View New Posts
12
1. ## Correlated parlays in the NBA: fact or myth?

First let's look at paralays. A standard, two team, winning parlay at a reputable book pays 2.6 x each unit wagered. Also there are 4 possile outcomes (2^2). In the case of team A and team B hooked to the Over and Under, the 4 possibles are: team A/over, team A/under, team B/over, team B/under. Suppose from the book's standpoint that 10 units are wagered on each of the 4 possible outcomes, which equals a total wager pool of 40 units. Ignoring pushes, the book pays the winning player 26 units for the win plus returns the original 10 unit wager (10+26=36) resultng in a profit of 4 units for the book. The vig in this case is 4(profit)/40(total wagered) = 10%.

On the other hand most local books don't pay 2.6 but rather something like 6 pays 14. Following the above logic, wagering 6 units on each possible leg (6 to keep math in integers), the book books 24 units in action and returns 14+6=20 to the winning player. As before the profit to the book is 4 units but the vig is 4/24 (as opposed to 4/40) = 16.66%

As discussed in other threads for the player to win at 10% juice, he must cover ~53% of his plays. That is just barely breaking even. The actual calculaton for this is betting 110 to win 100 =
110(wager amt)/210(collected amt) = 52.38%.

Using the numbers for the reputable book above, the break even point =
10(wagered)/36(collected) = 27.77%.

Now looking at actual data from 2008-2009 NBA season. Tossing out games with
pushes, here is the breakdown of winning parlays in 1196 games:

300/1196=25.08% fav/ov
291/1196=24.33% fav/un
300/1196=25.08% dog/ov
305/1196=25.50% dog/un

and so it goes...

2. Every book in the world allows same game side/total parlay's in nba and for good reason.

3. To put it mildly, NBA is not the ideal medium for side/total CPs.
There are occasional gems but when they have arisen in the past, it has generally been the case than I have had a very strong lean on the total.

Good analysis, though.

4. You need to drill down into specific subsets. You won't find strong correlation using an aggregate database

5. Myth in NBA; Fact in CBB with low totals.

6. try the H favorites by 12 or more and the HD of 8 or more subsets.

7. Originally Posted by LT Profits
Myth in NBA; Fact in CBB with low totals.
Please define a 'low total' in CBB. Thanks.

8. Originally Posted by Peep
try the H favorites by 12 or more and the HD of 8 or more subsets.
So, the pars would be H fave by 12< and over, and H dog of 8< and under?

9. Originally Posted by Munson15
So, the pars would be H fave by 12< and over, and H dog of 8< and under?
Well, that if for you to figure out.

You said you were interested in places to look. I told you a place to look.

10. Originally Posted by Munson15
Please define a 'low total' in CBB. Thanks.
Good to see you Munson.

11. Originally Posted by Peep
try the H favorites by 12 or more and the HD of 8 or more subsets.
Any chance some of the other database guys throw us a bone on this?

12. I could run it, over 12K games, as could a bunch of other guys here.

And come up with a number. But even that number is a "so what". The guy who started this thread really wants to learn how to look for his own favorable situations. I think that was the point of his original post.

"Feed a man a fish, teach him to fish, all that jazz".

If no one else will run it, I will. And we'll see what it says. I don't think there will be much there personally, but the two extremes are the place to look for this sort of stuff.

13. Nothing much in the favorites by 13 or more except doggies are good plays in general.

1) 533 games in total.
2) 243 covered.
3) 271 did not cover.
4) 19 pushed.

Of the 243 that covered.
1) 116 went over
2) 122 went under
3) 5 pushed.

Of the 271 that did not cover
1) 126 went over
2) 144 went under
3) 1 pushed.

There you go Tim, a bone at least. Maybe a thin one, but still a bone.

14. Originally Posted by Peep
Nothing much in the favorites by 13 or more except doggies are good plays in general.

1) 533 games in total.
2) 243 covered.
3) 271 did not cover.
4) 19 pushed.

Of the 243 that covered.
1) 116 went over
2) 122 went under
3) 5 pushed.

Of the 271 that did not cover
1) 126 went over
2) 144 went under
3) 1 pushed.

There you go Tim, a bone at least. Maybe a thin one, but still a bone.
Appreciate it my man

15. Originally Posted by arwar
300/1196=25.08% fav/ov
291/1196=24.33% fav/un
300/1196=25.08% dog/ov
305/1196=25.50% dog/un
I don't really see what these numbers really demonstrate, except that blindly betting anything is a losing proposition...

16. Exactly. People look for every angle in the book, but 85% of them still lose. If the truth be known in here, that figure would probably be a bit lower, but not by much. You will never see a sharp busting his gut trying to figure these numbers. They have more important things to do, like findling teams and fading the very poeple who like to play this system.

17. Originally Posted by OSUCOWBOYS
Good to see you Munson.
Thanks, OSU, this think tank has a little more action.

18. Originally Posted by Peep
Well, that if for you to figure out.

You said you were interested in places to look. I told you a place to look.
I used 12 point faves, not 13, and since my database skills are lacking, I simply tracked all results from this season. I came up with 22 winning parlays and 50 losers, for a net of +7.2 units. The way I used them was to parlay 12

19. No, I f***ed that up. No 30 units there. Also wasn't sure whether it would be correlated in the football sense where you play fave/over AND dog/under, or if one or the other was best. I would think dogs/unders would be the way to go since more dogs hit. I already play all 10

20. Originally Posted by Munson15
Please define a 'low total' in CBB. Thanks.
No.

21. Originally Posted by Munson15
Please define a 'low total' in CBB. Thanks.
Key is actually the ratio of spread/total, and it is easier to achieve +EV correlation as totals get lower.

22. Originally Posted by LT Profits
Key is actually the ratio of spread/total, and it is easier to achieve +EV correlation as totals get lower.
Would 10% correlation be a good place to start? Say, 12 point spread and 120 for a total?

23. Originally Posted by LT Profits
No.

24. Originally Posted by LT Profits
Key is actually the ratio of spread/total, and it is easier to achieve +EV correlation as totals get lower.

Exactly. People look for every angle in the book, but 85% of them still lose. If the truth be known in here, that figure would probably be a bit lower, but not by much. You will never see a sharp busting his gut trying to figure these numbers. They have more important things to do, like findling teams and fading the very poeple who like to play this system.

26. Originally Posted by Munson15
Would 10% correlation be a good place to start? Say, 12 point spread and 120 for a total?
Probably not even close to good enough to make it profitable. In football the correlation needed is around 30%

27. Originally Posted by Igetp2s
Probably not even close to good enough to make it profitable. In football the correlation needed is around 30%
I use 33% in football, but you'll never get that % in hoops. I'm looking for a corresponding % in hoops that would work.

28. Posters talking about total/spread ratio have a good point. Obviously lower totals are easier to work with, better ponds to fish in IMO.

So you have subsets within the subsets I outlined to look at as well.

BUT....

By that time your data becomes less, you are at high risk for backfitting. And that ain't someplace you want to go.

29. Just looked at some games, and I'm figuring the spread being 1/6 of the total would be a minimum for NCAAB.

30. Went back as far as Jan. 1st, and found only one game qualified. And, no parlay hit on that game. I feel the NCAA is a dead-end, but the NBA idea still has hope.

31. Hint

1st halfs

32. Originally Posted by Pancho sanza
Hint

1st halfs
College or NBA?

33. Completely different game dynamics in basketball.

34. Nice write up

35. Originally Posted by Munson15
College or NBA?
College

12 Last
Top