[Justin7]

[bigboydan]

Good breakdown Justin.

[jjgold]

Solid info, it is nice with these vids that we can go back to them by clicking his name on sbr.tv


The more I look at Justin I look similar to him with my rug on.

[SBR Lou]

Another excellent video, thanks for sharing Justin.

[onthewhat]

another great vid

[SlickFazzer]

Solid info.

[HedgeHog]

Good basic info. But at what point does correlation become relevant. For example, is an NFL game Line/Total of 14/38 significantly correlated for a fav/OV parlay (or dog/Un parlay)? Perhaps a video on advanced correlated parlay analysis is needed.

[Justin7]

Perhaps a video on advanced correlated parlay analysis is needed.

Yes, but I'd get mugged if I did it. It's one thing to point you in the right direction so you can figure it out yourself. It's quite another to kill a golden goose.

[acw]

Justin7,

http://forums.eog.com/online-sportsb...s-players.html

Next video on how you really can beat the bookies, please!

[MrX]

Yes, but I'd get mugged if I did it. It's one thing to point you in the right direction so you can figure it out yourself. It's quite another to kill a golden goose.

Exactly. There is plenty of information given (and well explained, IMO) in the video for a motivated person to find their own way.

[Doc JS]

Justin,
Thanks for the great videos!!

Doc

[big joe 1212]

Good basic info. But at what point does correlation become relevant. For example, is an NFL game Line/Total of 14/38 significantly correlated for a fav/OV parlay (or dog/Un parlay)? Perhaps a video on advanced correlated parlay analysis is needed.

Books have caught up with this and wont let you parlay correlations in same game.

However, find a local bookie and they usually are pretty dumb to this!

Use to hammer my guy in College football. Good corr. are teams like USC and over. Or big dogs and the under.

[HedgeHog]

All of my Books disallow same game pars for college games where there is fav of 14 or more. I can parlay just about any NFL game, regardless of line/total size at some of my books. Overall though, for on-line players at least, correlated action is severely limited from just a year or two ago.

[ldrapeau]

Hi Justin,

Very interesting, thank you very much, this helped me a lot in understanding correlation and I will sure keep my eye open, thanks again for the great tip.

I don't know if it's ok to go with a question here, but I'm going to try. We have 50 different matches going on (for the sake of the question team 1 against team 2, team 3 vs team 4, team 5 vs team 6, ..., so we have 100 teams).

Now, I plan to organize a contest with an expected 150 participants, each participant must pick 3 teams he think will win, my question is what are the odds that two contestants pick the same 3 teams. As you said in your video, if you can explain this instead of just giving the number I would appreciate it very much, this will help me in getting my math a bit better. Thanks, LD

[Justin7]

Each person selects 3 out of 50 games. This is 50 * 49 * 48 / 3 / 2 / 1. He has 8 ways to play that - so there are 50 * 49 * 48 * 8 / 6 total ways a player could make his picks. This is 156,000.

The odds of any two people having the exact same selection is 1/156000.

The odds of the first two having different selections is 155999 / 156000.

If the first two are different, the odds that the third is also different is 155998 / 156000.

For N people, the odds of no duplicates are the product of:
(156000-k)/(156000)
For k=1 to N.

Does that help?

[I.R.B]

Great Vid Justin

[Ganchrow]

For N people, the odds of no duplicates are the product of:
(156000-k)/(156000)
For k=1 to N

Above should read:

(156800-k)/(156800)
For k=1 to N-1

[smitch124]

I'm not sure of the nature of the contest, but this percentage is obviously if the odds of picking each game were equal. In these contests people tend to pick the more visible games. For example during the week of the USC-Ohio St. game, I'm sure one saw a much higher percentage of people using this game in a contest than what one would expect by random selection.

[ldrapeau]

Thanks Justin,

I have to continue to spend some time on the first part to make sure I catch everything, but as for the second part, I wrote this little program that produced the following;

ps = 50 * 49 * 48 * 8 / 6 * 1.0

N = 150

p = 1.0

for k in range(1,N+1):

p = p * (ps - k) / ps
if k % 10 == 0: print '%03d %1.10f' % (k, p)

010 0.9996492884
020 0.9986615524
030 0.9970386400
040 0.9947836107
050 0.9919007267
060 0.9883954387
070 0.9842743695
080 0.9795452924
090 0.9742171074
100 0.9682998128
110 0.9618044742
120 0.9547431900
130 0.9471290538
140 0.9389761138
150 0.9302993301

So the answer would be 7% chances of having 2 out of 150 participants chose the same selection, or about 1/14 (1/0.07) does this sound OK, am I interpreting these numbers correctly I'm not too confident in my math at this point? Thanks again for the explanations, LD

smitch124: agreed 100% - I'm just looking to get my math ok, then I will factor in the "popularity factor", your observation is very correct, thanks.

[Ganchrow]

for k in range(1,N+1):

p = p * (ps - k) / ps
if k % 10 == 0: print '%03d %1.10f' % (k, p)

For the 150th item on your list to reflect the correct answer the above should read:

for k in range (1,N-1):

p = p * (ps - k) / ps
if (k+1) % 10 == 0: print '%03d %1.10f' % (k+1, p)

Without the above correction, the solution to the problem, (~ 93.12%), would correspond to item #151 on your list.

So the answer would be 7% chances of having 2 out of 150 participants chose the same selection, or about 1/14 (1/0.07) does this sound OK, am I interpreting these numbers correctly I'm not too confident in my math at this point?

There is a ~ 1-93.12% = 6.88% probability at least 2 participants making identical picks.

[ldrapeau]

Thanks Ganchrow for those precisions, and is my conversion of 7% to 1/14 ok or am I in the fog with that? Thx, LD

[Ganchrow]

Is my conversion of 7% to 1/14 ok or am I in the fog with that?

7% ≈ 1 14.29 (and 6.88% ≈ 1 14.53 ), so yes 1 14 is quite close.

[fiveteamer]

Ganchrow = Will Hunting.

[Lippsman]

I will usually lay off the correlated parlays where the OU is 60 or more. Just too much room for the fave to win 45 or so to nothing. I have seen it way too many times this year. I like to stick to 45% or higher.

Great video

[reno cool]

Good job.

I never considered the example you gave at the end as a correlated parlay(one team loses the other likely to rest players). that type of thinking I'm guessing might still be ahead of the books.

[smitch124]

Good job.

I never considered the example you gave at the end as a correlated parlay(one team loses the other likely to rest players). that type of thinking I'm guessing might still be ahead of the books.

Thinking back on it I can remember some instances where one team lost missed the playoffs and the 2nd team that now clinched got completely trounced...

Used to happen with the 9ers alot when they used to sniff the playoffs.

[pico]

good video

[Dark Horse]

Clear and instructive as always, but this time a little too general; not a whole lot of meat.