[JayDee]

I'm testing a new hockey system and I want to see how much I make/lose by playing Moneyline bets instead of 1X2 odds. I only have the historical 1X2 odds (the highest offered for each choice at the time) and I thought that I possibly could make an fair assumption regarding the offered ML odds for the same game.

I was thinking something along these lines:
Game :Toronto - Pittsburgh
Bookie offered odds: 1: 2.29, X: 3.97, 2: 2.45

Convert the bookie odds into probablities:
1: (1/2.29) = 43.67 %
X: (1/3.97) = 25.19 %
2: (1/2.45) = 40.82 %

Since the odds are from different places, this doesn't add upp to 100 % probablity.

Still, based on the bookies that offered the 1 and 2 choices, converted to moneyline probalities that would be:
Toronto win = 0.4366 / (0.4366 + 0.4082) = 0.5169 = 51.68 %
Pittsburgh win = 1 - 0.5169 = 48.32 %

Convered to moneyline odds:
Toronto win = 1 / 0.5169 = 1,93
Pittsburtgh win = 1 / 0.4832 = 2,07

But, taking to account that the odds doesn't reflect the true probablity by the bookies and to reflect that I won't always will get the best odds possible, I'm thinking about putting a discount on the odds offered.

So, when calculating what odds I get offered I'm thinking about playing the odds calculated as follows:
1.93 * 0.9 = 1.74
2.07 * 0.9 = 1.86

What do you guys think? Am I on the right track? Would you be comfortable with using that kind of odds input when backtesting your systems?

Thanks for your input!

/JayDee

[FourLengthsClear]

On the basis that you are trying to work out the advantage (or otherwise) of betting ML as opposed to 12X, making assumptions about the ML odds is flawed.

You need to find the actual data.

[JayDee]

You are absolutely correct. But for the sake of argument: let's assume that we don't want to play 1X2 odds to eliminate the chance/risk of draws and don't have access to ML odds. Would you trust my assumption regarding ML odds? :-)

[FourLengthsClear]

The first stage (getting the no vig probabilities) is OK.

Simply taking 90% of the decimal odds is wrong, though. I would calculate at something around -106 lines as a worst case (-110 is the worst you will find). -106 e/s equates 'only' to 2.91% juice.

As such for Toronto you would calculate the odds as follows:

0.5169 x 1.0291 = 0.5391
or 1.8799 in decimal form.

[RickySteve]

Nothing inherently wrong with deriving ML probability from 1X2 primary market. The hypothesis is fine.

In practice I think you'll find the draw is rarely represented accurately by the implied odds. Also I hope you realize these MLs are not inclusive of overtime and few US-facing books have 60 min MLs.