January 20, 2017, 10:01:24 PM
News: If it is not in arbusers, it doesn't exist.

Value on the long side

Pages: [1]
« previous next »
Author Topic: Value on the long side  (Read 715 times)
Gamblers Ruin
Newbie
*

Karma: -1
Posts: 29


View Profile
« on: September 28, 2016, 12:58:44 PM »

Hello fellow arbers!

As with many I've been quietly watching this forum for months now after realising I wasn't the only one in the world doing this  :)

After reading about value bets I decided to cut the data from all my arbs any way I could to look for other ways of finding the value, and interestingly the "everyone bets on the favourite" theory is the one that stands out as surprisingly accurate. While odds of under-2.0 return on average only 99% of what the decimal would imply they should, my over-2.0 section averages out to nearly 105% (and I have 102.8% in total, which is reassuring!). Note this only covers ~1000 bets so far.

The clear implication, if this becomes statistically significant, is that as well as not-betting-pinny we can also not-bet-the-favourite for the same effect.

Obviously other users on here will bet different markets catering to different demographics of punters etc etc, but since I'm sure some of you have much larger datasets to interrogate, I would be very interested to see if anyone else has seen the same trend.

(if this has already been discussed and I missed it, please feel free to delete/move this post)

Thanks,
Logged

Gamblers Ruin
Newbie
*

Karma: -1
Posts: 29


View Profile
« Reply #1 on: November 02, 2016, 09:21:08 AM »

In case anyone actually read this, after only a further ~200 arbs this disparity has all but disappeared. My profit is still slightly on the side of the underdog, but it's a lot closer: 2.9% vs 2.4%.

I clearly need a much larger sample size, I guess I'll be back with an update in about a year  :)
Logged

dealer wins
Probably a Pro
****

Karma: -16
Posts: 351



View Profile
« Reply #2 on: November 02, 2016, 10:16:12 AM »

It shows that to get an accurate sample takes 1000s and 1000s of trials.   
Logged

Never trust a goose!!!

maletaja
Has some experience
**

Karma: -26
Posts: 169


View Profile
« Reply #3 on: November 02, 2016, 10:55:20 AM »

favourites smaller variation, underdogs bigger variation

1000 sample you cannot take any assumptions, try million bet.,..
And even dogs have some tiltied margin, why you have to deal with your head. U still bet heavy favourites because its+ev all that  matters
Logged

dejected
Newbie
*

Karma: -8
Posts: 54


View Profile
« Reply #4 on: November 02, 2016, 11:46:30 AM »

It shows that to get an accurate sample takes 1000s and 1000s of trials.   

You can use statistics to determine if your results are just luck, or you actually have an edge. I have recently added the calculation of p-value to the statistics I have of various tipsters. You can do the same calculation for your own value bets. When the p-value gets low enough you have enough data. Although I think for the pini example of how to calculate it only works for level stakes.

 https://www.pinnacle.com/en/betting-articles/betting-strategy/how-good-are-betting-tipsters
Logged

my_username
Has some experience
**

Karma: -3
Posts: 102


View Profile
« Reply #5 on: November 02, 2016, 11:50:45 AM »

Here's the thing about statistics... it only shows what already happened.

Logged

dejected
Newbie
*

Karma: -8
Posts: 54


View Profile
« Reply #6 on: November 02, 2016, 01:43:14 PM »

Here's the thing about statistics... it only shows what already happened.

True, but what you can do is compare the ROI from your bets to what would be achieved if the selections were random. It is possible to calculate the percentage chance of your real ROI occurring if you just made the same number of random selections with the same average price.  This is what the p-value does. If you have a low p-value it is unlikely that your ROI or better could be achieved through luck. A p-value less than 5 (1 in 20 chance) is normally said to be statistically significant. So if your number of selections, ROI and average price gives you a value grater than 5 you need more selections before you can start to draw conclusions about your results and return.
Logged

Bubbles
Getting better
***

Karma: 6
Posts: 218


View Profile
« Reply #7 on: November 02, 2016, 04:38:02 PM »

Here's the thing about statistics... it only shows what already happened.
I would say it's the only thing we have, and a powerful one, that can help to take a peek into the future. If I were younger I would take an extra degree in econometrics in Sheffield like Coates. Now it's only Oldgit for me :(
Logged

Pages: [1]
Print
« previous next »