year just look for good home dogs. Good teams won't cover big numbers all the time and some teams with bad records still get huge home dog points even though they have improves significantly but are undervalued simply because of their overall record. So, they offer great value Look for that instead

 

[tsm123]

First, ....................

Don't take advice from anyone who has been handicapping less than thirty years
and is under the age of sixty. They haven't been schooled nearly enough.
Even those who meet the age and handicapping requirement will be quick to tell you,
"I'm not good enough to even offer my opinion."



Welcome to the world of eleven to ten.

 

[PoundFoolish]

How many games equal a good sample to use for data purposes.
For example

Using closing numbers on hoops:
If a home dog wins 78% of the time when the line is 9.5
How many games would be considered a good sample for data purposes?

Interested in your thoughts for those that get into the data end of it all.
Thanks in advance 

in my opinion, NONE

 

[azbob]

instead of people randomly throwing out numbers, why don't you google margin of error table and you will know the confidence level based on sample size? You could apply this for "static" elements of the calculations (such as your home dog example), but not for calculations where the variables change...for instance, Pittsburgh as covered vs. Cleveland ten straight times after a bye week. (Pitt and Clev have different players/coaches/etc over that time frame).

 

[CBBPHENOM]

instead of people randomly throwing out numbers, why don't you google margin of error table and you will know the confidence level based on sample size? You could apply this for "static" elements of the calculations (such as your home dog example), but not for calculations where the variables change...for instance, Pittsburgh as covered vs. Cleveland ten straight times after a bye week. (Pitt and Clev have different players/coaches/etc over that time frame).

Agreed. This is a stat question. I know what I'm talking about so I'll chime in. 20 is minimum. 50 is considered large sample. The higher you go, the less standard error and highest confidence interval. Good luck