Converting and re-converting, numbers to combination numberPrev Topic Next Topic

New Topic New Poll

luckyguyxxx

Italy
Member #184,208
August 30, 2017
11 Posts
Offline

Aug 30, 2017, 12:29 pm

Hello, i had an idea.

Let's say we have a lottery like, 6 numbers from 1 to 90.

We would have a number of permutations, like:

1) 1-2-3-4-5-6
2) 1-2-3-4-5-7
3) 1-2-3-4-5-8
...
84) 1-2-3-4-5-90
85) 1-2-3-4-6-7
86) 1-2-3-4-6-8
etc......
untill the
622.614.630) 85-86-87-88-89-90

We would have around 622,614,630 possible permutations.

Now, what i want to do, is convert lottery drawings results, to permutations number.

This mean, if the lottery drawing today is 1-2-3-4-5-6, it would be converted in: 1

If we have for example 85-86-87-88-89-90, it would be converted in 622.614.630.

So we would have a permutation n" history of drawings, just 1 single number for each drawing, instead of 6 different numbers.

We would be able then to study them, and hopefully forecast what will be the next lottery drawing permutation number (and then convert that number to the six matching numbers to be played.

Does it make sense?

Is there a software or something that is able to do such conversion?

tnx very much!
Raven62

25

New Jersey
United States
Member #17,842
June 28, 2005
180,983 Posts
Offline

Aug 30, 2017, 12:58 pm

Quote: Originally posted by luckyguyxxx on Aug 30, 2017
Hello, i had an idea.

Let's say we have a lottery like, 6 numbers from 1 to 90.

We would have a number of permutations, like:

1) 1-2-3-4-5-6
2) 1-2-3-4-5-7
3) 1-2-3-4-5-8
...
84) 1-2-3-4-5-90
85) 1-2-3-4-6-7
86) 1-2-3-4-6-8
etc......
untill the
622.614.630) 85-86-87-88-89-90

We would have around 622,614,630 possible permutations.

Now, what i want to do, is convert lottery drawings results, to permutations number.

This mean, if the lottery drawing today is 1-2-3-4-5-6, it would be converted in: 1

If we have for example 85-86-87-88-89-90, it would be converted in 622.614.630.

So we would have a permutation n" history of drawings, just 1 single number for each drawing, instead of 6 different numbers.

We would be able then to study them, and hopefully forecast what will be the next lottery drawing permutation number (and then convert that number to the six matching numbers to be played.

Does it make sense?

Is there a software or something that is able to do such conversion?

tnx very much!

It's called Lexicographic Numbering:

https://www.lotterypost.com/thread/230350

Lottery Post Forums Search is Your Friend:

https://www.lotterypost.com/search/forums?q=lexicographic&t=all

A mind once stretched by a new idea never returns to its original dimensions!

Catch-22: A dilemma or difficult circumstance from which there is no escape because of mutually conflicting or dependent conditions.

Corruptissima re publica plurimae leges: When the republic is at its most corrupt the laws are most numerous.

The best way to learn is to never stop being an Experiential Student!
JADELottery

Native American Eagle Sun

MN
United States
Member #21
December 7, 2001
4,812 Posts
Offline

Aug 30, 2017, 3:42 pm

We call it Combinatorial Indexing.

If you have Excel, you could try this:

http://www.jadexcode.com/files/excel/CombinatorialIndex.xls

or

ftp://www.jadexcode.com/excel/CombinatorialIndex.xls

The One Over None
I Know...
luckyguyxxx

Italy
Member #184,208
August 30, 2017
11 Posts
Offline

Aug 31, 2017, 7:49 am

Thank you very much, to both posters, that was exactly what i was looking for!

I will keep the thread updated in case i have some advancements in my project..
luckyguyxxx

Italy
Member #184,208
August 30, 2017
11 Posts
Offline

Sep 3, 2017, 10:38 am

Just a small update:

I've been able (thanks to the help i received here) to convert 1200 lottery drawing results to their relative index values (lexicographic index).

To sum up, we have now just one unique number for each lottery drawing (instead of 6), all these lottery history drawings (unique numbers) form a time-series chart.

The question is, do the past drawing are somehow linked each other, and can a future drawing can be forecasted by using past-ones?

A backpropagation neuralnetwork will try to give us an answer to that question, all i can see for now is that the created chart do have a very good distribution, and it's ideal for the NN to learn.

Inputs: 48 past drawings unique numbers. (this number might be reduced in future, need to play around a bit to understand what's the best thing to do)

Neurons: 48 hidden neurons

Output: 1 (the following drawing unique number).

The network is now training on the past 1200 drawings, when the error rate will be exactly 0 (yes, 0!!!!), it will come the time to attempt our forecast.

From a scientific point of view, the randomness of lottery drawing would theoretically make it impossible to forecast the future, even if a neuralnetwork learn how to forecast correctly the past 1200 drawings, this wouldn't make her any better than a random-number-generator in forecasting the n1201 draw.

Still.. i kinda like to play with neural networks and lottery, so why not.. i will make another post in case some unexpected good result comes out of this!

cheers
JADELottery

Native American Eagle Sun

MN
United States
Member #21
December 7, 2001
4,812 Posts
Offline

Sep 3, 2017, 11:01 am

Quote: Originally posted by luckyguyxxx on Sep 3, 2017
Just a small update:

I've been able (thanks to the help i received here) to convert 1200 lottery drawing results to their relative index values (lexicographic index).

To sum up, we have now just one unique number for each lottery drawing (instead of 6), all these lottery history drawings (unique numbers) form a time-series chart.

The question is, do the past drawing are somehow linked each other, and can a future drawing can be forecasted by using past-ones?

A backpropagation neuralnetwork will try to give us an answer to that question, all i can see for now is that the created chart do have a very good distribution, and it's ideal for the NN to learn.

Inputs: 48 past drawings unique numbers. (this number might be reduced in future, need to play around a bit to understand what's the best thing to do)

Neurons: 48 hidden neurons

Output: 1 (the following drawing unique number).

The network is now training on the past 1200 drawings, when the error rate will be exactly 0 (yes, 0!!!!), it will come the time to attempt our forecast.

From a scientific point of view, the randomness of lottery drawing would theoretically make it impossible to forecast the future, even if a neuralnetwork learn how to forecast correctly the past 1200 drawings, this wouldn't make her any better than a random-number-generator in forecasting the n1201 draw.

Still.. i kinda like to play with neural networks and lottery, so why not.. i will make another post in case some unexpected good result comes out of this!

cheers

Good luck with your project.

We've been working with indexing of combinations; in addition to indexing fractal subsets of combinations.

It's an interesting field of study.

The One Over None
I Know...
alchemist7

United States
Member #65,709
October 4, 2008
151 Posts
Offline

Sep 4, 2017, 8:57 pm

Quote: Originally posted by JADELottery on Sep 3, 2017
Good luck with your project.

We've been working with indexing of combinations; in addition to indexing fractal subsets of combinations.

It's an interesting field of study.

Hi Jade,

Can you explain "Fractal subsets of combinations", Please!

Thanks.
JADELottery

Native American Eagle Sun

MN
United States
Member #21
December 7, 2001
4,812 Posts
Offline

Sep 6, 2017, 8:29 am

Quote: Originally posted by alchemist7 on Sep 4, 2017
Hi Jade,

Can you explain "Fractal subsets of combinations", Please!

Thanks.

To explain, we will need to create a new topic aside from this one.

We'll post when we complete our training on a different subject this week.

The One Over None
I Know...
luckyguyxxx

Italy
Member #184,208
August 30, 2017
11 Posts
Offline

Sep 6, 2017, 4:27 pm

Update:

So, the neuralnetwork has been trained up to an error of 0, correctly guessing all the 1200 drawings in the history.

The forecasted drawing however turned out to be simply not the winning one (and quite far from the winning one too!).

Clearly, an "overfitting".

48 inputs * 48 hidden neurons + 48weights to the final output neuron, making a total of 2.352 variables within the network, i think this was the cause of the overfitting (1200 drawings got absorbed quite easily in the network and this didn't provided any correct forecasting capabilities on the network).

I've tried various settings, but seems like whenever the number of total weights variables is less than the number of training drawings, the network isn't able to get to a 0 error level.

I will write again in case some unexpected update comes out! But for now it seems the neuralnetwork idea isn't producing anything of value. :(

Cheers!
Johnebench

New Hampshire
United States
Member #181,609
May 8, 2017
1,069 Posts
Offline

Sep 6, 2017, 8:21 pm

Quote: Originally posted by luckyguyxxx on Sep 6, 2017
Update:

So, the neuralnetwork has been trained up to an error of 0, correctly guessing all the 1200 drawings in the history.

The forecasted drawing however turned out to be simply not the winning one (and quite far from the winning one too!).

Clearly, an "overfitting".

48 inputs * 48 hidden neurons + 48weights to the final output neuron, making a total of 2.352 variables within the network, i think this was the cause of the overfitting (1200 drawings got absorbed quite easily in the network and this didn't provided any correct forecasting capabilities on the network).

I've tried various settings, but seems like whenever the number of total weights variables is less than the number of training drawings, the network isn't able to get to a 0 error level.

I will write again in case some unexpected update comes out! But for now it seems the neuralnetwork idea isn't producing anything of value. :(

Cheers!
luckyguyxxx

Italy
Member #184,208
August 30, 2017
11 Posts
Offline

Sep 7, 2017, 9:48 am

I would like to ask some help, in terms of inspiration.

So, we are talking here about probabilities, chances, precision, and difficulty.

We can't eventually ask a neural network to be able to correctly predict one number from 1 to 600.000.000, as the number of possibilities is astonishing (600 millions!), thus it is like finding a particular sand grain on a whole beach.

What if we lower down the chances, from 1 to 600.000.000 to.. 1 to 2???

That would make a huge difference, right?

Here we are, the drawing results have been converted in binary (+1 or -1), if the number is > than 300.000.000, it is converted to 1.

if it is < of 300.000.000, it is converted to -1.

Same for the inputs, we only have +1s and -1s.

48 inputs >> 48 neurons >> final result.

Incredibly, the learning is quite faster, and the result is somehow promising (correctly forecasting!).

Now, we know that the next result lexicographic number will be between 1 and 300.000.000, bear with me.

What we want to do then, is keep getting more precise, so the next step would be to know if it's between 1 and 150mln or between 150mln and 300mln. (another neural network)

then we will do it again, lowering again the range, then again and again. Unless we will find the exact number!!!!

By now if you think i'm crazy, you are probably right in your forecast :D

Anyway.. the first step was easy, on excel, if(number>300mln,1,-1).

But for the second step, i am finding difficulty on how to do it, because the variable 300mln is now not the same for every row, as some are +1 and some are -1, we would need a more complex formula.

Not even sure if it make sense, but overall what i'm looking for should be clear by now, i just want to disassemble the number to a string of +1 and -1 that would then be given to many neural networks.

Any idea, even crazy or complex ideas, are well accepted!!!

thanks
alchemist7

United States
Member #65,709
October 4, 2008
151 Posts
Offline

Sep 7, 2017, 10:44 am

Quote: Originally posted by JADELottery on Sep 6, 2017
To explain, we will need to create a new topic aside from this one.

We'll post when we complete our training on a different subject this week.

Thanks!
luckyguyxxx

Italy
Member #184,208
August 30, 2017
11 Posts
Offline

Sep 7, 2017, 11:54 am

nevermind about my previous post, i found the solution already.
RL-RANDOMLOGIC

United States
Member #59,352
March 13, 2008
5,626 Posts
Offline

Sep 21, 2017, 6:58 pm

Here is a tool that wheels lexie digits, this shows the results for a 5-39 matrix wheeling the first 4 lexie values

for 100 lines. Have you tried feeding the NN the lexie digits one at a time? Seems to me to be the best method

so far but it's far from what I would like. I wrote a lexie predictor that hit a 5 of 6 right off the bat but in forward

test the best it ever got was several lines with 4 of 6 numbers correct. I've been working with index values for

over 15 years and have not found anything solid as prediction goes, but keep hoping.

RL

P.S. Waiting to see what Jade's got cooking.

Hope it turns out to be a real

....
thamizhpayan

United States
Member #173,010
February 13, 2016
1,629 Posts
Offline

Sep 22, 2017, 12:27 pm

Quote: Originally posted by luckyguyxxx on Aug 31, 2017
Thank you very much, to both posters, that was exactly what i was looking for!

I will keep the thread updated in case i have some advancements in my project..

Hi,

Here is a link with python code

http://www.math.umbc.edu/~campbell/Computers/Python/probstat.html

C++ library:

https://people.sc.fsu.edu/~jburkardt/cpp_src/combo/combo.html

all links in the reference section of the article are links to books and corresponding source code released under open source license(mostly GPL/LGPL). Make sure you follow usage restrictions when you reuse the code as such (each license is different). Or if you are a programmer, you can write those algorithms in your favorite language.

For more information, you can search Internet with search keywords : combinatorics or lexical analysis

Good Luck