[tool21]

Personally, i think i have a pretty sharp way of figuring out what spreads should be in an NBA/NCAAB game but i have a few questions which is best described in an example.

Tomorrow is Monday, 3/31 and Denver is play Phoenix.


My math model shows this:

Denver 217.1539742
Phoenix -4.857452044

If I were to look at this I would love to take the under but books are expecting this game to be a track meet and most of the time numbers can't pick this out(although i may be right) Should i adjust my spreads so that my predicted total matches with the books total? Or does this completely go against what making a math model is about? In this case i would have

Denver 236 (from book)
Phoenix -5.27 give or take which is right where the books line is.

It's not much a difference, but I'm just curious as to what you guys think.

Do any of you guys use the books information to help you in figuring out lines? From my model it seems like totals are so random that I don't bet them but the spreads work more often than not.

[Arnold]

I think spreads are more random. Figuring out approximate book totals is quite easy.

As to your problem, how do you know that your math model works in first place?

[tool21]

I also use situational and other things including the math model so as to 'enhance' it. It's won me money this year.

Arnorld, how do you approximate book totals easilly? What do you get for the Den - Pho game? Maybe i can use it in my capping.

[Arnold]

I also use situational and other things including the math model so as to 'enhance' it. It's won me money this year.

I would just check your updated prediction against past games. This is how you would know if your prediction works.

Arnorld, how do you approximate book totals easilly? What do you get for the Den - Pho game? Maybe i can use it in my capping.

It's not useful for predicting over/unders. It's only good as a prediction of over/under line when you don't have one yet. You can do the simple combined score average: (Visitor_PPGF + Home_PPGA) / 2 + (Visitor_PPGA + Home_PPGF) / 2 = predicted line. This will be pretty close for anything under 205 or so. After that it's almost always way off by minus 5-15 points.

For Den/Pho game I think it showed around 215. I also use compensated stat of the above Points Per Game For/Against, and it showed 230. So the actual 236 is about right. It rarely shows a total that high.

Once again, these stats won't get you any winners, so don't waste your money on it.

[tool21]

Arnold, no offense but that way of finding totals is very elementary and that kind of stuff was done my very first day of capping when i was new to the scene. I'm pretty advanced as in using regression analysis and other forecasting techniques and i consider my spreads to be very sharp. If you were to do my math model with pencil and paper it would take you about an hour per game and it would be impossible to look at more than 1 game a day trying to stay sane. You would give up day 2 if you didn't have a way to make things easier. Thank god for programming to cut down this amount of time. IMO it's very sharp.

But for the sake of the topic i started i was just wondering if it could make my spreads more accurate by proportionally changing my predicted total to the books total. I guess nobody would have an answer to this and I could only figure this out by testing. It wouldn't even really matter because like the example above, adjusting a 217 total to 236 is only about .5 points difference in the spread. Thank you for your help though, it's always appreciated!

[Arnold]

Arnold, no offense but that way of finding totals is very elementary and that kind of stuff was done my very first day of capping when i was new to the scene.

That's why I said it is quite easy to figure out over/under lines. I never said it was complicated.

I'm pretty advanced as in using regression analysis and other forecasting techniques and i consider my spreads to be very sharp. If you were to do my math model with pencil and paper it would take you about an hour per game and it would be impossible to look at more than 1 game a day trying to stay sane. You would give up day 2 if you didn't have a way to make things easier. Thank god for programming to cut down this amount of time. IMO it's very sharp.

I'm not doing anything with a pencil. I have a computer program myself that calculates things like that for me. Also I'm not sure if anyone does regression analysis by hand today.

But for the sake of the topic i started i was just wondering if it could make my spreads more accurate by proportionally changing my predicted total to the books total. I guess nobody would have an answer to this and I could only figure this out by testing.

Exactly. I'm a little surprised that you haven't tested it already. That is if you have a database of past games.

[tool21]

Arnold, i usually just write down my lines and i move on to the next game without saving. Then i throw away the papers i write on. The biggest thing is that i update a MASTERSHEET daily with new NBA stats which would alter all the past saved spreadsheets if i were to save them as well. Guess that's not a good way to do things.

I just don't understand how the books can have a line that high and it frustrates me that i'll probably never have a model that will detect a line that high based on averages. I use value of 5 for spreads 5 for overs and 10 for unders and the spreads have been winning im certain but i'm not so sure about the totals something is just off.

Like i said, 30 points of a total is worth about .5 points of a spread for me so it makes sense that spreads work better this way. But those totals blow my mind.

I'm gonna look at my sheet and see if i can think of anything.

[tool21]

Bingo!

Total at 237. Simple math error in the sheet.



Total: 237
Spread: Pho -5

[Wheell]

However, the main question still remains. What to do when you differ by a LARGE amount. I can say that I don't change my models and place the burden on the books to beat me when I feel I have a large edge. Other people prefer to sit those games out. I don't know of anyone who changes their models based on the current lines.