Weight of Money (WOM)

[BetScalper]

Why is WOM calculated:

- back_volume / (back_volume + lay_volume)

and not

- back_volume / lay_volume

Is there a scientific/ mathematical reason ?

[gutuami]

because of percentage view.

[Derek27]

The first formula keeps WOM between 0-100%, the second can reach near infinity.

[ShaunWhite]

The guys above have said it but nothing beats a picture...

One gives you a linear range of values, the other an exponential one. The values on the line are easier to handle, the exponential curve also changes when the total vol changes, the ones on the line don't.

The horizonal values go from BV=100 LV = 0 through to BV = 0 LV = 100.

Untitled.png

[BetScalper]

Many thanks for the replies.

And Shaun, once again your a star with that graph as it’s given me another idea.

[jimibt]

to recap everything said before but to also give further clarity.

if you have two values (say 6 for back and 4 for lay), this means that the TOTAL you are concerned with is 10 units. thus, to find a percentage of one over 'tother, you need to sum the components (6+4=10) and then divvy out the required percentage. for back:

6 /(6+4) = 60%

If you were to measure one against the other, you'd be looking at a far different comparison altogether -i.e. how mnay units of A was contained in unit B. In the case above, this would equal 1.5 units as 6/4...

hope the new idea sparked from Shaun's graphs works overtime for you, but as a percentage, go with the former equation

[BetScalper]

to recap everything said before but to also give further clarity.

if you have two values (say 6 for back and 4 for lay), this means that the TOTAL you are concerned with is 10 units. thus, to find a percentage of one over 'tother, you need to sum the components (6+4=10) and then divvy out the required percentage. for back:

6 /(6+4) = 60%

If you were to measure one against the other, you'd be looking at a far different comparison altogether -i.e. how mnay units of A was contained in unit B. In the case above, this would equal 1.5 units as 6/4...

hope the new idea sparked from Shaun's graphs works overtime for you, but as a percentage, go with the former equation

Thanks Jim.

[BetScalper]

I been doing some testing around WOM. Yes, I know its old hat and some consider outdated.

However,

Obviously money can enter at any point, especially in the last 3 minutes of a UK horse race but I decided to go with the following:

- Place more emphasis on money towards the front of the queue and less towards the back (3 best back prices / 3 best lay prices). You spoofing muppets.
- Take an average by combined weighted money available

Example:

A = ((back_amount * 1.00) + (back_amount2 * 0.50) + (back_amount3 * 0.25)) / 3
B = ((lay_amount * 1.00) + (lay_amount2 * 0.50) + (lay_amount3 * 0.25)) / 3
C = A / (A + B)

Opening Trades:

If C < 30% Then Look to Open Back Trade
If C > 70% Then Look to Open Lay Trade

Closing Trades:

If C > 50% or 5 seconds to go before scheduled start Then Look to Close Back Trade
If C < 50% or 5 seconds to go before scheduled start Then Look to Close Lay Trade

Early days but seems to have done fairly well today when trading only the favourite. Haven't tried it on other selections but might include the first 3 favourites depending on the type of race etc.

Enjoy.

[jimibt]

I been doing some testing around WOM. Yes, I know its old hat and some consider outdated.

However,

Obviously money can enter at any point, especially in the last 3 minutes of a UK horse race but I decided to go with the following:

- Place more emphasis on money towards the front of the queue and less towards the back (3 best back prices / 3 best lay prices). You spoofing muppets.
- Take an average by combined weighted money available

Example:

A = ((back_amount * 1.00) + (back_amount2 * 0.50) + (back_amount3 * 0.25)) / 3
B = ((lay_amount * 1.00) + (lay_amount2 * 0.50) + (lay_amount3 * 0.25)) / 3
C = A / (A + B)

Opening Trades:

If C < 30% Then Look to Open Back Trade
If C > 70% Then Look to Open Lay Trade

Closing Trades:

If C > 50% or 5 seconds to go before scheduled start Then Look to Close Back Trade
If C < 50% or 5 seconds to go before scheduled start Then Look to Close Lay Trade

Early days but seems to have done fairly well today when trading only the favourite. Haven't tried it on other selections but might include the first 3 favourites depending on the type of race etc.

Enjoy.

might just be me, but i think your pro-rating should equal 1, therefore, your sums should be more along the lines of:

A = ((back_amount * 0.6) + (back_amount2 * 0.25) + (back_amount3 * 0.15)) / 3
B = ((lay_amount * 0.6) + (lay_amount2 * 0.25) + (lay_amount3 * 0.15)) / 3
C = A / (A + B)

otherwise, you're doing some weird abstractions that can't be reverse engineered

[BetScalper]

I been doing some testing around WOM. Yes, I know its old hat and some consider outdated.

However,

Obviously money can enter at any point, especially in the last 3 minutes of a UK horse race but I decided to go with the following:

- Place more emphasis on money towards the front of the queue and less towards the back (3 best back prices / 3 best lay prices). You spoofing muppets.
- Take an average by combined weighted money available

Example:

A = ((back_amount * 1.00) + (back_amount2 * 0.50) + (back_amount3 * 0.25)) / 3
B = ((lay_amount * 1.00) + (lay_amount2 * 0.50) + (lay_amount3 * 0.25)) / 3
C = A / (A + B)

Opening Trades:

If C < 30% Then Look to Open Back Trade
If C > 70% Then Look to Open Lay Trade

Closing Trades:

If C > 50% or 5 seconds to go before scheduled start Then Look to Close Back Trade
If C < 50% or 5 seconds to go before scheduled start Then Look to Close Lay Trade

Early days but seems to have done fairly well today when trading only the favourite. Haven't tried it on other selections but might include the first 3 favourites depending on the type of race etc.

Enjoy.

might just be me, but i think your pro-rating should equal 1, therefore, your sums should be more along the lines of:

A = ((back_amount * 0.6) + (back_amount2 * 0.25) + (back_amount3 * 0.15)) / 3
B = ((lay_amount * 0.6) + (lay_amount2 * 0.25) + (lay_amount3 * 0.15)) / 3
C = A / (A + B)

otherwise, you're doing some weird abstractions that can't be reverse engineered

Interesting, your probably correct Jim, I will give it a blast tomorrow. Cheers.

[jimibt]

Interesting, your probably correct Jim, I will give it a blast tomorrow. Cheers.

ooops - typo, should be no divide by 3 at the end:

A = ((back_amount * 0.6) + (back_amount2 * 0.25) + (back_amount3 * 0.15))
B = ((lay_amount * 0.6) + (lay_amount2 * 0.25) + (lay_amount3 * 0.15))
C = A / (A + B)

g'nite

[foxwood]

might just be me, but i think your pro-rating should equal 1, therefore, your sums should be more along the lines of:

A = ((back_amount * 0.6) + (back_amount2 * 0.25) + (back_amount3 * 0.15)) / 3
B = ((lay_amount * 0.6) + (lay_amount2 * 0.25) + (lay_amount3 * 0.15)) / 3
C = A / (A + B)

otherwise, you're doing some weird abstractions that can't be reverse engineered

Can't see a reason that it has to add up to 1 and you can't reverse engineer yours anyway - there are 6 unknowns in the formula and none of them can be correctly deduced from the answer.

To me his formula makes sense since it means he's saying he believes all the front money is genuine, only half of the second tranche is real and only a quarter of the third amount real as well. Probably a better analysis of the real money than forcing an arbitrary total of 1 which ends up disregarding 40% of the money on offer at the front.

Dividing by 3 or not is probably irrelevant since the same thing is done to both sides so the ratio between them remains the same.

Big thing though is that his formula (right or wrong ) worked ok so no sense in changing anything until more indication of how good it is as an indicator for him.

[BetScalper]

Good thoughts and feedback!

Going forward I am going to run 3 simulations and then try and make something out of the overall results after 7 days of racing. Small sample I know but that should give me some data to work with.

Simulation 1
A = (back_amount1 * 0.34) + (back_amount2 * 0.33) + (back_amount3 * 0.33)
B = (lay_amount1 * 0.34) + (lay_amount2 * 0.33) + (lay_amount3 * 0.33)
C = A / (A + B)

Simulation 2
A = (back_amount1 * 1.00) + (back_amount2 * 0.50) + (back_amount3 * 0.25)
B = (lay_amount1 * 1.00) + (lay_amount2 * 0.50) + (lay_amount3 * 0.25)
C = A / (A + B)

Simulation 3
A = (back_amount1 * 0.60) + (back_amount2 * 0.25) + (back_amount3 * 0.15)
B = (lay_amount1 * 0.60) + (lay_amount2 * 0.25) + (lay_amount3 * 0.15)
C = A / (A + B)

From the simulations and testing I have done already then I am very confident that Simulation 1 will give the worst results out of the 3 but I need to back that statement up with some refined data.

Its all fun.

Enjoy,

[jimibt]

To me his formula makes sense since it means he's saying he believes all the front money is genuine, only half of the second tranche is real and only a quarter of the third amount real as well. Probably a better analysis of the real money than forcing an arbitrary total of 1 which ends up disregarding 40% of the money on offer at the front.

hi foxy, will respectfully agree to differ on this one . What i'm using is a weighted mean formula which (as it says on the tin), gives uplift to certain values and downplays other values. if the sum were taken as a straight value as per:

A = (back_amount1) + (back_amount2) + (back_amount3)

.... then, each component of this sum would receive a 33.333 (to infinity - and beyond - lol) weighting in the calculation. This is based on the concrete number of entries being analysed (in this case 3) and apportioned out across each to give an equal weighting. We don't natuarally SEE this as being weighted as each component is of equal *weight*. However, if we were being techically correct, it would be thought of as:

A = (back_amount1 * 0.33333333) + (back_amount2 * 0.33333333) + (back_amount3 * 0.33333333)

If all the weights are equal, then the weighted mean equals the arithmetic mean (the regular “average” you’re used to). You can think of each number contributing 1/3 to the total mean (as there are 3 numbers in the set).

Right, as for my pathetic effort . My proposal was that far from degrading the 1st entry by 40%, we'd be uplifting its weight from 33.333% to 60%, thus giving it a net additional strength of 26.66%. This therefore allows us to fiddle around with weightings as long as the sum total of the weightings is constrained to 1 (which is our natural). This is known as compensating for bias. In BetScalpers case, he wants to push the spoof money out of focus.

So, referring back to said effort:

A = (back_amount1 * 0.60) + (back_amount2 * 0.25) + (back_amount3 * 0.15)

In effect, we're giving back_amount1 more than double (2.4 times) the strength of back_amount2... etc. Each of these decimal weightings is called a weighting factor and can be applied in any fashion as long as the sum of the factors adds up to 1. Thus, the spoof money (back_amount3) contributes only 15% to the perceived strength of the WOM.

It's an interesting topic (weighting) and I believe that the settings for WOM in BA actually follow a similar pattern, with the defaults being set to 34,33,33.

Sorry to grab our old friend mr google into the mix but these may help to explain far better than my example:

https://www.statisticshowto.datascience ... hted-mean/
https://www.statisticshowto.datascience ... ng-factor/
https://en.wikipedia.org/wiki/Weighted_arithmetic_mean

Anyway, I ramble too much - hope this makes sense.

[foxwood]

Well that got you scribbling early in the morning

Yes, understand all that - just prefer the unbounded myself.

Most people don't seem to reckon it as a metric anyway - never found it much use myself - maybe everyone should use the Jim Ratio (patent pending) instead