[dimes1111]

I am new to the forum and new to NBA betting as well. I have a developed a very basic model for predicting points scored for each team as well as the game total. I have only been doing this for not quite a full month, just 160 games. The lines that I have bet are hitting right at 50%. But I noticed that of the 480 projections only 98 have been over the line and 20 more have been just under the line by a couple tenths of a point or less. Of these 118 situations the over has hit 73 times and lost 45 times (62%). Obviously as a whole my model is flawed. But is it possible that since what I'm doing heavily favors the under, when it does predict the over can I continue to win at anywhere near this rate?

[nhjxmen]

Hi, I'm new to. Welcome. I am confused on how you got your formula and what you did. Can you walk me through it? As long as your formula predicts a winning rate higher than 52.45% with a juice of 10%, then you are good to go! I doubt it will work if it's based only on past totals. Plus, your data is only 160 games? Is that a couple months of regular season basketball? I have compiled a few databases myself, but couldn't find any patterns to make a formula for totals based on only past totals.

[easyliving]

your sample size is much to small.

[HeeeHAWWWW]

Of these 118 situations the over has hit 73 times and lost 45 times (62%). Obviously as a whole my model is flawed.

It's possible you might have struck on some unknown gold ......... but it's unlikely, to be honest. When you dig around the numbers enough, eventually you'll find something that appears to be a pattern. There are patterns everywhere, but most are meaningless - us humans are unfortunately rather good at interpreting these as meaning, where none exists.

As easyliving says above, sample size. Even ignoring that your selection of these 118 results may have introduced some pretty hefty bias, the chance of going 73-45 or better just by coinflipping is about 1 in 159. Reasonably unlikely, but it happens ....... and when you're specifically looking for "unlikely", you tend to find it pretty quickly.

Having said all that, it's unlikely enough to be worth investigating. Don't bet on it, try understand it first.

[AceSportsbook]

When you dig around the numbers enough, eventually you'll find something that appears to be a pattern. There are patterns everywhere, but most are meaningless

I would tend to agree with this. There is a difference between finding a pattern that happens to exist in historical data, and finding a model that can be used to predict future data. In other words, just because a pattern exists in past data, it does not automatically mean that pattern is predictive, nor that the variations it interprets are causally related to the outcomes that occurred.

I would recommend you use your model to predict your plays, and then just keep track of your W/L record on the side without actually playing yet. If after some testing, it turns out your pattern is predictive, then start betting with it... and sell it to me.

[thehoorse]

As others have pointed out, most patterns are meaningless. The ones that were not created out of coincidence, are most likely no longer reliable once you have noticed them since books adjust and compensate for past inefficiencies. One could say it might be best to buck the trend in hopes of catching books over-adjusting. That is not easy either since it is usually done gradual.