More on NBA Middles.

[grapefruit]

First off, great job and thanks a lot for the numbers.

I am just starting out as a middler in the NBA so any info like this is very helpful to me. I find it hard to anticipate line movements, so I usually place middles simutaneously, meaning I only make middle bets when both sides are available to bet right now. So I don't take the risk of betting the wrong side of a game and having to cover myself at a loss.

Because of this, I can't find 3 point middles, most often only 1.5 point middles, which are generally accepted to be profitable by everyone I have talked to and by other posters.

Would it be safe to assume that 1.5 point middles hit half the time that 3 point middles hit (13.46 X 1/2, 6.73%) or one and a half more times that 1 point middles (1.5 X 5.64, 8.46%)?

 

Grapefruit,

Not sure what all my data "means".

I have the line falling exactly on the number 3% of the time, within (.5) 5.64% of the time.

So it would "make sense" to me the 1.5 middle would be somewhat higher than the 1.0 I checked, your guess "makes sense" to me.

If I get full of beans someday, I should be able to check the 1.5 middle in two separate queries, >=1 and <=.5 AND >=.5 and <=1. [I think).

What I did this study for, and is a little surprising perhaps, is to test the value of the "middle" in the higher point spread catagories, "the extremes". It passed with flying colors, and I would go so far as to say THE MIDDLE HAS MORE VALUE IN THE EXTREMES THAN NEAR THE CENTER. The slight home dogs for instance, were the worst middle buy, they would have killed you (and common sense would point towards them being the best IMO.)

 

[Sodium Pantethol II]

ACTUALLY CAN BUILD ONE FROM THE GOLD SHEET ARCHIVES AND GO BACK QUITE A BIT FURTHER.

TRY USING TWO SUBSETS, PRE AND POST ALL STAR BREAK. OR PRE NEW YEARS AND POST NEW YEARS.

 

With Access and a query, breaking the db up either way (At Jan 1st or at All-Star) is possible.

What do you think would be there?

As far as "close to the number", my guess would be nada. Totals, maybe.......

 

[Sodium Pantethol II]

THE NUMBERS SHOULD CONTINUALLY GET TIGHTER AS ANY SEASON PROGRESSES.

THE BEST BETS ARE ALWAYS IN THE BEGINNING OF ANY SEASON BEFORE SLOW MOVING POWER NUMBERS CATCH UP.

IT WOULD BE SHOCKING IF THE LIKELIHOOD OF ANY PARTICULAR MIDDLE HITTING DOESN'T INCREASE AS THE SEASON PROGRESSES.

 

I could see the power numbers "tightening up" in college, but the NBA's funny IMO (as is the NFL).

Seems to me the power numbers don't catch up, they measure "history" a lot, and team's second halves don't often resemble their first halves (the Patriots and Heat two easy cases in point).

But if I get a chance, will check it, report back. Will use Jan 1st, think it is better now that I see where you are going with it.

 

[Blackman+]

Ok -1 at -110 and plus 2 at -110 is a good bet?

 

[grapefruit]

I agree that middles further away from pick em are more likely to land because with teams fouling with seconds to go it’s very rare to find buzzer beater shots to win the game with just seconds left or teams winning by just a few points. I can use common sense for that. I find your numbers surprising that middles from 10-14 are more likely than middles from 4-7, but numbers don’t lie.

What I’ve been doing for the past hour or two was take the average of your info and sick gambler’s info. You say that each number lands about 3% of the time, SG saying it lands 4.4%. I could care less who is right, I’m just taking the average (3.7% chance of each # hitting on average)

This is not entirely correct as some numbers will be more or less likely at hitting, as we already proved, so we would obviously prefer to make middles on games with spreads at 4 or more.

I will post a chart that I made showing the average advantage the middler has per type of middle (1 point middle, 1.5 point middle, etc.), and at different levels of juice. The advantage is in the form of a percentage. It’s similar to what I do for the NFL.

In the NFL the fave winning by 3 hits 9% of the time.

i.e. if I bet 110 to win 100 on the Giants –3, and 110 to win 100 on the Raiders +3.5, if they were playing each other.

.09 x 100 = $9 profit per middle on average
.91 x 10 = $9.09 loss per middle on aveage
total loss of $0.09 per middle of $100, and my advantage would be –0.09%

if I bet 105 to win 100 on both sides instead of 110..

.09 x 100 = $9 profit per middle on average
.91 x 5 = $4.55 loss per middle on average
total profit of $4.45 per middle, and my advantage would be 4.45%

 

[grapefruit]

ADVANTAGE OF EACH TYPE OF MIDDLE AT DIFFERENT LEVELS OF JUICE

IN THE CHART I REFER TO TOTAL JUICE AS: -110 ON BOTH SIDES HAS A TOTAL JUICE OF 20, -105 ON ONE SIDE AND -107 ON THE OTHER HAS A TOTAL JUICE OF 12, ETC...

TOTAL JUICE --- MIDDLES OF

2PTS 1.5PTS 1PT
20 6.76% 2.77% -1.22%
19 7.22% 3.23% -0.76%
18 7.68% 3.69% -0.30%
17 8.13% 4.15% 0.15%
16 8.60% 4.61% 0.62%
15 9.06% 5.07% 1.08%
14 9.47% 5.53% 1.54%
13 9.98% 5.99% 2.00%
12 10.43% 6.45% 2.45%
11 10.90% 6.91% 2.92%
10 11.35% 7.37% 3.38%

 

[AnotherAnon]

Peep,

Great work and thanks for sharing!

1) Do you think that the College "Duke -27" lines would outperform the NBA big favorites? Perhaps in college, that group would have the best pecentage of hitting as others have suggested? Your studies show the big favorites in the NBA are a great buy, but they weren't the best for the groupings you selected.

2) I think maybe you took Sick's comments a little too harshly. I chuckled at his comment, until I read your reply, where it appears you took it as an insult.


Grapefruit,

Thanks for the chart! I posted an example in the previous thread using 8% for the 1.5 middle hitting and full juice just to ask if I was doing it right, and got a similar number (2.8). Thanks for providing the rest of the chart.

 

[Sodium Pantethol II]

BUT ONE WOULD NEED TO BE DROPPING LSD TO NOT KNOW THAT NUMBERS ACCURACY IMPROVES OVER THE COURSE OF ANY SEASON.

DO YOU DOUBT 14 WAS THE RIGHT NUMBER IN THE SUPE?

DO YOU THINK YOU'RE GOING TO FIND ANY BARGAINS COME NBA PLAYOFF TIME WHEN EVERY NUMBER IS AS OBVIOUS AS YOUR NOSE?

 

[Mothe]

Peep, you may have explained this elsewhere, but when you give lists of games with lines both on integers and on half points and say that a certain percent of them fell within a certain number of points from the line, how do you end up treating them in a consistent manner? I see that maroon asked this a few weeks ago, but I never saw the answer. For example, let's say you're trying to determine how often games fell within .5 point of the line. If you have a game with a line of 5.5, it can land on either 5 or 6, but if it has a line of 5, it can only land on 5. To treat both types of games the same, wouldn't you have to do something like counting the times that a game that had a line of 5.5 fell on 5 or 6 and then divide by 2 (or, as an alternative I wouldn't prefer, perhaps arbitrarily picking either 5 or 6) in order to compare it with how often games that had a line of 5 landed on 5?

Has anyone been able to contact Rio or Skybook today?

 

I didn't really understand the question, so that is why I didn't answer him.

Don't know why I don't understand it, but I am just drawing a blank.

I listed the data from the query I ran, which is all games that fell within .5 of the number (I think), can you answer it yourself from that?

My thoughts are that a half point number would include both sides only, and a whole number line would have to hit "bang on" for the query to show a "winner". As I recall from the original thread, about 1/2 the numbers were "*.5"'s.

As I remember it, to get my 3% of the numbers fell bang on part, I excluded .5 lines altogether (see thread "NBA middles/Sick Gambler")

Does this help any? If not, post back.

 

[Mothe]

Peep, maybe the purpose of your study is different from mine. My reason for being interested in these numbers is to know how profitable certain middles are. If I'm thinking of middling a game that has lines of 4.5 and 5.5, I want to know what its chances are of landing on 5. If I'm thinking of middling a game that has lines of 4 and 5, I want to know what its chances are of landing either 4 or 5. Lumping games that have lines of 5.5 and including how often they landed on 5 or 6 with games that had lines of 5 and including how often they landed on 5 won't show how often any middle will hit, since some adjustment for the fact that one type of game has 2 numbers to land on and the other type only has one would then have to be done. I'd like to see 2 separate totals - one of how often games with lines on an integer landed on that integer and another of how often games with lines on a half point fell a half point away. I would then take an average of the latter figure divided by 2 and the former figure to estimate how often games land on their line.

 

I don't think you should have too much trouble getting those numbers out of the data provided here.

If you wanted to find out how many times the "five" hit dead on for instance, all you have to do is run throught the "4-7" data, count the number of fives, then the number of hits. It should be close to 3% if I am right.

And I would think if you checked the "4.5"'s for how many hit on the "4" or the "5", that would give you the answer to your second question.

Shrink said 1.5 dif is "O.K.", especially with reduced juice, and that seems to make sense.

What I was mostly interested in with running this latest set of data was seeing if Sick's idea that the middle was as effective with large number point spreads (i.e."-11") as it was with low point spreads (i.e. "-1"), and was surprised to find it was if anything A LITTLE MORE EFFECTIVE AT LARGER SPREADS. Think whoever pointed out "fouls late" as the reason is spot on.

 

[Mothe]

My summary of the Sick Gambler / maroon debate is that Sick Gambler said that games land on their betting lines about 4.4% of the time but that maroon believed that was a misinterpretation of your data, since it wrongly included all games that landed within .5 point of the line as a "hit," whether the line was on an integer or on a half point. If the figure of 4.4% isn't erroneously based on mixing apples and oranges like this when the real figure is 3%, what is its source?

 

I ran a bunch of numbers from a four year db to get my 3%, so I know where that came from. I only used whole numbered lines for the data, and it is listed under "Middling NBA/Sick Gambler" thread.

Sick I believe did his study by "hand" rather than computer, and logically reasoned that if a line was a "point five" he could shop it a bit and get a whole number the way he was leaning, hence he counted the "3.5" as a "hit" for the 3 as well.

Real worldwise, I didn't mind his thinking at all, hence I agree with his 4.4% number given his methodology.

 

Last one.

 

[BenDover]

"Peep, maybe the purpose of your study is different from mine. My reason for being interested in these numbers is to know how profitable certain middles are. If I'm thinking of middling a game that has lines of 4.5 and 5.5, I want to know what its chances are of landing on 5. If I'm thinking of middling a game that has lines of 4 and 5, I want to know what its chances are of landing either 4 or 5. Lumping games that have lines of 5.5 and including how often they landed on 5 or 6 with games that had lines of 5 and including how often they landed on 5 won't show how often any middle will hit, since some adjustment for the fact that one type of game has 2 numbers to land on and the other type only has one would then have to be done. I'd like to see 2 separate totals - one of how often games with lines on an integer landed on that integer and another of how often games with lines on a half point fell a half point away. I would then take an average of the latter figure divided by 2 and the former figure to estimate how often games land on their line."

I ran it and I can tell you that you do not need to know how often a 1.5 middle hits on 5, all you need to know is that is it profitable or not. I ran it in such a way that I looked how often -3.5/+5, -4/+5.5, -4.5/+6, and -5/+6.5 hit and you know what? It did not matter, either it made money or it did not make money.

My basic findings are here...

http://therxforum.com/6/ubb.x?a=tpc&s=100090022&f=988094022&m=2943075464

You can bust your a-ss like I did to come to the same conclusions. Also, I would not trust anyone than ran something by hand. Humans make errors, the computer makes 0 math errors unless you program it wrong which is less likely than a hand calc. The numbers start running into each other.

Also, the most profitable range seems to be -4/+7, not farther out due to fouling. CBB lines outside of double figures are much less profitable. Middling Duke -27/+30 is a joke. In college's the standard deviation is larger for those outliers...

 

[MadCapper]

Very Very valuable moneymaking thread!