[Sportslover]

I have a spreadsheet that calculates Moneylines for Over/Under bets increments of 0.5. There is a table starting at 0.0 and goes up by 0.5 for each line of the table and gives an estimated winning percentage and fair moneyline for each line. Then on the side there is a section where you put in the "expected total" and put in your desired EV or edge, once you've put in those details it completes the necessary calculations.

My question is this:

If I am betting a line with an Asian Handicap in Soccer of Over/Under 2.5 and 3 Goals, effectively a Over/Under 2.75 Line, how would I use the Poisson Table to get the Fair Moneyline price for this outcome?

I tried to make a new Poisson Table that went up in increments of 0.25 instead of 0.5. However, this just yielded the same Fair Moneyline price as the Over/Under 2.5 Line. That is because the spreadsheet doesn't recognise that the 2.75 Line represents half your stake on the 2.5 Line and half your stake on the 3.0 Line.

Would I simply take the Fair Moneyline for Over/Under 2.5, the Fair Moneyline for Over/Under 3, convert the American odds to Decimal Odds, add them together and divide the odds by 2 to get an average odds? Would this method give me the Fair Moneyline price for an Over/Under 2.75 Asian Handicap Line in Soccer? Or would I need to use another method to determine the Fair Moneyline price for this event?
Any insight into this subject would be much appreciated, thanks.

[Justin7]

Poisson is not very good for soccer.

If you wanted to use the tool you tried, you need to be mindful of push percentages when averaging. A lot of equity can be gained by making bets with a huge push chance, but a small advantage on your play.

[Sportslover]

Poisson is not very good for soccer.

If you wanted to use the tool you tried, you need to be mindful of push percentages when averaging. A lot of equity can be gained by making bets with a huge push chance, but a small advantage on your play.

The spreadsheet also calculates the percentage chance of a push.

For example, if we look at a Swedish Superettan game for today, Orgryte vs Jonkopings, let's say we want to investigate a line of Over/Under 2.5 and Over/Under 3.0 and compare Fair Moneyline prices.

I enter 2.964 for the "expected goals" and enter an EV of 2.0%.

Over/Under 2.5: Under - 43.1% Over 56.9%
Fair Moneyline: Under - (+136) Over - (-126)

Over/Under 3.0: Under - 43.1% Push - 22.4% Over - 34.5%
Fair Moneyline: Under - (-118) Over - (+131)

So it does take into account the chance of a "push". Do those odds look accurate to you? Or has it over/underestimated the "push" when determining the "fair price"?

Also Justin, do you have any thoughts on how to get a Fair Moneyline for Over/Under 2.5 and 3.0 bets (or Over/Under 2.75 if you like)? Is it simply the average Fair Moneyline of 2.5 and 3.0? Thanks for any insight into this!

[Justin7]

Sports,

If you want to attack soccer totals, there are a couple concepts you might want to look at. First, how are scores distributed? They don't really follow a Poisson distribution. Second,
how does the frequency of pushes affects pricing. You need to screw around with a spreadsheet to get a feel for this. Most people underestimate how much a push affects the prices.

[Sportslover]

Sports,

If you want to attack soccer totals, there are a couple concepts you might want to look at. First, how are scores distributed? They don't really follow a Poisson distribution. Second,
how does the frequency of pushes affects pricing. You need to screw around with a spreadsheet to get a feel for this. Most people underestimate how much a push affects the prices.

When you are talking about the distribution of scores are you talking about across a single game, referring to the timing of each of the goals? For example, average first goals scored by team x is approx 39.5 minutes? Or am I thinking of something different to what you were referring?

And would you be able to offer any insight into these Over/Under 2.75 or Over/Under 2.5 and 3.0 Lines?

[Justin7]

When you are talking about the distribution of scores are you talking about across a single game, referring to the timing of each of the goals? For example, average first goals scored by team x is approx 39.5 minutes? Or am I thinking of something different to what you were referring?

And would you be able to offer any insight into these Over/Under 2.75 or Over/Under 2.5 and 3.0 Lines?

I'm talking about a season-long distribution.

Consider a game between X and Y, at team Y's pitch. You need to estimate the odds of the game landing on each score to evaluate these bets.

[Sportslover]

I'm talking about a season-long distribution.

Consider a game between X and Y, at team Y's pitch. You need to estimate the odds of the game landing on each score to evaluate these bets.

Ok thanks I understand what you mean here now.

Any thoughts on the Over/Under 2.75 Asian Handicap Lines though? Could you offer some insight into whether or not simply averaging the odds of 2.5 and 3.0 would work?

[Data]

Would I simply take the Fair Moneyline for Over/Under 2.5, the Fair Moneyline for Over/Under 3, convert the American odds to Decimal Odds, add them together and divide the odds by 2 to get an average odds? Would this method give me the Fair Moneyline price for an Over/Under 2.75 Asian Handicap Line in Soccer? Or would I need to use another method to determine the Fair Moneyline price for this event?

No, this will not give you the Fair Moneyline. Calculating the Fair Moneyline requires calculating the push probabilities. However, if the odds are close to even for O/U 2.75, use 0.475 as a multiplier instead of dividing by 2 and you will not be too far off comparing to the rest of your table.

[Sportslover]

No, this will not give you the Fair Moneyline. Calculating the Fair Moneyline requires calculating the push probabilities. However, if the odds are close to even for O/U 2.75, use 0.475 as a multiplier instead of dividing by 2 and you will not be too far off comparing to the rest of your table.

Ok, if I estimate the push probability of O/U 2.75 to be 10.7% would I use 0.447 as the multiplier? Have I understood this correctly or am I way off track here?

Appreciate your help here.

[Sportslover]

Ok, if I estimate the push probability of O/U 2.75 to be 10.7% would I use 0.447 as the multiplier? Have I understood this correctly or am I way off track here? Appreciate your help here.

If I am on the right track here then I now understand what Justin7 was saying earlier about taking into account push probabilities when determining "fair price".

Sorry, it's only just clicked into my head now, Justin, after Data started talking about the 0.475 multiplier. Justin, you must have thought you were talking to a brick wall earlier but I know what you meant now!

[Data]

Over/Under 2.5: Under - 43.1% Over 56.9%
Fair Moneyline: Under - (+136) Over - (-126)

Fair Moneyline +/- 132

Over/Under 3.0: Under - 43.1% Push - 22.4% Over - 34.5%
Fair Moneyline: Under - (-118) Over - (+131)

Subtract from 1 half of the push probability for 3:
1-0.112=0.888

Multiply it by U2.5 decimal ml, 2.36:
0.888*2.36=2.08

2.08 is the ml for U2.75.

[Data]

use 0.475 as a multiplier instead of dividing by 2

Actually, in many cases dividing by two will be better but this is out of the scope.

[Sportslover]

Fair Moneyline +/- 132 Subtract from 1 half of the push probability for 3: 1-0.112=0.888 Multiply it by U2.5 decimal ml, 2.36: 0.888*2.36=2.08 2.08 is the ml for U2.75.

Thankyou so much for explaining this. Very much appreciated.

[Sportslover]

Fair Moneyline +/- 132 Subtract from 1 half of the push probability for 3: 1-0.112=0.888 Multiply it by U2.5 decimal ml, 2.36: 0.888*2.36=2.08 2.08 is the ml for U2.75.

Would 1.59 then be the ML for Over 2.75? Seems like it doesn't quite add up, have I misunderstood something here?

[Data]

Would 1.59 then be the ML for Over 2.75? Seems like it doesn't quite add up, have I misunderstood something here?

No. If decimal odds for the Under are 2.08 then the American line is +108 and the line for the Over is -108.

[Ganchrow]

Multiply it by U2.5 decimal ml, 2.36:
0.888*2.36=2.08

2.08 is the ml for U2.75.

If the OP were looking for fair odds on the Asian Handicap then I assume you meant to write:

0.888*=

is the fair ml for U2.75.

using fair U2.5 odds.

Or perhaps I've misunderstood.

[MonkeyF0cker]

Yeah. I was just about to say that, Ganch.

[Data]

Yes,

Fair Moneyline +/- 132

Thank you

[Sportslover]

No. If decimal odds for the Under are 2.08 then the American line is +108 and the line for the Over is -108.

If the OP were looking for fair odds on the Asian Handicap then I assume you meant to write:

0.888*= is the fair ml for U2.75.

using fair U2.5 odds. Or perhaps I've misunderstood.

Thanks a lot guys. Appreciate your help.

[robotzel78]

Hi Guys. I search on google about this problem, using Poisson Distribution for Asian Under Over, and i arrived here.

I don't understand exactly about ML, so i want to put my problem here, and i hope you understand me.

So, we have this game:

Vallerenga - Sarpsborg, with these odds:

Over /Under 2.75: 1.84 2.06
Over/Under 2.5: 1.47 2.45

The Average Goals is: Vallerenga 2.43 Sarpsborg 1.09. These give me the total goals for this game at 3.52.
Now, Poisson estimate me for Fair % for Over/Under 2.5 at 47% for Over 2.5 and 53% for Under 2.5. So, i have Value for Under 2.5
But, how can i calculate the Fair% for Over/ Under 2.75, or for Over/Under 3 ???

This is my question, about Poisson for Asian Over Under Lines ...

Thanks in advance for you answer.

[Dash2in1]

If the average goals is 3.52 and you use a poisson distribution then:

Probability under 3 goals = Punder2_5 = 0.3172
Probability under 4 goals = Punder3_5 = 0.5323

Probability over 2 goals = Pover2_5 = 0.6828
Probability over 3 goals = Pover3_5 = 0.4677

Probability on 3 goals = Pgoals_3 = 0.2152

So actual odds would be
Oover2_5 = 1/Pover2_5 = 1/0.6828 = 1.46
Ounder2_5 = 1/Punder2_5 = 1/0.3172 = 3.15

Oover2_75 = 1+Punder2_5/(Pover3_5 + 0.5*Pgoals_3) = 1.55
Ounder2_75 = 1+(Pover3_5 + 0.5*Pgoals_3) / Punder2_5 = 2.81

Oover3 = 1 + Punder2_5/Pover3_5 = 1.68
Ounder3 = 1 + Pover3_5/Punder2_5 = 2.47

Oover3_25 = 1+(Punder2_5 + 0.5*Pgoals3)/Pover3_5 = 1.91
Ounder3_25 = 1+Pover3_5/(Punder2_5 + 0.5*Pgoals3) = 2.10

Hope I'm not mistaken, but seems to be accurate.

[robotzel78]

Thanks Dash2in1, i will check now if it is correct.

[v-zero]

Soccer totals are not Poisson, they are first-order linear. Making goal estimates is difficult as it has nothing to do with past scoring intensities for each team.

[_bester_tuff]

Soccer totals are not Poisson, they are first-order linear. Making goal estimates is difficult as it has nothing to do with past scoring intensities for each team.

so does winning and losing games.

[v-zero]

so does winning and losing games.

No, that's wrong.

[Dash2in1]

No, that's wrong.

Indeed, you are right. On that matter and the matter of scores not being correctly poisson distributed.

On that topic, it might be good for this topic to explain how to get to those actual good distributions (I think I've done my part)?

[robotzel78]

Ok, i understand that Poisson it's wrong. But why ? Can you give an example ?

[v-zero]

Soccer is not adequately described by the interaction of two independent distributions, there is covariance between teams that must be modelled. This is less important for the standard match odds than for live modelling, though it is necessary to inflate the draw probabilities output by Poisson, and reduce the away team probabilities.
However, the distribution of over and under is entirely wrong for soccer if using Poisson, and estimating goals by average past goals also makes no sense. Soccer matches are more like a superposition of two distributions that find a joint equilibrium position, and the match will have a general tendency to that equilibrium position. In simple terms this means matches between closely matched teams will be low scoring, and mismatches will be high scoring. It doesn't matter if the closely matched teams are both bad or both good, only that they are closely matched relative to each other. You can see from that information that there is no way you can take past data for a particular team and use it to build a totals model, you must model team strengths and from that model the game types, and hence derive totals. These totals, though, will still not be Poisson. They will be linear to first order, which will be adequate.

[robotzel78]

Ok, thank's for your thoughts. I think it's better to know other ideeas about Poisson before i start to make this system. I like very much to play and bet on Overs, i search on google, and i find information about Poisson, and i want to try it, but now i look for another over (totals) systems.