Monday, October 15, 2007

Accuracy of databases

In the past, I've compared the numbers in my databases with other people. I'm actually in the process of doing so right now. It really is easy to screw up a negative or positive sign and have the home team as a 12 point favorite when they should be a 12 point underdog. So it is useful to compare and contrast numbers to get out those stupid errors.

But at the same time, there are other lines that may be different between different databases. What should the two people do about that? Should person 1 change or person 2? Actually, I don't think either has to necessarily change as both could be right.

You must be thinking "what the heck is he talking about? If two people have different lines, how could they both be right?"

The answer is that the line moves. One person may capture the line that was stagnant from Tuesday to Saturday at -7. Another person may capture the line at 12pm Eastern, an hour before the NFL 1PM kickoffs, and capture the line of -8 because the line moved on Sunday morning. In my opinion, both people are right, neither is wrong. Of course, it is helpful to have both lines, so you could look at line movements as well. Unfortunatley, I don't have that information, and frankly, I can't recall when I captured the data for my database, nor can I tell when the other person captured his data. So I'm left with having to choose a number. Either keep mine, change mine to his, or do something in between. Checking with other sources (such as can be helpful, but only to the extent that the line didn't move. For example, let's say I used a closing line and so did Covers, but the other person used a Saturday line. That doesn't mean I'm right and the other guy is wrong just because Covers' data agreed with me.

So the moral of this post is simply try to be diligent when inputting your database numbers, but you can only do so much. Besides, results from databases should always be looked at askew, especially in the NFL since the sample sizes are so small.