Ideas for Bayesian Networks & predictionPrev Topic Next Topic

New Topic New Poll

SerenityViolet

Melbourne
Australia
Member #84,611
January 1, 2010
6 Posts
Offline

Apr 24, 2010, 9:14 pm

Has anyone looked at Bayesian Networks for tackling the prediction problem? As I understand it Bayesian Networks can offer predication based on incomplete information. The key is the rules that are put in place. For example in weather prediction certain conditions indicate rain, cloudiness, cloud type, humidity, temperature etc. When certain conditions are present rain can reasonably be predicted, even though the full information set isn't known.

I guess the problem with applying this to lotteries is defining valid rules. What are your thoughts about rules that could be applied?
LANTERN

Tx
United States
Member #4,570
May 4, 2004
5,180 Posts
Offline

Apr 25, 2010, 1:04 pm

Prediction rules?

I kind of made filtering rules, not exactly strict and I didn't write them down.

That was more or less how I used to predict.

The information used were of course the patterns made by past draws.

So you are on the right track.

But people and also those who make programs don't seem to be interested on it.

BibleOnline ParishesOnline ChristianRadioOnline MassOnline Mass

"Ten measures of beauty descended to the world, nine were taken by Jerusalem."
RJOh

100

mid-Ohio
United States
Member #9
March 24, 2001
20,272 Posts
Offline

Apr 25, 2010, 1:46 pm

Sounds like what is needed here is a lottery strategy and system shop, a shop that does for lottery players what a machine shop does for inventors. If an inventor gets an idea he makes a few drawings and hires a machine shop which has a group of skilled machinists to make a working model of his idea. Like wise a lottery player who has an idea for winning lotteries could write down his ideas and hire a lottery strategy and system shop with its skilled programmers to develop his ideas into a working program. With the economy in bad shape, there must be some highly skilled programmers out there looking for work as a reasonable price.

* you don't need to buy every combination, just the winning ones *
SerenityViolet

Melbourne
Australia
Member #84,611
January 1, 2010
6 Posts
Offline

Apr 25, 2010, 8:45 pm

I think you are probably right. I've seen several posts where people think they are onto something but don't have the programming skills to implement it. While I did programming I don't use it regularly so my skills are pretty rusty - and I don't have the time, which means that I'm pretty much in the same boat. The good commercial products on the market do seem to pick up ideas as they become mainstream. The problem with single ideas is that you end up with a one-trick pony. I guess that's OK if it turns out to to be the thing that solves everything. Otherwise you want to ombine it with the rest of the tools you are using and that means recreating one of the better commercial ones.
mikemerkury

Canada
Member #87,732
March 2, 2010
10 Posts
Offline

Apr 25, 2010, 8:52 pm

Did many lottery number experiment with Bayesian Networks (Neural and Support Vector Machine) in the past few months . No great success so far. Many models used in order to forecast first number of next Pick3 combo and/or first number direction (higher or lower than last number). Experimented using rules extraction exploration of last 200 winning numbers, ephemeris data with various planetary longitude-lattitude (mercury, venus, mars, etc.) at date/location of drawing, earth geomagnetics data at date/location of drawing, sunspot data, Biorhythm 23/28/33/38/43/48/53 days cycle, solar data, etc.

Any suggestion is welcome and always open to public or private collaboration.
mikemerkury

Canada
Member #87,732
March 2, 2010
10 Posts
Offline

Apr 25, 2010, 9:03 pm

"system shop with its skilled programmers to develop his ideas into a working program" : i have access to analyst, programmer, modeler, etc. Do not hesitate to privately contact me and we can workout something if you have a new/great model that can be validated on historical drawing data. Signed confidentiality agreement, shared percentage profit, fixed cost, and many options available.
SerenityViolet

Melbourne
Australia
Member #84,611
January 1, 2010
6 Posts
Offline

Apr 25, 2010, 9:07 pm

Thanks for your post Lantern. I think that you touch on a crucial point.

The idea of the network is to create a way of extrapolating outcomes from incomplete information. It can be used to model strategies for example. The problem with lotteries is that I am not sure we know what the rules should be. If the rules are simply filtering strategies then is anything additional achieved by using a network? For example we may know that 80% of all potential lines in a pick 6 game contain 2, 3 or 4 odd numbers. We can compare this of actual results. But it doesn't necessarily achieve anything more than a good filtering program does anyway.

I've been thinking about how Bayesian Networks might be applied for a while. I'm sort of stuck, at this point and was wondering if other people had more insight.

I'm also intrigued by your statement about patterns in past draws. I've forgotten most of my math, but another thing I've been wondering about is whether you can calculate a value (eigenvalue?) by treating each drawing as a vector and, as in the above example, whether that would tell you anything additional anyway.
SerenityViolet

Melbourne
Australia
Member #84,611
January 1, 2010
6 Posts
Offline

Apr 25, 2010, 9:25 pm

Thanks MikeMerkury, it's good to know that such options are available. Unfortunately, I don't have much more than some fairly broad ideas at present.

Trying to think about what to pursue. I'm fascinated to hear that you have investigated this angle. Interesting to hear about the types of rules you have applied. Sounds like you are also looking at Matrix math?
mikemerkury

Canada
Member #87,732
March 2, 2010
10 Posts
Offline

Apr 25, 2010, 11:00 pm

Support vector machines are a valuable tool for

making classi?cations, but their black-box nature means that

they lack the natural explanatory value that many other

classi?ers possess.

No hard rules applied. Support vector machines are tool for making classi?cations, but their black-box nature means that they lack the natural explanatory value that many other classi?ers (i.e. Bayes probabilistic model) possess. What i did was to feed historical lottery numbers (and associated data, ephemeris, biorythm, etc.) to tool (tested many : mattlab, cran R project, etc.) and let the svm black box do the (next day, next 10 days, next 100 days, etc.) forecast.
LANTERN

Tx
United States
Member #4,570
May 4, 2004
5,180 Posts
Offline

Apr 26, 2010, 11:33 pm

Quote: Originally posted by mikemerkury on Apr 25, 2010
Did many lottery number experiment with Bayesian Networks (Neural and Support Vector Machine) in the past few months . No great success so far. Many models used in order to forecast first number of next Pick3 combo and/or first number direction (higher or lower than last number). Experimented using rules extraction exploration of last 200 winning numbers, ephemeris data with various planetary longitude-lattitude (mercury, venus, mars, etc.) at date/location of drawing, earth geomagnetics data at date/location of drawing, sunspot data, Biorhythm 23/28/33/38/43/48/53 days cycle, solar data, etc.

Any suggestion is welcome and always open to public or private collaboration.

Mike wrote:

Many models used in order to forecast first number of next Pick3 combo and/or first number direction (higher or lower than last number).

--------------------------

First Digit = 1/10

Chances might also be 1/10

That is why I never predicted, instead I filtered out.

--------

Characteristics are needed

1 Digit out of 10 that is 1 characteristic

Higher or Lower than last, that is another

2 Charcteristics are not enough.

We try to have as many characteristics as possible.

--------------

I don't know anything about:

Bayesian Networks (Neural and Support Vector Machine)

------------------

Instead meditate on what might appear to you to be logical prediction ways.

Ways in which your mind might be able to predict, by some kind of logic.

First one must have (Clear) logical ideas of what prediction is.

Somebody like me, with little to no knowledge of Math, can't exactly do prediction by Math.

But to some extent can make use of simple patterns produced by the past draws-digits-numbers.

Need to have very many patterns and then use the ones that appear to be best for doing the prediction-filtering out with, or whatever.

How to go about predicting? We don't all have the same idea(s), We don't all see prediction in the same way.

Some of us have a 1 track mind and or are 1 trick ponies.

Use whatever works best for you, there is only trouble if whatever you are using is not working for you, then you might have to find some other way(s), even if it is somebody else's way, unless you can develop some other way that works better for you.

-------

Astrology might no exactly be taylor made for lottery prediction, might be best to develop your own way.

It is a thing of making and putting together your own data-patterns-stats into a predictable system, more or less.

----------------

BibleOnline ParishesOnline ChristianRadioOnline MassOnline Mass

"Ten measures of beauty descended to the world, nine were taken by Jerusalem."
LANTERN

Tx
United States
Member #4,570
May 4, 2004
5,180 Posts
Offline

Apr 26, 2010, 11:51 pm

Quote: Originally posted by SerenityViolet on Apr 25, 2010
Thanks for your post Lantern. I think that you touch on a crucial point.

The idea of the network is to create a way of extrapolating outcomes from incomplete information. It can be used to model strategies for example. The problem with lotteries is that I am not sure we know what the rules should be. If the rules are simply filtering strategies then is anything additional achieved by using a network? For example we may know that 80% of all potential lines in a pick 6 game contain 2, 3 or 4 odd numbers. We can compare this of actual results. But it doesn't necessarily achieve anything more than a good filtering program does anyway.

I've been thinking about how Bayesian Networks might be applied for a while. I'm sort of stuck, at this point and was wondering if other people had more insight.

I'm also intrigued by your statement about patterns in past draws. I've forgotten most of my math, but another thing I've been wondering about is whether you can calculate a value (eigenvalue?) by treating each drawing as a vector and, as in the above example, whether that would tell you anything additional anyway.

Look, things are simple:

There is no such thing as prediction, unless one is such as "Nostradamus" or "The Oracle At Delphi"

For the rest of us, there is: "Chances are that chances are".

That is "Probabilities"

What is easier (More accurate of a prediction), to predict 1 out of 10, or to filter out 1 out of 10?

To predict 1 out of 10 = 1/10 chance of being right.

To filter out 1 out of 10 = 9/10 chance of being right.

I don't know Math so maybe that is not right.

Anyhow, I never figured out how to predict, but I did figure out how to filter out.

--------------

As I said to others, Why keep on reposting over and over again?

For some ideasusethe search engine that the Lottery Post has and look at old posts, even at some or many of my own posts.

------------

Look for and you might find, search engines are a great invention.

BibleOnline ParishesOnline ChristianRadioOnline MassOnline Mass

"Ten measures of beauty descended to the world, nine were taken by Jerusalem."
GASMETERGUY

NASHVILLE, TENN
United States
Member #33,371
February 20, 2006
1,044 Posts
Offline

Apr 28, 2010, 10:13 pm

Quote: Originally posted by LANTERN on Apr 26, 2010
Mike wrote:

Many models used in order to forecast first number of next Pick3 combo and/or first number direction (higher or lower than last number).

--------------------------

First Digit = 1/10

Chances might also be 1/10

That is why I never predicted, instead I filtered out.

--------

Characteristics are needed

1 Digit out of 10 that is 1 characteristic

Higher or Lower than last, that is another

2 Charcteristics are not enough.

We try to have as many characteristics as possible.

--------------

I don't know anything about:

Bayesian Networks (Neural and Support Vector Machine)

------------------

Instead meditate on what might appear to you to be logical prediction ways.

Ways in which your mind might be able to predict, by some kind of logic.

First one must have (Clear) logical ideas of what prediction is.

Somebody like me, with little to no knowledge of Math, can't exactly do prediction by Math.

But to some extent can make use of simple patterns produced by the past draws-digits-numbers.

Need to have very many patterns and then use the ones that appear to be best for doing the prediction-filtering out with, or whatever.

How to go about predicting? We don't all have the same idea(s), We don't all see prediction in the same way.

Some of us have a 1 track mind and or are 1 trick ponies.

Use whatever works best for you, there is only trouble if whatever you are using is not working for you, then you might have to find some other way(s), even if it is somebody else's way, unless you can develop some other way that works better for you.

-------

Astrology might no exactly be taylor made for lottery prediction, might be best to develop your own way.

It is a thing of making and putting together your own data-patterns-stats into a predictable system, more or less.

----------------

Now I concentrate on the P5 game and I know ya'll are talking about P3 games. However, the discussion carries over.

Filtering and only filtering will not, IMHO, accomplish what you are trying to find. Filtering and stringing the filters together might do a little better. I don't know at this point

We try to have as many characteristics as possible. If we can think of a filter as a pattern then, yes, you will need as many characteristics as possible. The more the better. But know this, the more patterns/characteristics/filters you have, the less likely you will have a winning combination. That might not be a bad thing. I don't know at this point.

First one must have (Clear) logical ideas of what prediction is. We do need a clear understanding of just what we want to accomplish. Having a clearly defined goal will aid in developing an algorithym.

It is a thing of making and putting together your own data-patterns-stats into a predictable system, more or less. Correct. One must find their own way for there is no lengthy discussion on this site about how to string patterns together. There are at most four people who are also considering patterns and what they might mean. The vast majority are into +1/-1 math or "take the last three drawing and ..............". While I do not condemn those who take this path, it would seem to me that after 5 years of trying without success, one would consider something else. That is why I went to patterns almost as soon as I entered this site.

If you are into patterns/filters/characteristics, I would like to hear more from you.
LANTERN

Tx
United States
Member #4,570
May 4, 2004
5,180 Posts
Offline

Apr 29, 2010, 1:22 am

Quote: Originally posted by GASMETERGUY on Apr 28, 2010
Now I concentrate on the P5 game and I know ya'll are talking about P3 games. However, the discussion carries over.

Filtering and only filtering will not, IMHO, accomplish what you are trying to find. Filtering and stringing the filters together might do a little better. I don't know at this point

We try to have as many characteristics as possible. If we can think of a filter as a pattern then, yes, you will need as many characteristics as possible. The more the better. But know this, the more patterns/characteristics/filters you have, the less likely you will have a winning combination. That might not be a bad thing. I don't know at this point.

First one must have (Clear) logical ideas of what prediction is. We do need a clear understanding of just what we want to accomplish. Having a clearly defined goal will aid in developing an algorithym.

It is a thing of making and putting together your own data-patterns-stats into a predictable system, more or less. Correct. One must find their own way for there is no lengthy discussion on this site about how to string patterns together. There are at most four people who are also considering patterns and what they might mean. The vast majority are into +1/-1 math or "take the last three drawing and ..............". While I do not condemn those who take this path, it would seem to me that after 5 years of trying without success, one would consider something else. That is why I went to patterns almost as soon as I entered this site.

If you are into patterns/filters/characteristics, I would like to hear more from you.

https://www.lotterypost.com/thread/134410/606727

https://www.lotterypost.com/thread/134410/607076

Hi!

GASMETERGUY

Click on those links, start with the first one.

That is one of the better samples of what can be done with filtering for the pick 3 game.

That was done with filtering and only filtering.

------------

I don't know what you mean by:

Filtering and stringing the filters together.

-----------

You said:

If you are into patterns/filters/characteristics, I would like to hear more from you.

----------

Yes, I am or Yes, I was.

In order to "Hear" more about that from me, you will have to use the Lottery Post's search engine option, look for old posts made by me, but mostly look for old threads started by me.

Maybe you could look for old posts made by "EXCALIBUR" and also for old posts made by "LANTERN"

Use all the history, go back all the way or at least as far back as Sep 2003.

The threads and posts might mostly be on the "Lottery Systems" and on the "Pick 3" forums, at least the posts which have information that you might want to see, maybe mostly on the "Lottery Systems" forum.

-----------

It might be hard as I posted so very much, many posts are good and many others are trash, you might have to go thru all of them in order to find what you want.

BibleOnline ParishesOnline ChristianRadioOnline MassOnline Mass

"Ten measures of beauty descended to the world, nine were taken by Jerusalem."
GASMETERGUY

NASHVILLE, TENN
United States
Member #33,371
February 20, 2006
1,044 Posts
Offline

May 1, 2010, 1:18 pm

Lantern

After your hiatus from Lottory Post, you returned with the comment that you would post again some of your thoughts. I replied at that time that you had a lot of posting to do. The reason I said that was due to the fact I search this site all the time. I will go back to day one and re-read all that was posted for many, many hours. I have yet to travel from Day One to Present Day in one sitting; there are just too many posts. But you had a lot to say"back in the day" and I found them extremely informative.

It was your ideas that led me to where I am today. In other words, "IT'S ALL YOUR FAULT!!" LOL.

I do not think that filtering is the only way to go. I do think filtering is the best area for research as of today. But who knows? Maybe Badger42 will prove that some form of math will make any P5 game profitable thus beating me to the finish line. Or that someone will find the +1/-1 algorithm that makes Lotto profitable. But who knows?

For an idea about stringing patterns together, read my blog. I put some very simple, straighforward comments there. I have also aluded to the same information in various threads on this site. Unfortunately, no one bothered to reply. So I just hunkered down (as they say in the mountains of East Tennessee) and went about my business.
LANTERN

Tx
United States
Member #4,570
May 4, 2004
5,180 Posts
Offline

May 1, 2010, 9:19 pm

Quote: Originally posted by GASMETERGUY on May 1, 2010
Lantern

After your hiatus from Lottory Post, you returned with the comment that you would post again some of your thoughts. I replied at that time that you had a lot of posting to do. The reason I said that was due to the fact I search this site all the time. I will go back to day one and re-read all that was posted for many, many hours. I have yet to travel from Day One to Present Day in one sitting; there are just too many posts. But you had a lot to say"back in the day" and I found them extremely informative.

It was your ideas that led me to where I am today. In other words, "IT'S ALL YOUR FAULT!!" LOL.

I do not think that filtering is the only way to go. I do think filtering is the best area for research as of today. But who knows? Maybe Badger42 will prove that some form of math will make any P5 game profitable thus beating me to the finish line. Or that someone will find the +1/-1 algorithm that makes Lotto profitable. But who knows?

For an idea about stringing patterns together, read my blog. I put some very simple, straighforward comments there. I have also aluded to the same information in various threads on this site. Unfortunately, no one bothered to reply. So I just hunkered down (as they say in the mountains of East Tennessee) and went about my business.

GASMETERGUY

I do for sure have posted a lot.

For sure, filtering is not the only way to go, for one thing there is "prediction by probability", there might be more than one way in which that might work.

One of those possible ways was used for the digits predictions posted on the Tx Tread some time ago and I guess that it worked fair enough, until the Tx Lottery found some way to defeat that method.

But know that predictions by filters is also another way of doing "prediction by probability"

But unlike the way in which I did prediction by filters, I used the help of prediction software and some other software in order to be able to do the digits predictions that I posted time ago on the Pick 3 Forum.

As to Jackpot games,if I myself was able to make softwares and if I was good enough at that, then maybe I would have a good chance of making the right filters software for those games, which in time would allow me to make fair or good filters predictions for those games.

Math Formula(s) for prediction?, Maybe, but I doubt it.

+1/-1? Maybe, but I don't think so.

Maybe sometime I will take a look at your blog, thanks!

BibleOnline ParishesOnline ChristianRadioOnline MassOnline Mass

"Ten measures of beauty descended to the world, nine were taken by Jerusalem."

New Topic New Poll

Subscribe to this topic

Ideas for Bayesian Networks & predictionPrev TopicNext Topic

Ideas for Bayesian Networks & predictionPrev Topic Next Topic