Welcome Guest
You last visited December 10, 2016, 7:11 pm
All times shown are
Eastern Time (GMT-5:00)

# If lotteries weren't completely random

Topic closed. 9 replies. Last post 12 months ago by ArizonaDream.

 Page 1 of 1

Member #170667
December 11, 2015
9 Posts
Offline
 Posted: December 17, 2015, 1:26 pm - IP Logged

For entertaining purposes, let's assume for the moment that lotteries weren't completely random.

That they follow some sort of law.

Now for a pick 3? for the people who know about neural networks...what would be a good model to a neural network?

Right now i just got this simple of 30 inputs, 30 hidden units, 30 outputs.

I am training it as if i enter a draw as inputs, i train it to output the next draw.

30 inputs because if the result of a draw was for example 2-5-8, I would set all my inputs to zeros then set input[2] to 1, and input[15] to 1 and input[28] to 1 (that's why there are 30, 0-9 for the first digit, 10-19 for the second digit and 20-29 for the third digit).  Similarly, this structure is also used for outputs.

So have you had experience with neural networks and are able to suggest better/other models for pick 3?

I want to be able to use this neural net to post predictions for pick3 (specifically California Daily 3 for now) to see if my odds are good  using this forum's predictions feature (I think it's a great feature by the way).

Greenville, SC
United States
Member #169870
November 4, 2015
374 Posts
Offline
 Posted: December 17, 2015, 3:00 pm - IP Logged

This got me thinking about lotteries switching to RNGs to pick winning numbers.  Is it really random? I wouldn't be surprised if they choose a set of numbers that isn't played in the current drawing. Of course, that is corruption.

May the balls bounce in your favor!

Member #170667
December 11, 2015
9 Posts
Offline
 Posted: December 17, 2015, 3:12 pm - IP Logged

I was thinking of the same thing too like they could choose numbers that played the least amount of times to maximize their profit but you're right that would be major corruption.

United States
Member #51269
April 3, 2007
529 Posts
Offline
 Posted: December 17, 2015, 10:26 pm - IP Logged

It's completely random.  You could test this very easily, using Excel, actually.  Just play virtually and you can confirm if anyone's tampering with the results.  After all, no one could know what lines you have on your computer, so they can't exclude them.

Dump Water Florida
United States
Member #380
June 5, 2002
3106 Posts
Offline
 Posted: December 22, 2015, 7:05 pm - IP Logged

Completely random, as random as humans can manage, random enough for a lottery, take your pick.

Neural networks need to work with numbers that represent something, are a measure of something, numbers that carry weight or value.  If there is absolutely no difference between a five ball and a ten ball, there is nothing to work with.  Find a way to give value to the numbers besides frequency, then maybe.

Back in the nineties there was a program called NeuroLotto that needed the original GH database to run and by the time I got my hands on it that version of the GH database was gone and the new one wouldn't work with it.

What it did was graph the number hits over a range and generate endless nets trying for a at best match up where we'd play what was projected for the next draw.  Would have been interesting to work with.

Then there was ProbaLotto (dos) which gave weights to each ball based on frequency and drew your plays from that hat.  Kind of like bobbing for sinkers and floaters, where the strong numbers rise to the surface and the weaker are less likely, but still possible to be captured for wheeling play.

BobP

United States
Member #121745
January 16, 2012
4806 Posts
Online
 Posted: December 22, 2015, 8:34 pm - IP Logged

Randomness is an interesting concept in many areas. People often dismiss modern abstract painting as being so simple a child could paint it. I used to do an exercise in class where I gave adult students 20-30 little plastic pieces/shapes and asked them to arrange them in a random abstract pattern. After about 30 seconds, they will look at their too quickly finished product and think...oops...I need to rearrange this...lol. Then they start to obsess over it.

J'aime La France.

South Africa
Member #129822
June 28, 2012
366 Posts
Offline
 Posted: December 23, 2015, 6:38 am - IP Logged

The winning numbers is not the best numbers BUT the least played numbers. Less numbers played means more profit for them and at the end of the day they run a business

United States
Member #169277
October 10, 2015
630 Posts
Offline
 Posted: December 23, 2015, 7:08 am - IP Logged

The winning numbers is not the best numbers BUT the least played numbers. Less numbers played means more profit for them and at the end of the day they run a business

I think that would be too easy to spot over time if they did so consistantly.

For the lotto games, what I sometimes wonder about, is an anti-jackpot bias that could be turned on and off. Turn it on to cause roll-overs and build interest in the game,  then, once at the level they want, turn it off and let the game run normally. I think that would be very hard to spot, much less prove.

Texas
United States
Member #150797
December 31, 2013
815 Posts
Offline
 Posted: December 23, 2015, 7:14 am - IP Logged

Welcome to the forum tdottolotto.

Personally, I do believe that randomness follows some law.  However, I believe that the law is outside the realm of what humans can comprehend, either too simple, too complex, or too different from what we are expecting.  Maybe your neural net can figure it out.

If you haven't done so already, I would recommend that you do a search for neural networks to see what members have proposed in the past.  And, visit the Mathematics forum and read the current thread there on Neural Networks.

"There is no such thing as luck; only adequate or inadequate preparation to cope with a statistical universe."

~Robert A. Heinlein

United States
Member #169277
October 10, 2015
630 Posts
Offline
 Posted: December 23, 2015, 8:03 pm - IP Logged

Here's an idea I had (and tried on a small scale.)

Once you have a Neural Net designed, run it on some true random data, like from random.org.  In back and forward testing, the NN should perform in line with the game's calculated odds with truly random data.   Now train the same NN on your lottery data and back/forward test. Does it perform the same, better or worse than it did on the known random data?   If the performance differs, there's something non-random going on in that game.

 Page 1 of 1