A Survey--What do you like better: Better Vig? or Better Spread?

[BSTPNM68]

When shopping for lines, what do you prefer, a better vig or better spread? What is more profitable in the long-term, "generally speaking?"

For example,

In tonight's, Milwaukee-Denver game, let's say later on when I am shopping for a line Pinnacle has Milwaukee at +6.5 (-105) and Bowman's has Milwaukee +7 (-110).

What is the better choice in the long run? It seems with a larger spread, the lower vig would be the better choice--and vice versa.

Thanks,

BT

[JoshW]

Take the +7. If you want to be conservative you should give most half points (on single digit spreads) a value of around 9 cents. Some are worth less, some a little more.

[The Great One]

I'd put more value on the lower vig, in this case +6.5.

[Doug]

buy any half-point in hoops for 5 cents, so +7 -110, no brainer. I say most halves are worth about 7 cents.

[Ganchrow]

In tonight's, Milwaukee-Denver game, let's say later on when I am shopping for a line Pinnacle has Milwaukee at +6.5 (-105) and Bowman's has Milwaukee +7 (-110).

Just as Doug and lakerfan have already said, take the +7.

And it's not even close. Assuming a reasonable (not insanely high) total I'd take the +7 all the way up to about -116 or so.

[Bill Dozer]

An NBA half is worth around 9 cents max. I don't think a full game half is worth more than 7. This is always a good topic with different opinions.

[JoshW]

Pinnacle will let you sell a half point from 6.5 on to 7 for 9 cents, while making you pay 12 cents to buy the half onto 7. I would say that is a good indication 9 cents isn't a max, but rather a minimum IPO (in Pinnacle's opinion ). 6 and 7 are key number in NBA from every database I have seen as well as just from experience betting NBA.

[Ganchrow]

An NBA half is worth around 9 cents max. I don't think a full game half is worth more than 7. This is always a good topic with different opinions.

From the 1999-2000 season through the 2004-2005 season, approximately 4.5% of games opening with a spread of 6.5 or 7 points ended with the favorite winning by exactly 7 points. (sources: databaseBasketball.com and covers.com).

So assuming a 4.5% probability of hitting the 7 and an even money probability on the +6.5 spread, it follows that 52.356% of the time the bet doesn't push the +7 will win and 47.644% of the time the bet doesn't push the +7 will lose. Hence, the theoretical hold of 2.381% on a spread of +6.5 at -105/-105 is about the same the expectation on a +7 at -115.7.

[Ganchrow]

Pinnacle will let you sell a half point from 6.5 on to 7 for 9 cents, while making you pay 12 cents to buy the half onto 7. I would say that is a good indication 9 cents isn't a max, but rather a minimum IPO (in Pinnacle's opinion ). 6 and 7 are key number in NBA from every database I have seen as well as just from experience betting NBA.

This, along with the 4.5% probability figure from above, together imply that Pinnacle is charging only about an extra penny off its typical -105 line set both when buying and selling on to the 7 from the 6.5.

[Illusion]

I prefer the lower vig.

[juuso]

Lower vig for me too. The worth of a half point varies from game to game imo.

[Dark Horse]

Someone posted a link here the other week with the percentage of basketball games that ends within a half pt of the line; for different spreads. The average, I think, was 3.7%. I don't know how accurate those records were - they didn't look too good for the NFL -, but even if that is close the lower vig wins out.

In any case, it shouldn't be too hard to find out the exact percentage of games that end within a half point of the spread.

[pags11]

I thijnk a lot of this depends on how you think a particular game will go...if it's a football game and you can get +15 (+100) or +15.5 at (-107) you have to factor in what you would consider your key number areas to be (and in this case be more inclined to go with the even money wager)...in hoops, a fav at -1.5 (-103) isn't as of much value as say the same team at -1 (-109), again being a key number area...so it will depend on the game...I always try to look for good juice, but don't want to cut myself short on the number either...

[THE SHRINK]

If you have an account with Pinnacle, you will see a drop down menu offering half points and more on NBA games...

You will also see that they charge different amounts to buy depending on the line...

As a general rule, the only number that many sharpies consider it OK to BUY a half point for 10 cents is ONTO 2, period...

One other wise guy I know also buys ONTO 3, but I don't...

THE SHRINK

[JoshW]

ganchrow provides a fairly definative answer. Databases that go back even further also show they same trend with the half point around 6 and 7 worth ten cents or slightly more. This is in line with the opinion of what most people think is the smartest book on earth. Pinnacle doesn't let you sell half points for 9 cents because they are nice guys LMAO. Choose to go against it at your own loss, but it amazes me when leave money on the table by taking the worse bet.

I know most people are answering in general, but in the specific example was what I was commenting on. In general there is very few cases where one wouldn't pay 5 cents for a half point, even with double digit spreads.

[JayEgdarWho]

A little sideways from the topic, but the degree to which the Pinny line hops around at key times during the day and just prior to post provides a great opportunity, for the patient, to have your cake and eat it too.

[Ganchrow]

Someone posted a link here the other week with the percentage of basketball games that ends within a half pt of the line; for different spreads. The average, I think, was 3.7%. I don't know how accurate those records were - they didn't look too good for the NFL -, but even if that is close the lower vig wins out.

In any case, it shouldn't be too hard to find out the exact percentage of games that end within a half point of the spread.

This is a link to that discussion: http://forum.sbrforum.com/showthread...selling+points

And a link to the half point chart:

http://www.sportsbookfreepicks.net/halfpointchart.htm

But be warned that the poster's claims of having used 10 years of NBA data as source material are clearly fallacious and that he doesn't make public his methodology.

EDIT: Even if the push % were completely accurate you should obviously still take the $BE column with a few grains of salt as the math itself doesn't lend to such a broad generalization.

[Ganchrow]

As a general rule, the only number that many sharpies consider it OK to BUY a half point for 10 cents is ONTO 2, period...

One other wise guy I know also buys ONTO 3, but I don't...

With all due respect Shrink, this is simply not borne out by the statistical record.

[pags11]

good to see you posting here shrink...

[turbobets]

But be warned that the poster's claims of having used 10 years of NBA data as source material are clearly fallacious.

What evidence do you have that my claims are fallacious?

and that he doesn't make public his methodology..

Ask away!

EDIT: Even if the push % were completely you should obviously still take the $BE column with a few grains of salt as the math itself doesn't lend to such a broad generalization.

I agree the past does not dictate the future.

[imgv94]

Lower Vig all the way!!!!

[BuddyBear]

getting the best number is key for me..even at -110. I mean if you bet long term you are gonig to win your share of games by half a point or push when you should have lost. All that adds up to the juice you save....

IMO, it's a wash...but I don't have anything scientific to back me up just experintial knowledge

[BSTPNM68]

Great feedback from everybody. This clearly takes me up the learning curve faster, not only for vig vs. points but also the value of selling points at Pinnacle.

[Ganchrow]

But be warned that the poster's claims of having used 10 years of NBA data as source material are clearly fallacious.

What evidence do you have that my claims are fallacious?

I'm sorry if I've come off as antagonostic, but all I meant by this was that you present data for too few games to constitute 10 full NBA seasons.

and that he doesn't make public his methodology.

Ask away!

OK, so what was your methodology?

Even if the push % were completely accurate you should obviously still take the $BE column with a few grains of salt as the math itself doesn't lend to such a broad generalization.

I agree the past does not dictate the future.

The issue isn't whether historical results are necessarily indicative of future performance, it's that breakeven dollar value shouldn't really be considered an experimental statistic but rather a value derived from the relevant push likelihood given certain market conditions.

In order to deternine the fair price of buying or selling the half point on to or off a figure given a push probability, one would have to calculate it from first principles. Simply stating that the breakeven is x cents ignores the fact that actual breakeven values will vary depending not only on associated push probability but also to varying extents on:

1. direction (buy vs. sell)
2. movement (on vs. off)
3. line set
4. price
5. total (although in analyses such as these, the impact of the total on half point pricing is typically ignored largely because in most cases such impact is quite small and can't be calculated from first principles)

For example, using the data from your chart we see that given a spread of 4½, 5, or 5½ the margin of victory will be favorite by exactly 5 about 5.2% of the time (although of course the methodology used will dictate the precise set up). However, the fair value of buying the half from 4½ to 5 given both a line set and a price of -110 will differ by more than 2 cents from the cost of selling the half from 5 to 5½ given both a line set and price of -105. Admittedly, this is one of the more extreme cases (the higher the push probability the higher the divergence) but you get the idea. When I get a chance I'll try to put together a spreadsheet to do these calculations.

Anyway though, my thanks for having shared your chart. Appreciated.

[Doug]

Do whatever you guys feel is correct. For me I'll take a half point in hoops for 5 cents more juice, every time, foots is totally different.

[turbobets]

you present data for too few games to constitute 10 full NBA seasons.

Non-integers are not included in the chart only whole number pointspreads are used.

The issue isn't whether historical results are necessarily indicative of future performance, it's that breakeven dollar value shouldn't really be considered an experimental statistic but rather a value derived from the relevant push likelihood given certain market conditions.

It will always be an experimental statistic because it is impossible to derive every possible market condition or variable that affect the odds of a game ending in a push.

For example, using the data from your chart we see that given a spread of 4½, 5, or 5½ the margin of victory will be favorite by exactly 5 about 5.2% of the time (although of course the methodology used will dictate the precise set up). However, the fair value of buying the half from 4½ to 5 given both a line set and a price of -110 will differ by more than 2 cents from the cost of selling the half from 5 to 5½ given both a line set and price of -105.

I agree with the above intimation.

After all it is up to each handicapper to decide for themselves the amount of subjectivity they should add to their brew of systems, formulas and trends. In case you didn't notice the half point page was written in html not stone.

[Ganchrow]

Non-integers are not included in the chart only whole number pointspreads are used.

So then I'd argue that you're not really using 10 full nba seasons and as such, by blindly discarding half your data points, your methodology is likely suboptimal.

If you have the time you might want to take a look at Stanford Wong's simple and starightforward methodology for valuing half point data, presented in his book Sharp Sports Betting.

The issue isn't whether historical results are necessarily indicative of future performance, it's that breakeven dollar value shouldn't really be considered an experimental statistic but rather a value derived from the relevant push likelihood given certain market conditions.

It will always be an experimental statistic because it is impossible to derive every possible market condition or variable that affect the odds of a game ending in a push.

Saying that given a stated push probability, breakeven dollar value is an experimental statistic is like saying that given a measured value for average annual Fahrenheit temperature, the average annual Celsius temperature would still need to be independently measured. The fact is that for a stated push probability it's an algebraically facile task to derive a precise half point valuation (the dependent variable) as a function of the first four (independent) variables I listed above (direction, movement, line set, and price) just from first principles of probability.

[turbobets]

Funny thing is I cannot find a page in Wong's book where he calculates push percents using non-integer pointspreads.

I did find the following.

"The data in Tables 17 to 19 are summarized in table 20. Some subjectiveness is involved here. My intention in creating table 20 was to select the integers that I think best represent the frequency of pushes in this years NFL season."

There is no subjectiveness in my charts just raw data. I am not saying this method is better or worse just a alternate approach of presenting the figures.

I believe a nearly accurate estimate can be made and used to decide if it is advantageous to buy on or off a non-integer (1/2pt) pointspread using the the data I have provided.

It would be beneficial to use a significant sample of games to calculate how often a non-integer pointspread landed on the whole number just above and below it. If it has been done I have not seen it. Not from Wong, yourself or any other person.

[Ganchrow]

It would be beneficial to use a significant sample of games to calculate how often a non-integer pointspread landed on the whole number just above and below it. If it has been done I have not seen it. Not from Wong, yourself or any other person.

The underlying axiom of Wong's method is that for very small differences in pointspread the difference between the associated "true" probabilities of a given MOV (margin of victory) occuring is small in relation to the sampling error. As such it may be possible to generate a more robust unbiased estimator of MOV probabilities for given spreads by also considering likelihoods for adjacent spreads. The subjectivity (or the art) surfaces in the attempt to determine what exactly is meant by "very small differences".

(Another important axiom might also be that for a given MOV M, and two spreads S1 and S2 where S1 < S2, Prob(M|S1) > Prob(M|S2) if M <= S1 and Prob(M|S1) < Prob(M|S2) if M >= S1. The fact that this is not always borne the data is, as far as I'm concerned, an excellent argument in favor of the Wong-style aggregation.)

By way of example, when considering the likelihood of a pointspread of let's say 5 ending in a push, one might look not only at the historical likelihood of such occuring, but also at the historical likelihoods of pointspreads of 4½ and 5½ ending with a fave MOV of exactly 5 (although when dealing with a half point radius you'd be double counting non-integer spreads and so you may want to adjust for that by weighting integers spread MOV's twice as much -- although in reality this shouldn't bias the resuls.) Or you might want to go further, and look at MOVs for spreads within a full point of your target (and this is what I personally think best in the NBA).

[pags11]

damn granchow, that's some unbelievable stuff...

[jjgold]

Long term the low vig will win out