Lottery Prediction by Recurrence Analysis
March 28, 2005
|Posted: June 7, 2005, 11:26 am - IP Logged|
OK!!! Let's put this thing to the test! I am going to compare this thing to random numbers selected straight from random.org to see which can predict the actual numbers more accurately ... or better so to say.
I ran this analysis against the Minnesota North Star Cash drawing (5/31). I used the updated data and here is what it produced:
Dim = 17, Delay = 5: 8,4,7,20,30
Dim = 17, Delay = 4: 3,10,17,21,27
Dim = 16, Delay = 5: 8,4,15,20,28
Dim = 16, Delay = 4: 3,15,10,21,26
Dim = 15, Delay = 5: 8,4,23,20,28
Dim = 15, Delay = 4: 3,15,10,21,26
Dim = 14, Delay = 5: 8,7,15,20,28
Dim = 14, Delay = 4: 1,15,10,21,26
Dim = 13, Delay = 5: 8,7,18,20,28
Dim = 13, Delay = 4: 3,10,21,21,29
Now, feeding in the correct information to random.org, here are the random numbers that I received:
Now, let's see which system gives the most hits, thus the most money in return. This test should be interesting and is very exciting!
Tune in tomorrow to see which system wins!
From last night's test of this system versus random numbers picked from random.org in regards to last night's Minnesota North Star Cash drawing, here are the results:
Numbers Drawn: 1,8,17,19,29
Dim = 17, Delay = 5: 8,4,7,20,30 = 8; $0.00
Dim = 17, Delay = 4: 3,10,17,21,27 = 17; $0.00
Dim = 16, Delay = 5: 8,4,15,20,28 = 8; $0.00
Dim = 16, Delay = 4: 3,15,10,21,26 = None
Dim = 15, Delay = 5: 8,4,23,20,28 = 8; $0.00
Dim = 15, Delay = 4: 3,15,10,21,26 = None
Dim = 14, Delay = 5: 8,7,15,20,28 = 8; $0.00
Dim = 14, Delay = 4: 1,15,10,21,26 = 1; $0.00
Dim = 13, Delay = 5: 8,7,18,20,28 = 8; $0.00
Dim = 13, Delay = 4: 3,10,21,21,29 = 29, $0.00
The results from random numbers from random.org are:
10,26,23,18,12 = None
25,29,13,6,7 = 29; $0.00
20,13,29,11,31 = 29; $0.00
23,20,13,7,30 = None
2,22,29,26,19 = 19, 29; $1.00
30,28,3,14,27 = None
5,29,13,17,14 = 17, 29; $1.00
31,13,5,12,26 = None
9,27,3,1,4 = 1; $0.00
17,23,22,31,15 = 17; $0.00
As you can see from this test, pure random numbers beat this system by $2.00. That's not saying much though. If you look at the picks, you will see that this system gives fewer picks with no numbers drawn in comparison to random number picks.
What I can take from this test is that the visual recurrence system in regards to lottery picks needs to be refined ...... A LOT. I guess what I am referring to are the Dimension and Delay choices. These need to be refined and standardized so that they pick winners per input. An example would be to test a load of PICK 3 drawings and see what combinations of Dimension and Delay work best.
If anyone has loads of time on their hands, please test the Dimension and Delay combinations against PICK 3 draws and see which combinations work the best.
To close, I believe in this system! Even though it lost out to random number picks, it still has a ton of potential in this realm. If Dimension and Delay combinations can be discovered that pick winners for PICK 3, 4 and 5, then you would have an awesome system. The potential is there! LET'S UNLOCK IT SOON!!!
BOW WOW WOW ......
...... YIPPY YOH YIPPY YAY!!!
June 1, 2005
|Posted: June 7, 2005, 12:51 pm - IP Logged|
Thank you for those results, Atomic Dog. It confirms what I've discovered myself. Consistency is the key. It's just a matter of gathering more data. Now that the UK lotteries run more than once a week it doesn't take as long to build up data as it did a few years ago. I think I'll turn my attention to the main draw for a bit (6 from 49) since there are over 900 results to analyse.
June 1, 2005
|Posted: June 7, 2005, 7:59 pm - IP Logged|
Just another thought, Atomic Dog: with Delay = 5 try lower values of dimension, say 10 to 15.
May 14, 2005
|Posted: June 12, 2005, 1:30 pm - IP Logged|
I have been working with this VRA system for a bit now and I have a question for you. In your analysis for "Thunderball" Draws 399 through 449 are you analyzing each digit separately, i.e. first digit of the draw on one line at time, and then the second digit, and so on? Or are you using another procedure which lets you do all five numbers at once ... that would be nice.
May 4, 2004
|Posted: June 12, 2005, 3:39 pm - IP Logged|
As far as I know he does one number or "Pair" at a time, such as 05 being 1 number or "Pair" and also such as 27 being also 1 number or pair, that is 1 column at a time as far as I know, but I might be wrong.
June 1, 2005
|Posted: June 12, 2005, 4:26 pm - IP Logged|
Currently I try and predict all numbers for each draw. I could try a 'columnar' approach, but I suspect it might not work as well, since the numbers aren't drawn like that. My approach tries to preserve the time aspect.
May 4, 2004
|Posted: June 12, 2005, 9:35 pm - IP Logged|
True, but also, while all the main balls come out of the same "Container" if they are not computer drawn, they come out one at a time, from first to last, so we might say that in a way that there is one sequenced column, a horizontal sequence of numbers or balls intead of a vertical position column, so the history of the main numbers is many broken or displaced sections (One on top of another) of one very long horizontal column, the first number of the new or last draw is the next one in series after the last number from the second last draw and so on.
It is a relative thing more or less.
June 1, 2005
|Posted: June 13, 2005, 12:57 am - IP Logged|
Quite true. However, the whole point of recurrence analysis is that it works on time-series data. I'm trying to preserve any time-related structure in the results because that's what matters. In fact, to be even more faithful I should take the numbers in the precise order in which they are drawn (so far I've been examining results that have been placed into ascending numerical order after they've all been drawn).
Of course, looking at them in columns might also reveal something. It might even work better that way. Any patterns would be different, meaning the model parameters would be different. Again, I'll get round to it when I can.
Lantern, I'm still having trouble with the ABM software. I've got it to run, but now it says the data is missing a comment, and the help file is corrupted. I guess it just doesn't like my PC! But the free version can only use a 100 numbers it wouldn't be much use anyway.
Saint Louis, Missouri
June 9, 2005
|Posted: June 13, 2005, 12:59 pm - IP Logged|
I have actually used a technique similar to this for picking the GA Fantasy 5 numbers. On May 23, 2002 the numbers which fell were 07-09-14-20-32. I used a website on infospace to get most of these numbers. This site listed combinations of two numbers which fell together often as well as triple numbers which fell often. I used that months draws to come up with the combination of 20 and 32. Then from there I checked which triple numbers came along with that set to get the 07 and 09. These numbers not only fall with 20 and 32 but they also fall together a good bit. In fact, they fell in last nights Fantasy 5 draw. Unfortunately for my last number I picked 18. This is my birthday number. So there was no analysis involved for the 18. I was just about ready to go ahead and buy all combinations with 07-09-20-32 but for some reason I didn't do it. I have hit several times for the 4 out of 5 and 3 out of 5 using this same method. Unfortunately the site that listed these combinations died so I haven't even hit 3 out of 5 for about 2 years now. I have looked all over the web for something like it but have had no luck. I would love to download this software but alas our computer uses Linux and I don't have the technical know how to build a spread sheet that would do this. Alas! -Bonnie
What would the 4 numbers have been worth? I think that is what I am coming to grips with, what hits more often versus what pays out the most per "investment"
Great site folks, great reading
If the eyes are the windows to the soul, let me see you without the shades drawn
What's a Hades Bunny? Well let's just say you don't wanna be jumpin' over MY back fence. But hey, you ain't afraid a lil white rabbit now are ya?
June 1, 2005
|Posted: June 13, 2005, 1:29 pm - IP Logged|
Summary of Results.
For those who have jumped to the end of this thread first, here's a rundown of what I've been doing. Recurrence Analysis (RA) is a statistical method for analysing trends or patterns in noisy, chaotic, non-linear time-series data. I thought of applying it to lottery numbers.
Using a program called VRA (go to http://home.netcom.com/~eugenek/download.html to download the latest version), I am searching for the parameters that will give reliable predictions for at least three numbers. For the UK Thunderball draw (five main balls), the current best values are:
using Nearest Neighbour predictor with Euclidean measure. There are now several UK lottery games, and I'd like to find parameters for them all. How the results are ordered is an important factor. I prefer to keep the arrangement close to the original draw order, but a pattern is a pattern, and RA should find it regardless. Numbers can be looked at by column, by draw order, or by ascending order.
I'm going to set up a web page with all my results, ASCII files of UK draws, links, et cetera. I'll add screenshots of VRA for those who are still struggling with it (a picture is worth a 1,000 words...)
Do the lottery numbers have a strange attractor? It would be nice to see all those numbers expressed as a function.
A computer simulation of a lottery machine (really only an academic exercise rather than serious predictive tool). A first approximation would be to model just the motions of the balls. You would need a routine for each ball, to handle its x,y,z coordinates and velocity. The physics engine would call each ball routine in turn and look for collisions, updating each routine accordingly. I suggest a time step of 1/100 of a second. The program would look something like this:
1 Initialise variables: time, positions, velocities, ball count
2 Set balls in motion; start timer.
3 Call each ball subroutine in turn; miss out drawn balls
4 Increment timer by 1/100 second
5 Draw a ball after a specified time; decrement ball count
6 Repeat from 3 until all required balls drawn.
Recently the Millennium Consortium ran a simulation of the universe on one of the world's largest supercomputers, so a few lottery balls should easily be within the power of a high-end PC to simulate. A more advanced simulation would take into account the size and shape of a lottery machine, and whether the balls are agitated by a wind blower or mechanical arms.
May 4, 2004
|Posted: June 13, 2005, 11:31 pm - IP Logged|
I guess then that if you are still having trouble with that software, just forget about it, the VRA is probably better and cheaper any way.
If I was a computer programmer and if I was good enough to do it, then I would make a program or programs that would attack the problem thru all sides, it would maybe evaluate digit's relationships in multidimentional ways, I am of course talking about the history of the winning numbers.
June 1, 2005
|Posted: June 16, 2005, 10:17 pm - IP Logged|
After three days of coding, my site is finally up:
Obvously it contains everything I've explained here so far, and a a little bit more.
May 24, 2004
|Posted: June 17, 2005, 12:42 pm - IP Logged|
I've been using a homegrown recurrence analysis system for the last 2 months. Can't say it's gotten me any closer to the numbers yet, but good luck!
June 1, 2005
|Posted: June 17, 2005, 4:52 pm - IP Logged|
Good luck with yours too. Of course, if it works you won't need luck...
December 10, 2004
|Posted: June 17, 2005, 5:15 pm - IP Logged|
Thanks for setting up that website. It gives a good overview on how to use the program.
If it proves it can do the job it will be worth buying the software.