Value Predictor

[stan]

[quote=Maryland]
Should your input be as the Trend Predictor with lastest at bottom for the Value Predictor?
[/quote]

correct

[Paulo]

[quote=stan:1156532322]
[quote=Maryland]
Should your input be as the Trend Predictor with lastest at bottom for the Value Predictor?
[/quote]

correct
[/quote]

Hi Stan,

You have a example to configuration in this new uptade plugin?

What is the best statistics to predicted with this plugin?

I like to predict WN History.

Thanks

[stan]

[quote=Paulo]
[quote=stan:1156532322]
[quote=Maryland]
Should your input be as the Trend Predictor with lastest at bottom for the Value Predictor?
[/quote]

correct
[/quote]
You have a example to configuration in this new uptade plugin?

What is the best statistics to predicted with this plugin?

I like to predict WN History.
[/quote]

wn history data are fine for predictor. just use all the default settings, enter 7 or 10 into the rows for validation field. i recommend more than 200 hundred rows in the input

[Paulo]

[quote=stan]
[quote=Paulo]
[quote=stan:1156532322]
[quote=Maryland]
Should your input be as the Trend Predictor with lastest at bottom for the Value Predictor?
[/quote]

correct
[/quote]
You have a example to configuration in this new uptade plugin?

What is the best statistics to predicted with this plugin?

I like to predict WN History.
[/quote]

wn history data are fine for predictor. just use all the default settings, enter 7 or 10 into the rows for validation field. i recommend more than 200 hundred rows in the input
[/quote]

Stan,

My lottery have 152 drows.

Is the value that I should consider that?

Trend prediction: Decrease
Predicted value: -2.0

Thanks

[stan]

[quote=Paulo]
My lottery have 152 drows.

Is the value that I should consider that?

Trend prediction: Decrease
Predicted value: -2.0
[/quote]

if you're trying to predict wn history sum values then the value is obviously wrong. the current version of the value predictor knows nothing about possible minimum / maximum values. it will be fixed in some future version.
besides 152 rows may not be enough - look at your sum charts, are they distorted at their beginnings? (starting with small values and gradually rising?)

[Paulo]

[quote=stan]
[quote=Paulo]
My lottery have 152 drows.

Is the value that I should consider that?

Trend prediction: Decrease
Predicted value: -2.0
[/quote]

if you're trying to predict wn history sum values then the value is obviously wrong. the current version of the value predictor knows nothing about possible minimum / maximum values. it will be fixed in some future version.
besides 152 rows may not be enough - look at your sum charts, are they distorted at their beginnings? (starting with small values and gradually rising?)
[/quote]

Stan,

You can please to interpret this example:
___________________________________________________________
Trend prediction: Same
Predicted value: 3.0
========================================================
Best pattern size: 6
Best weight: 1.0
Matching pattern: [1.0 -3.0 2.0 -1.0 -2.0 -7.0] => 2.0
Marker: 104
Back-tests: 30
Mean Square Error: 37.748259515288005
Mean Absolute Percentage Error: 1.9083068783068784
Valid trend predictions: 17 (56,7%)




Trend prediction: Decrease
Predicted value: -29.32307692307692
========================================================
Best pattern size: 13
Best weight: 1.0
Matching pattern: [-2.0 1.0 1.0 2.0 -2.0 0.0 -5.0 -3.0 -4.0 4.0 4.0 2.0 1.0] => -3.0
Marker: 132
Back-tests: 30
Mean Square Error: 91.82547521642283
Mean Absolute Percentage Error: 3.137006557861258
Valid trend predictions: 18 (60%)

[stan]

first of all i'm not sure whether the predictor can handle input with negative values. it may need some more tuning on my part.

as for the results - the first set is based on the lowest MSE on past results and the second set is based on the greatest number of valid trend predictions.

basically you just need to track the predicted trend and value, you can ignore the rest. also valid trend predictions field is interesting as it shows how many trend predictions were correct when the predictor back-tested itself on the validation rows. the greater number the more chances are that the trend prediction will be valid for the coming draw.

[scenium]

Hi Stan,

I'm looking forward to the new predictor.
Are you just making a layout changes to the predictor
or will there be improvement for better prediction?


Thanks!

[stan]

[quote=scenium]
I'm looking forward to the new predictor.
Are you just making a layout changes to the predictor
or will there be improvement for better prediction?
[/quote]

there will be mainly improvements in the core code to improve the prediction precision. although there might be some ui modifications as well.
i have a lot of input and comments from george to process - i just wish i had more free time...

[Alcymart]

Good work on the updated Value Predictor! We are going to have ourselves a lot more fun now! George, you are doing an excellent Job. You are the man! I love the updated version, I just had to say it. It predicts much better than I. This will definetly remove human error when estimating! I am so happy about purchasing Expert Lotto, I have to scream it everywhere.

Sorry about that, I am just super excited when new plug-ins or new updates come along. This software is always making big steps forward!

Thank you, thank you, thank you!!

[stan]

[quote=Alcymart]
Good work on the updated Value Predictor! We are going to have ourselves a lot more fun now! George, you are doing an excellent Job. You are the man! I love the updated version, I just had to say it. It predicts much better than I. This will definetly remove human error when estimating! I am so happy about purchasing Expert Lotto, I have to scream it everywhere.

Sorry about that, I am just super excited when new plug-ins or new updates come along. This software is always making big steps forward!

Thank you, thank you, thank you!!
[/quote]

well, tell your friends then. an extra residual income from expert lotto licenses helps to compensate my girlfriend for the long hours i spend at my computer

[Midas]

The credit is entirely Stan's and EL, I've just helped out my fellow lotto players.

Surely gratuities to our gurus must be the only way expert lotto winners can truely show their appreciation!.

We don't want to see Stan unlucky after all his outstanding efforts. I for one do not want to be accountable to his girlfriend!.

Enjoy Value Predictor

Best Wishes
George

[scenium]

Hi Stan,

I am just thinking if including a variable like "no of rows in history columm" in your pattern matching algorithm would help to decide accurately the next direction of a history columm(+ or -).

i.e...less rows....pattern matching number valid trend prediction that next direction would be positive.

.....more rows...pattern matching number of valid trend prediction that next direction would be negative.


additionally...maybe including the number of gaps within the rows would be helpful.
ie...if there are more rows....but only few numbers occupy the last rows..
would not make an accurate forecast.

[scenium]

Maybe there are some other statistics which would fit your pattern matching algorithm perfectly.
Finding a few statistics variable which are closely linked & fit your current prediction algorithm, hopefully we would find a model for accurate prediction.

[stan]

[quote=scenium]
I am just thinking if including a variable like "no of rows in history columm" in your pattern matching algorithm would help to decide accurately the next direction of a history columm(+ or -).
[/quote]
do you mean the number of rows in the history differences table? that's actually a good indicator for wn history estimates and we're working on a separate wn history predictor that will use the differences table as one of its inputs.