Skybook lines... again

[Dark Horse]

Have said this before, but it bears repeating, because this book uses its free 1/2 point as carrot.

NBA finals game 6. Line DAL -6 everywhere. At -110 or -105, depending on the book. Sportsbook is -115 for DAL -5.5. Now Skybook...

MIA +6 -104
DAL -6 -116

They pull this all the time. Just trying to keep them honest.

[bigboydan]

thanks for the heads up bud.

[Dark Horse]

FWIW, the limit is 300 for free 1/2 pts, so the only way to hammer this, and let Skybook rethink its lines, is through many bettors. Right now:

Pinnacle DAL -6 -103
Skybook MIA +6.5 -104 (with free 1/2 pt)

[DrSlamm]

Damn those sportsbooks for trying to make a buck

[Terris]

Damn those sportsbooks for trying to make a buck

What iam thinking, its not bad but also not great, they are no pro book like Pinny etc so why would they want all those scalps.

[DrSlamm]

they dont their lines shift often even after a 300 dollar bet

[Ganchrow]

If the 'true' -110 lineset market for a 6-point spread were -110/-110 (implying equal odds of either side coivering the spread), then with Skybook's free ½ point, the approximate expctations would be:
Miami +6+½ ~ +2.5%
Dallas -6+½ ~ -2.7%

And the approximate 6-point spread line equivalencies would be:
Miami +6 ~ +105.0
Dallas -6 ~ -105.4

So, in other words, if you believe that Miami has the same chance of covering the spread as Dallas, then the Miami +6 -104 with the free ½ point represents a positive expectation bet.

(Note that this is based on a roughly 4.5% chance of Dallas winning by exactly 6 points. I've derived this probability form the historical record and I believe it to be a good predictor. Indeed a proabability of an exactly 6-point margin of victory by Dallas as low as only 2% would still yield a positive expectation bet on Miami.)

[Doug]

Ganchrow: What really happens though is Skybook uses the vig shift to mimic Pinnacle's lines, they are almost always in line with Pinny.

Sky is OK but that free half is over-rated, as the juiced side is the same side Pinny juices.

If a game were +/- 6 -105 at pinny, it's probably +6.5/ -5.5 (-110) at Sky, and at -104 ( or cheaper) at Mansion.

To simplify it, one can say the half point is worth about 7 cents ( pretty close for most spreads).

So getting that half is very equivelant to getting -104 at a half point worse at Mansion. The half is slightly prefered but not by much. In the long run it won't make a lot difference if you play it at Mansion, Pinnacle, or Skybook.

Skybook is kind of over-rated and charges for transactions. They're good to have for when their line really is different, but it doesn't happen much.

[Ganchrow]

Ganchrow: What really happens though is Skybook uses the vig shift to mimic Pinnacle's lines, they are almost always in line with Pinny.

Sky is OK but that free half is over-rated, as the juiced side is the same side Pinny juices.

If a game were +/- 6 -105 at pinny, it's probably +6.5/ -5.5 (-110) at Sky, and at -104 ( or cheaper) at Mansion.

To simplify it, one can say the half point is worth about 7 cents ( pretty close for most spreads).

So getting that half is very equivelant to getting -104 at a half point worse at Mansion. The half is slightly prefered but not by much. In the long run it won't make a lot difference if you play it at Mansion, Pinnacle, or Skybook.

Skybook is kind of over-rated and charges for transactions. They're good to have for when their line really is different, but it doesn't happen much.

Good points. Although I may be misinterpreting your post, I'll add the following:

Firstly, the 6 does happen to be a fairly important NBA number and as such the ½ point off the 6 is a bit more valuable than 7c. Based on historical frequencies I put it at around 9c for lines close to even.

Secondly, you're clearly 100% correct about Skybook's juiced line +½ being roughly equivalent to Pinnacle's. However, the idea is that the 'non-juiced' side is significantly more valuable at Skybook. As a simple means of comparison, notice that Pinnacle is offering Miami +6.5 -115, while VIP is offering Miami +6.5 -104. Putting it another way, while Pinnacle is offering up the spread at a hold of +2.38%, VIP is effectively offering it up at a hold of only about 0.09%.

So the conclusion would be that from a strict expectation standpoint (and fees aside), while a player should be essentially indifferent between the Dallas lines at Skybook and at Pinnacle, the Miami line at Skybook is vastly preferable to that at Pinnacle.

[pags11]

good stuff as usual ganchrow...I just think skybook does this so they don't get middled, but most of the time I'd take the reduced juice, vs. taking a side at -116 and getting the free half point...I know you (ganchrow) ran some numbers last year regarding this, but I just hate watching a game knowing I've spent a lot of juice on the wager...that's just me though...

[Doug]

good stuff as usual ganchrow...I just think skybook does this so they don't get middled, but most of the time I'd take the reduced juice, vs. taking a side at -116 and getting the free half point...I know you (ganchrow) ran some numbers last year regarding this, but I just hate watching a game knowing I've spent a lot of juice on the wager...that's just me though...

No half point in hoops is worth 12 cents. I'd never take +6.5 -116, over +6 -104.

7 cents is a good starting point to evaluate it's worth for those not as mathy as Ganchrow.

Here's a general guidline that will work for most ( hoops not football, some football numbers are worth almost nothing):

Buy any half point for 5 cents or less of additional vig.

never pay 10 cents or more for it.

points on sides are worth more than totals

use 7 cents for a quick estimate.

2 and 7 are worth more to get on/ off of, you might want to consider these worth 8-9 cents.

1 rarely hits in NBA, so on/off 1 is worth less.

[Ganchrow]

The following are ½ point NBA buy equivalencies for spreads from ½ to 10 based on the publicly available (but somewhat dubious) push frequencies here.

These are assuming a baseline of -105/-105 and equal implied juice on either side. Obviously the numbers will be a bit different as the associated money lines skew, if one side has more value than the other, or as the theoretical hold varies.

Value of buying a ½ point (-105/-105 baseline) on the underdog

spread	cents
½	 4.5
1	 4.4
1½	 8.6
2	 8.3
2½	 8.4
3	 8.1
3½	 5.4
4	 5.3
4½	12.5
5	11.8
5½	12.0
6	11.4
6½	 9.9
7	 9.5
7½	 8.1
8	 7.8
8½	 9.4
9	 9.0
9½	10.9
10	10.4

My own derived frequencies are in most cases a bit lower, and consequently so are most of my ½ point valuations, but they nevertheless still serve to illustrate that the conventional wisdom claiming that most spreads are worth about 7 cents is simply wrong.

[Doug]

Ganchrow: It seems I've seen many versions of the chart, and little agreement. Here's one that credits to a poster from SSB ( Computerpicks) it's a few years old. I basically use the simple guideline because you can find so little agreement.

What do you think of this guy's version ?

In case anybody missed it here is part of a handy chart I compiled on the historical value of half points in the NBA for games with a closing line of +4 to +8



Fav win by 8 7.5 7 6.5 6 5.5 5 4.5 4
11 4.13%
10 4.68% 4.35% 4.17%
9 4.13% 5.71% 3.92% 3.62% 3.77%
8 0.55% 7.61% 3.43% 5.66% 3.99% 2.69% 2.68%
7 5.23% 3.80% 5.15% 2.94% 4.66% 3.10% 4.74% 3.18% 6.30%
6 3.31% 3.80% 2.21% 3.17% 4.43% 4.75% 4.33% 3.38% 3.57%
5 3.58% 2.72% 4.17% 2.94% 3.99% 4.55% 4.12% 5.17% 4.41%
4 4.41% 3.85% 3.55% 4.34% 2.06% 2.58% 3.36%
3 4.43% 4.55% 2.89% 4.57% 5.25%
2 3.71% 3.58% 3.57%
1 2.31%


The columns are the spreads and the rows are the frequency each spread has landed on each number.
As an example if you were offered +7.5 -103 or +8 -110 then historically from the chart you would be far better off taking the +8 -110.
This chart is useful if you find a nice middle and would like to know the historical frequency that the number land.

[Doug]

Ganchrow: By your chart if the line was 5.5 as a consensus and 5.5 -105 at Pinny, and you could get +6, or -5 ( say both -110, and at different books) would that have long-term +EV ?

[Ganchrow]

Ganchrow: It seems I've seen many versions of the chart, and little agreement. Here's one that credits to a poster from SSB ( Computerpicks) it's a few years old. I basically use the simple guideline because you can find so little agreement.

What do you think of this guy's version ?

Unfortunately, vBulletin doesn't handle formatted plain text to well, but this is the same chart displayed with a fixed width font:

Fav                                  Closing Spread
Win   	    8	  7.5	    7	  6.5	    6	  5.5	    5	  4.5	    4 
By 
11	4.13%								 

10	4.68%	4.35%	4.17%						 

9	4.13%	5.71%	3.92%	3.62%	3.77%				 

8	0.55%	7.61%	3.43%	5.66%	3.99%	2.69%	2.68%		 

7	5.23%	3.80%	5.15%	2.94%	4.66%	3.10%	4.74%	3.18%	6.30% 

6	3.31%	3.80%	2.21%	3.17%	4.43%	4.75%	4.33%	3.38%	3.57% 

5	3.58%	2.72%	4.17%	2.94%	3.99%	4.55%	4.12%	5.17%	4.41% 

4			4.41%	3.85%	3.55%	4.34%	2.06%	2.58%	3.36% 

3					4.43%	4.55%	2.89%	4.57%	5.25% 

2							3.71%	3.58%	3.57% 

1									2.31%

For spreads of 4,5,6, and 7, the computerpicks push frequencies of 3.36%, 4.12%, 4.43%, and 5.15% respectively are at least the same order of magnitude as those available from the link posted. This implies ½ point buy prices (under the same assumptions I had indicated above) of 7.5c, 9.3c, 10.0c, and 11.7c.

For a spread of 8, however, computerpicks' figure of 0.55% is pretty far out of whack both with the posted frequencies as well as my own proprietary one. The real warning sign is apparent when comparing the 8 point victory frequencies of 8, 7½, and 7 point spreads. Does it really make sense to expect a fave to win by exactly 8 points only 0.55% of the time when the spread is 8, but 7.61% of the time when the spread is 7½? Are 7½ and 8 point spreads really so different?

I think this is indicative of computerpicks' data's major problem, namely that he's almost certainly used a small sample size to compile his results (and as a corollary, hasn't employed any smoothing techniques to clean up data anomalies). What we'd hope to see from well-behaved data would be frequencies that monotonically decreased when moving left to right across any particular row. In other words, we'd expect to see 8 point favorites win by exactly 8 points more frequently than 7½ point favorites, which in turn we'd expect to win by exactly 8 points more frequently than 7 point favorites, etc. (Obviously one could bring up that varying spreads might also be indicative of variations in team tenor, but over a broad sampling, and against a variety of different opponents all with their own variations in tenor we'd expect this to smooth out.)

Anyway, Stanford Wong details a simple methodology for predicting push (or ½ point win/loss) frequencies ATS in Sharp Sports Betting. It's unfortunate that the author of the chart I linked to above didn't utilize this when coming up with his figures. If you're interested in doing this yourself , this data can easily be obtained from a combination of Covers.com and databasebasketball.com. Although at some point maybe I'll just wind up put it together so I can finally stop posting that link to sportsbookfreepicks. Or who knows? Maybe one day I'll even get off my high horse and post own proprietary figures.

[Ganchrow]

Ganchrow: By your chart if the line was 5.5 as a consensus and 5.5 -105 at Pinny, and you could get +6, or -5 ( say both -110, and at different books) would that have long-term +EV ?

Well if the consensus line were ±5½ -105, then, according to the sportsbookfreepicks numbers, the fair price of +6 would be -117.0, and the fair price of -5 would be -117.5.

Based on computerfreepicks numbers of 4.55% when moving from the 5½ to the 5 and 4.75% when moving from the 5½ to the 6, the fair prices drop to -5 -115.8 and +6 -116.3.

So yes, both sets of numbers do imply a positive EV.

[pags11]

Doug, I agree with those statements...

[Doug]

Doug, I agree with those statements...

I was trying to make it simple for a quick comparison. I think Ganchrow goes over a lot of people's heads. I think his data can be rated as highly trustworthy, but complicated.