[HUY]

How do you guys convert bookmaker moneylines to no-juice moneylines? Before people say that they simply multiply the odds for each event by the single number that will make the line juice-free, let me just say that this strategy is too naive, as the strongest the favorite the less juice it has. Here is the function I tend to use (for decimal odds, 2.5 % edge, 3-outcome events, e.g. pinnacle soccer). I have found though that it tends to assume that the favorites have less juice than they really have.

[HeeeHAWWWW]

Take past results, calculate juice in different odds bands, and make a formula. For instance, in tennis most of the juice is on the dogs (although less so with closers).

There will be all sorts of oddities unique to each sport though. In tennis there's a considerably difference between 3 and 5 setters, in football I'd imagine national differences, and in the US sports perhaps college vs pro might have quite a split.

[allin1]

you have 3 way ml: 2.6, 3.1, 3

1/2.6 + 1/3.1 + 1/3 = 1.0405

2.6 x 1.0405 = 2.7
3.1 x 1.0405 = 3.22
3 x 1.0405 = 3.12

I don't know if this is what you want, and I don't know if it's accurate.

[JoeyBagels]

http://sportsbettingsites.org/bettin...ig-calculator/

[HUY]

you have 3 way ml: 2.6, 3.1, 3

1/2.6 + 1/3.1 + 1/3 = 1.0405

2.6 x 1.0405 = 2.7
3.1 x 1.0405 = 3.22
3 x 1.0405 = 3.12

I don't know if this is what you want, and I don't know if it's accurate.

It's not accurate, as it adds the same amount of juice to each selection. This is exactly what I warned against in my original post.

[allin1]

ok. now I understand. interesting

[chunk]

No correct answer here.

[EXhoosier10]

Take past results, calculate juice in different odds bands, and make a formula. For instance, in tennis most of the juice is on the dogs (although less so with closers).

There will be all sorts of oddities unique to each sport though. In tennis there's a considerably difference between 3 and 5 setters, in football I'd imagine national differences, and in the US sports perhaps college vs pro might have quite a split.

This is probably the way to do it, but you'll most likely run into SSS issues in US sports (sans baseball, probably) unless you want to go back 5+ years.

[HeeeHAWWWW]

This is probably the way to do it, but you'll most likely run into SSS issues in US sports (sans baseball, probably) unless you want to go back 5+ years.

I'd assume so, yes. I've only done it for tennis, which has a tidy 2800ish matches a year (2300 if you exclude those troublesome 5-setters).

The major issues for me are more ....

a) the tradeoff between band size and statistical validity, not an uncommon one to face.
b) various difficulties all caused by inconsistent variance between the bands (heteroscedasticity)

[HUY]

I'd assume so, yes. I've only done it for tennis, which has a tidy 2800ish matches a year (2300 if you exclude those troublesome 5-setters).

The major issues for me are more ....

a) the tradeoff between band size and statistical validity, not an uncommon one to face.
b) various difficulties all caused by inconsistent variance between the bands (heteroscedasticity)

You're doing it the wrong way. The correct way is to record bookmaker odds and compare them against betfair odds (which have no juice). I haven't done this because I can't collect betfair odds.

[satana]

http://www.thinkbetprofit.com/forums...ad.php?t=22881 ??

[HUY]

http://www.thinkbetprofit.com/forums...ad.php?t=22881 ??

Thanks for this link! Interestingly they are discussing the exact same thing, but they don't have an answer. I still think that the function in my original post is the best presented so far.

[HeeeHAWWWW]

You're doing it the wrong way. The correct way is to record bookmaker odds and compare them against betfair odds (which have no juice).

How so? That's just comparing one set of implied probabilities with another (and at least in tennis, pinnacle's closers are historically more accurate than betfair).

I'm comparing implied probabilities with actual results, split into various odds ranges. For example, if you bet every pinnacle opener between 1.5 and 1.6 for the last ten seasons, excluding slams, the result would be 1439 wins from 2243 matches (64.2%) - a tiny loss overall. This yield actually holds pretty much flat all the way from 1.0 to 1.6 (10488 of 13846), above which the juice starts to increase. At 3.0+, you're paying an average 10% juice.

Presumably pinnacle set out their juice this way because it optimises profit (most balanced action? attains sharpest line most quickly?).

[HUY]

How so? That's just comparing one set of implied probabilities with another (and at least in tennis, pinnacle's closers are historically more accurate than betfair).

But pinnacle lines include juice, which means that their accuracy depends on the way you choose to calculate the no-juice line. So I can't see how one can measure pinnacle lines' accuracy. It's just not possible. Of course, even in betfair there is the back/lay spread, but for big markets this is negligible.

I'm comparing implied probabilities with actual results, split into various odds ranges. For example, if you bet every pinnacle opener between 1.5 and 1.6 for the last ten seasons, excluding slams, the result would be 1439 wins from 2243 matches (64.2%) - a tiny loss overall. This yield actually holds pretty much flat all the way from 1.0 to 1.6 (10488 of 13846), above which the juice starts to increase. At 3.0+, you're paying an average 10% juice.

Can you post the complete set of numbers?

Presumably pinnacle set out their juice this way because it optimises profit (most balanced action? attains sharpest line most quickly?).

There are many scientific articles on the favorite-longshot bias, each one suggesting different explanations. My personal best guess is that favorites' prices are raised due to competition with other bookmakers, since most people like to play the favorites, especially on parlays.

[buby74]

To correct for fls I use (k^d)/d to work out the true winning %. Where d is the decimal odds so 3/1 or +300 gives decimal odds of 4.0. K is a constant for each sport/bookie/juice level which you have to calculate by looking at actual odds. For uk horse racing k is 0.975 iirc. The effect of fls is quite small in two team/player sports unless there is a big difference in ability.

[bihon]

best guess is that favorites' prices are raised due to competition with other bookmakers, since most people like to play the favorites, especially on parlays.

Probably true.

O.T.:
Just out od curiosity, I wonder if you can accurately calculate no juice line what possible edge one can have?
Even if you assume no juice on favorite, what can you do with it?

[HUY]

Probably true.

O.T.:
Just out od curiosity, I wonder if you can accurately calculate no juice line what possible edge one can have?
Even if you assume no juice on favorite, what can you do with it?

Pinnacle lean.

[satana]

To correct for fls I use (k^d)/d to work out the true winning %. Where d is the decimal odds so 3/1 or +300 gives decimal odds of 4.0. K is a constant for each sport/bookie/juice level which you have to calculate by looking at actual odds. For uk horse racing k is 0.975 iirc. The effect of fls is quite small in two team/player sports unless there is a big difference in ability.

what's k for tennis? what doeas fls stand for?

[Unknown User]

"that favorites' prices are raised due to competition"

This is a very interesting observation. By prices being raised you mean that the payout/odds are raised on the favorite right?

That makes guessing where the vig is, a lot harder :S

[gui_m_p]

"that favorites' prices are raised due to competition"

This is a very interesting observation. By prices being raised you mean that the payout/odds are raised on the favorite right?

That makes guessing where the vig is, a lot harder :S

I don't think the statement is always true. I rather think the opposite regarding recreational books. In tennis, for example, recreational books have very low odds on favorites; 1st rounds of grand slams is common to see 1.005 odds and the dog with higher odds. They know the public will bet on favorites, so they charge the juice there.

[HeeeHAWWWW]

I don't think the statement is always true. I rather think the opposite regarding recreational books. In tennis, for example, recreational books have very low odds on favorites; 1st rounds of grand slams is common to see 1.005 odds and the dog with higher odds. They know the public will bet on favorites, so they charge the juice there.

That's probably just a natural result of the extremely dominant results of the top guys. When you're charging 8% juice, any fav much above 90% is rapidly tending towards a 1.0000000000000000 "fair" price. With some rec books I've seen them just not offering that side of the moneyline at all.

I haven't run the numbers on rec books, but a couple of years ago I checked out returns of various odds ranges at pinnacle. At slams, you would actually make a very small profit betting all short odds (<1.25 iirc). In best of three, there's a clear fls distribution of returns though.

[arwar]

i admit i am confused here. there are formulas for converting point spreads to money lines and money lines to decimal probabilities. at pinnacle for example the overload (juice) works out to 1.024 which is one of the main reasons they are considered a reasonable benchmark. I am not sure about the 3 way plays on soccer or hockey or even tennis, though i have worked with all of these. i guess you are talking about the juice on individual plays. obviously if a favorite (which say gets 75% of the action) loses, then more juice is on the chalk. but on a -110 play on both sides it seems to me the juice is the same on either side. please explain in more detail

[Unknown User]

In best of three, there's a clear fls distribution of returns though.

What does "fls" stand for?

[HeeeHAWWWW]

In best of three, there's a clear fls distribution of returns though.

What does "fls" stand for?

Sorry ..... it's favourite-longshot. Basically pretty much any book pricing always puts more juice on the bigger dogs, and conversely little to no juice on the big favourites.

Lazy quick example: say you have a two-way moneyline, and the books want to offer 2/3 one side, 1/3 the other, at a total overround of 5%. The favourite fair price is thus 1.50 (aka -200). That will maybe get juiced at 2%, so they will offer 1.50*0.98 = 1.47. The other side will thus be 1/ (1.05 - (1/1.47)) = 2.705, against a fair price of 3.00, juice of 9.8%. Exact %s were pulled out of my ass, but you get the general idea.


You can see this in action in pretty much any market you care to look at. For example, pinnacle's average overround in tennis moneylines over the last decade has been 102.25%, but if you bet all odds at 10ish (+900) you'll end up with an RoI in the ballpark of -30%.

[Unknown User]

Cool. I still don't understand how they get away with that and get their books somewhat in balance but it's important to know.