Predicting lottery number spread using neural networks

[TwoPointOh]

Hey guys. I had this idea a few days ago. I was talking to someone about lottery odds, and it came to pass that we got to discussing how to win by using statistics. My idea is this: say you buy 200 tickets. If you get the spread of numbers in each ticket right, you stand a very good chance of making more than £200. But how to calculate the spread? Well, I had a think with my computer scientist hat on, and I came up with neural networks.

For those of you who don't know, a neural network is basically a way of making an analog learning system. You feed it a set of data and a set of results from that data, then you tell it to go extrapolate*.

So what are people's opinions on this, and mathematically targeted gambling in general? I've also been toying around with several spread betting algorithm ideas, like predicting the currency market and stuff. Obviously, the best way to test these out is to use the vast amount of data available (previous lottery numbers, etc.) and plug the numbers into a spreadsheet or similar.


*Well, this is one way of using neural networks. It's actually a rather vast subject.

[chriwi]

I think neuronal Networks only help if htere is an underlying system what is not yet known or understood, so it might work for econnomy or as you say currencys. For Lottery there should be by de3finition no underlying system but only random, so also neuronets will not help there.

Thats my opinion.

[TwoPointOh]

I see where you're coming from, but my idea is not to find an underlying system to the numbers. What I want to do is find the probabilities of certain spreads of numbers. I'm finding it hard to illustrate this in words, so I might draw up a diagram tonight.

[March_Hare]

I'd say that the chance of winning something in a random lottery is simply governed by (a) the number of tickets in the lottery, and (b) the number of tickets owned. The numbers on the tickets haven't got anything to with the chance to win, as long as we can exclude unusual factors such as (a) multiple tickets with the same number, and (b) tickets on sale that don't get included in the actual lottery.

Each ticket (each number) has a chance of winning equal to 1/T (where T is the number of tickets), so at the first draw you have a chance of winning that draw equal to n (number of tickets) x 1/T.

At the next draw (assuming there are multiple prizes), the chance of a win on 1 ticket becomes 1/(T-1) because the ticket that won in the first draw is not eligible for another prize (usually). So
- with n tickets (not having won in the first draw) we get a chance of n / (T-1)
- with n-1 tickets (we won in the first draw) it's (n-1) / (T-1)
Et cetera.

The chance that you win ANYTHING in this lottery follows from these relationships as well but unfortunately I regret to say that I have a little bit forgotten how to calculate that

Someone? Help? Please?

[LarryS]

If one fully understands the rules of probability theory, then the only law that you need to observe is...

You can't lose if you don't play

[spencer]

Isn't this the domain of trisectors ?





Although, a group of MIT students, made a blackjack killing, and did get kicked out of Las Vegas.

[DCWhitworth]

I'm only really familiar with the UK lottery and your 'system' definitely won't work with that.

The chance of any particular ticket winning is entirely independent of any other ticket. The only way they interact is the bigger prizes are split between however many tickets win.

So, although you cannot increase your chances of winning you *could* increase the amount you win because although the numbers drawn are perfectly random the numbers people pick are not. So if you avoid commonly selected numbers then if you *do* hit the jackpot you have a smaller chance of having to share it.

[chriwi]

You just have to keep in mind that despite of the fact that one number has not been drawn for a long time or another number has been drawn just last sttureday both still have the same probability to be drawn next week, everything else would be an underlying system and should be ruled out as soon as it is observed.
For Blackjack and the laws how it has to be handled in the casinos in a certain country there might be such an unknown underlying system for any lottery just draing balls from a reservoir there is not and if there is it must be changed according to the rules.

[March_Hare]

Although, a group of MIT students, made a blackjack killing, and did get kicked out of Las Vegas.

Ok, but Blackjack isn't a lottery. It's more than a game of random numbers that win or don't.
And if players cooperate at a Blackjack table, there probably are ways to break the bank (BTW: not saying the people you refer to did that).

[PorchFlamingo]

Hello TwoPointOh !

chriwi is on the mark with the application of neural networks to random problems, it just won't work. So the lottery has to be non random to use NNs.

Lotteries run the machines before the draw, at the draw, and after the draw and their statitician checks for randomness. So one can imagine that the lotteries think their games are random. Also, they change out the sets of balls regularly. (For a good discussion of randomness visit random.org).

If the lottery is not truely random it would be because of the machine, paddle or drum, vacuum or gravity, and the balls which vary +/- 2.5% in weight and 50 mm +/- 0.1 in diameter and the ink which is at least two orders of magnitude lighter than the balls.

So, if you stil think there is a chance, find a good joone programmer, I've got a data set they would be interested in.

Cheers

[Tau]

A neural network has the same weakness as the human brain that it sees patterns even if there aren't any.

[annresearcher]

Hey guys. I had this idea a few days ago. I was talking to someone about lottery odds, and it came to pass that we got to discussing how to win by using statistics. My idea is this: say you buy 200 tickets. If you get the spread of numbers in each ticket right, you stand a very good chance of making more than £200. But how to calculate the spread? Well, I had a think with my computer scientist hat on, and I came up with neural networks.

For those of you who don't know, a neural network is basically a way of making an analog learning system. You feed it a set of data and a set of results from that data, then you tell it to go extrapolate*.

So what are people's opinions on this, and mathematically targeted gambling in general? I've also been toying around with several spread betting algorithm ideas, like predicting the currency market and stuff. Obviously, the best way to test these out is to use the vast amount of data available (previous lottery numbers, etc.) and plug the numbers into a spreadsheet or similar.


*Well, this is one way of using neural networks. It's actually a rather vast subject.

I had my thesis on the topic of artificial neural networks, yes they are a vast topic and their uses can be numerous if you can find the right way to create training sets...

the idea of using Neural Networks to predict lottery numbers is not to be dissmissed, it has been used successfully in predicting DNA sequences and that is way more complicated than predicting a number. the process of learing in the neural networks is the hard one as you haev to create training sets to get the results you need,

in the case of lottery, if you are buying a ticket with preset numbers then it is unlikely to use neural networks on that ...you're not predicting here, you're given a ticket with a number that is already chosen for you...and even if you want to predict the drawn numbers...you still have to count for several factors regarding the medium of draws, is it balls, or wheels, or random computer numbers...all these matter.

there is the OTHER lottery types, in which you chose 6 numbers out of 49 numbers, and then a draw is made using balls. this one has something called trends and statistics of the most frequent numbers drawn. if you want to apply that on neural networks you're getting closer to something doable, but not easy !
1- you need to get ALL the winning numbers in all the draws and watch how many times they showed up in the course of a year,
2- check how many times these most frequent numbers appeared with less frequent numbers but still highly frequent, and so on
of course a computer can make these calculations much much more faster than a human can do. and that can become your first raw training data ..and you're still not done...you have to put that raw training data in a format that can be - understood - or - learned - by the ANN ...and also how many neurons you want in the network.
you have a lot to research my friend....if you have at least the time to do this, you'd be on your way for being the first to defeat the Lotto odds...and they ar way over 24 million potential combinations. but only 1 set of numbers can be the 1st prize winner...

wanna work on that ? i suggest some reading on the ANN...it is worth it i assure you...try a quick ANN to get an idea what it can do ...here http://leenissen.dk/fann/
fiddle with it get to know what training sets are and how they are created and based on what criteria....remember ..neural networks are not like the human brain, even though they were invented to simulate SOME not all functions of a brain neuron.... so what we need to do is to find a way to train your ANN to predict, based on criteria...so can you come up with a training set for this ( red, white, and ....) usually people who have knowledge of this sentence will automatically say, blue...because they heard it before and thus it is the most logical and accurate...aother training set for numebrs for example if i say what comes after ( 1, 3, 5, 7, .... ) ... of course anyone would say 9 ...and that is what we want to do ... we want to see the trends from the winning numbers, and then create a net of numbers combinations. Good luck .

[Kasuha]

the idea of using Neural Networks to predict lottery numbers is not to be dissmissed, it has been used successfully in predicting DNA sequences and that is way more complicated than predicting a number.

The important fact here is, DNA sequences are far from random.

[shiraz]

I think to get it to work with NHL, NBA and MLB is to only use the last 5-15 games played or so. I'm only starting to experiment with this, but it does make the most sense to analyze them that way. Effectively taking advantage of the streaks of teams. In NFL since there are so few games it doesn't matter, but for the other sports where they play more than 80 regular season games it does.