Value Predictor

[stan]

you can download and evaluate a new predictor plugin. it's based on pattern matching (see the attached document kindly provided by george). the algorithm is quite close to the 'trend predictor' but the results of the new 'value predictor' is a single trend and value prediction so it should be easier to interpret and back-test.

the download link is:
http://www.expertlotto.com/files/previe ... 0.0.plugin

(again, it's a non-official beta version release and you'll need 'expert lotto pro' license to use it)

pls let me know if you think the predictor results are any good.

[stan]

btw, the above attached document on pattern matching seems to be some kind of university paper so i hope there are no copyright issues.

(otherwise let me know and i'll remove it)

[Da80th]

hi Stan,

I get this error "There were some errors while loading application plugin(s)." on both the full version and beta version of ExpertLotto Pro.

Thanks again,


Da80th

[stan]

[quote=Da80th]
I get this error "There were some errors while loading application plugin(s)." on both the full version and beta version of ExpertLotto Pro.
[/quote]

email me your error.log pls

[Midas]

Stan,

There are no copyright issues with the documentation or the algorithms mentioned. I am pleased to report that Dr Sameer Singh is delighted that some of his ideas are being used, and wishes us all the best in our lottery application. I have a few other interesting papers by the same author and his collaborators, should further information on the subject be required.

Many thanks for the rapid implementation, lets get testing people!.

[stan]

[quote=George]
There are no copyright issues with the documentation or the algorithms mentioned. I am pleased to report that Dr Sameer Singh is delighted that some of his ideas are being used, and wishes us all the best in our lottery application. I have a few other interesting papers by the same author and his collaborators, should further information on the subject be required.
[/quote]

i'm glad to hear that:)
btw, there's no 'noise injection' in the value predictor as i'd have to find some java library for fourier transformation and refresh my university math...

(i've also made some minor modifications to the algorithm from the paper - like preventing division by zero, use absolute values etc)

[Da80th]

hi Stan,

See error message below:

*** Error Sat Jul 29 11:25:53 PDT 2006
*** App Version: 4.2.1 Build: 793
*** Java Version: 1.4.2_03
*** OS Version: Windows XP 5.1
*** Lottery: CA Fantasy Five
*** Message: com/expertlotto/predictor/value/ValuePredictorPlugin (Unsupported major.minor version 49.0)
*** Stack:
java.lang.UnsupportedClassVersionError: com/expertlotto/predictor/value/ValuePredictorPlugin (Unsupported major.minor version 49.0)
at java.lang.ClassLoader.defineClass0(Native Method)
at java.lang.ClassLoader.defineClass(Unknown Source)
at java.security.SecureClassLoader.defineClass(Unknown Source)
at java.net.URLClassLoader.defineClass(Unknown Source)
at java.net.URLClassLoader.access$100(Unknown Source)
at java.net.URLClassLoader$1.run(Unknown Source)
at java.security.AccessController.doPrivileged(Native Method)
at java.net.URLClassLoader.findClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
at com.expertlotto.plugin.PluginManager.loadPlugins(PluginManager.java:141)
at com.expertlotto.Main.run(Main.java:45)
at java.lang.Thread.run(Unknown Source)

[stan]

[quote=Da80th]
hi Stan,

See error message below:

*** Error Sat Jul 29 11:25:53 PDT 2006
*** App Version: 4.2.1 Build: 793
*** Java Version: 1.4.2_03
*** OS Version: Windows XP 5.1
*** Lottery: CA Fantasy Five
*** Message: com/expertlotto/predictor/value/ValuePredictorPlugin (Unsupported major.minor version 49.0)
*** Stack:
java.lang.UnsupportedClassVersionError: com/expertlotto/predictor/value/ValuePredictorPlugin (Unsupported major.minor version 49.0)
at java.lang.ClassLoader.defineClass0(Native Method)
at java.lang.ClassLoader.defineClass(Unknown Source)
at java.security.SecureClassLoader.defineClass(Unknown Source)
at java.net.URLClassLoader.defineClass(Unknown Source)
at java.net.URLClassLoader.access$100(Unknown Source)
at java.net.URLClassLoader$1.run(Unknown Source)
at java.security.AccessController.doPrivileged(Native Method)
at java.net.URLClassLoader.findClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
at com.expertlotto.plugin.PluginManager.loadPlugins(PluginManager.java:141)
at com.expertlotto.Main.run(Main.java:45)
at java.lang.Thread.run(Unknown Source)
[/quote]

i see. i compiled the plugin for java version 1.5
please download and install again, it should be fixed now

(btw, you should upgrade to java 1.5)

[Da80th]

Thanks Stan,

works now.

What strange is the Java version in IE shown:

Java Plug-in 1.5.0_06
Using JRE version 1.5.0_06 Java HotSpot(TM) Client VM

Anyway, working now.

Thanks again


Da80th

[stan]

[quote=Da80th]
What strange is the Java version in IE shown:

Java Plug-in 1.5.0_06
Using JRE version 1.5.0_06 Java HotSpot(TM) Client VM
[/quote]

you have both versions of java runtime installed, version 1.4 is probably in windows PATH variable, while IE picks the latest version by default (probably)
just uninstall java 1.4 and you should be fine

[Maryland]

Stan, should I be getting values higher than the highest lottery number when using the drawings for input?

Trend prediction: Increase
Predicted value: 42.51294117647059
========================================================
Best pattern size: 5
Back-tests: 30
Mean Square Error: 28.460273666009055
Mean Absolute Percentage Error: 0.13302601792074217
Valid trend predictions: 18 (60%)

[stan]

[quote=Maryland]
Stan, should I be getting values higher than the highest lottery number when using the drawings for input?

Trend prediction: Increase
Predicted value: 42.51294117647059
========================================================
Best pattern size: 5
Back-tests: 30
Mean Square Error: 28.460273666009055
Mean Absolute Percentage Error: 0.13302601792074217
Valid trend predictions: 18 (60%)
[/quote]

well, the algorithm doesn't know anything about minimum/maximum values possible - see the paper. so it may happen that there's a historical pattern which is a best match for the latest pattern and when an increase is predicted and the latest value is the maximum then the predicted value will be outside the range...

[stan]

some more comments:

1) it seems the algorithm (or my implementation) doesn't handle sequences with negative values too well. so i don't recommend using it on negative values until i fix it.

2) it might be good idea to check how the winning numbers are actually drawn - e.g. when the ball set changes, when a different machine is used etc. then set the 'validation rows' to the last change in the drawing process

3) experiment with it

[Maryland]

Experimenting is what I'm doing that's how I came up with the higher numbers. The part about how the balls are actually drawn makes more sense in using the predictors. Now to try and get the actual drawn positions is like trying to get the next winning numbers . I guess I could start now with a new data base till I hit about 250 draws to input.

[Maryland]

After reading the papers on the value predictor it seems to say I should use less data rather than more for a good output. Does anyone think I'm interpeting this wrong?