Hi all,
I am watching quietly LP for a long time now and I am an active researcher in lottery subjects.
Have done all the usual drills, such as passing from neural nets to ARIMA models and from statistical methods to randomizers...
A brief conclusion so far from me: The main reason these methods fail is simple, known and let me put it in a simple yet intuitive way which will help me later to present my latest finding:
(useful remark: for practical reasons I will refer to a 5/45 Lottery game, but of course my method applies to all types)
Let's say that you have a history of draws of your local lottery.
Also let's assume that the history goes like this (numbers don't matter just get the idea):
3-5-12-17-24
1-7-12-29-34
....
...
18-23-24-35-41
and you apply your "super" method to derive your next draw from your history.
BUT! Assume that another fellow somewhere in the universe participates in his/her local lottery and until now has observed the SAME draws history as you.
Well, the problem is that from the SAME history the two lotteries (one here and the other on Jupiter...!) may produce totally different next draws.
See the problem of lottery forecasting? If the two fellows use the SAME method on the SAME history files they are going to get the SAME next draw as a prediction.
But the next draw will NOT be the same for the two different games no matter they have the same history!
That's the reason that ANY method will fail, UNLESS your prediction IS NOT based on your lottery history...
But that's my story.
The reason I am sharing this is to enrich the discussion with new ideas, maybe crosscheck results and get some input from community.
Let's start.
What is a linear congruential generator?
Well is a simple method to produce pseudorandom number sequences. Meaning, it is a formula that you can use to get a series of numbers that they LOOK like random BUT they are not. They are PREDICTABLE as long as we know the exact parameters of the formula.
What I am using is this:
(NEXT NUMBER IN SEQUENCE)=((MULTIPLIER*PREVIOUS NUMBER IN SEQUENCE)+SUMMAND) MOD BASE
Ok, so how I start as the first number has no previous...?
Well I need another number which I call SEED
MOD is the usual modular operator and simply means this: 9 mod 4= 1 as if we divide 9 by 4 we get 1 as the remainder.
Conclusion ONE:
If we know just FOUR numbers: MULTIPLIER, SUMMAND, BASE and SEED I can produce with my formula above a pseudorrandom sequence of numbers.
Nothing new so far as this is everywhere in Internet, math books etc.
BUT...
WHAT IF I TRY TO FIND MY FOUR NUMBERS (MULTIPLIER, SUMMAND, BASE and SEED ) IN SUCH A WAY THAT MY NUMBER SEQUENCE PRODUCED FROM THE FORMULA IS SIMILAR BUT NOT EXACT TO MY DRAWING HISTORY???
Let's go back to my hypothetical history file and let's also imagine that I have also found my four numbers which produce the following:
HISTORY
3-5-12-17-24 and my SEED IS 10 (FIRST NUMBER IN SEQUENCE)- (NO HIT)
1-7-12-29-34 SECOND NUMBER 12 (HIT)
....
...
18-23-24-35-41 MY N-TH NUMBER IN SEQUENCE 19 (NO HIT)
and so on...
Conlcusion TWO:
I can find a well-defined pseudorandom sequence of numbers which can have TWO important features:
- The sequence will resemble BUT NOT be the same as the actual draws history
- The sequence is defince by four numbers which I can search through the combinations and find some good fits.
FINAL REMARK
Obviously the next number in my sequence will be the prediction.
So, what do you think?
If you care I can elaborate more, but I need your input and your feeling on this concept.
Thanks all!