Just for fun
I have created a javascript file that lets you choose a past result file it reads all numeric entries separated by spaces.
And then with a Quick Pick button, it'll random choose 6 numbers from the read entries making sure they're all unique in order for you to play the Canadian Lottery 6/49.

The idea here is that if the there is slight flaws in the lottery balls that causes some numbers to occur more often than others, we'll have a slight (by slight i mean very very little amount) chance that our randomly choosen numbers is based on past results.

To use, just save the lotto649.html (this is the javascript file) and 649.txt (this is the past result file) to your computer.
Then open up lotto649.html in your browser,
click "Browse" and browse to the 649.txt file to load up the past result entries.
Then click on "Quick Pick" button to get a quick pick result (this can be done repeatedly to get different numbers).

If you do end up using this Quick Pick javascript to play 649 and end up winning big, don't hesitate to share your winnings with me

I'd imagine the Quick Pick would be better if there was more data but all i could find online was 1 year's worth of data (Sept 2013 to Sept 2014)

Here's what the html looks like when it produces Quick Pick results.

It would be really easy to alter the javascript to make a Quick Picker for Lotto Max.
Shall i do it?

Well, I did one for Lotto Max too.

a page called "lottomax.html" and uses the result file "max2014.txt" of 2014 (data from Jan 2014 including max million draws as well).

I don't know how far back i can go with the data, since i don't know when they might have changed to a new drawing machine.

You don't have to use my result file, you can just create a text file with your favorite numbers occuring more often and use the javascript file to randomly choose from your own numbers if you like.

here's what the lottmax.html looks like.

When you win the lottery Tran, let us know. lol

When.. for sure but chances are very slim
since if you play once every draw all your life (from 18yrs of age to 80yrs of age). You'd have to reincarnate thousands of lives before you have a chance to win once at the jackpot.

a Powerball quick picker for American friends


use the file named "powerball.html"
browse to one of the "whiteballsXXXX.txt" file for the first file (to load up white balls results)
browse to one of the "powerballsXXXX.txt" file for the second file (to load up power balls results)
click on "Quick Pick" and you'll be shown a result like below


I have 4 data files
"whiteballs2014.txt" to be used along with "powerballs2014.txt"
"whiteballs2010-2014.txt" to be used along with "powerballs2010-2014.txt"

Current Estimated Jackpot for powerball is $127 Million

I don't know if the method did make a slight improvement or not but
my uncle bought 12 tickets and we won 2 free tickets so our winning odds was 1/6.
It's a slightly better than the 6/49 posted odds which is 1/6.6.
It could just be fluke.

here's the updated 6/49 result file that includes last night's draw

Cool; promising then.

chances is are it's just fluke, i'll have to observe see if that odds stays consistent in the future.

If anyone else plays the 6/49 or the lottoMax using these javascript files,
if YOU COULD PLEASE post your winning lines(of any prize) count and your lines bought count here for example (with my result so far)
6/49
bought 12 lines
won 2 lines
so i can add them up and find out what the odds of winning is using the method and see (with more confidence) that it is better than 1/6.6. Also, if you post stats of lottoMax, only post counts of actual chosen lines by javascript not the 2 free lines given by lottoMax that accompany your bought line, since that will mess up our actual stats

or if you're American and play the powerball why not post the result count too? just to see if this method works or not. it's all in good fun if lots of people do it, since we don't have to spend much individually, if one person or few people do it it'll take a long time or a lot of money. if many people do it we should have some decent stats.

I took another step and emailed Lottery Center to ask for the date when they last changed machine(s)/lottery balls.
And how i can get earlier results before August 2013 (just in case if they changed machine(s)/balls before that date)
to get a more accurate probability representation of a result file.
Hopefully they'll email me back with all the needed info soon

Here are some fun diagrams of what the method simply does

First the good news.

Proof of concept works, this method does in fact improve your odds if the machine/lottery balls combination is flawed in their probability distribution.

I have create a test page (called "theoryTest.html") which takes a file ("theoryTest.txt") with entries that define a number followed by frequency of occurrences on each line to define a probability model. You can download these files and change the theoryTest.txt file with different numbers or frequencies to test for yourself.



For proof of concept I am only making it selecting 2 numbers (instead of 6 or 7 for 6/49, lottoMax respectively).
I make it quick pick 2 numbers based on probability model (defined by input file) (let's call this probabilityModelQuickPick).
select a random result of 2 numbers based on same probability model (let's call this the drawnResult).
then select a random quick pick of 2 numbers based on an even distribution model (let's call this evenDistributionQuickPick).

I ran the test repeatedly (32000 times) and way more often than not (I did run into a fluke case where it's not) the probabilityModelQuickPick has more match counts than the evenDistributionQuickPick. You can change the "theoryTest.txt" file, to give it a different probability model.

Now, the bad news.

I ran into some sites that mentioned that the Lottery Corp. often change their lottery balls/machines and sometimes even have different sets of balls and sometimes even swap balls between sets before their draws.

Wishful thinking.

The balls are probably made by same manufacturer and if they're a little flawed, and that the machines that make these balls haven't changed then any flaw is repeated over time so that even if they change new balls, these flaws are present and overall, the average still gives us some sort of consistent probability model.

with this wishful thinking in mind, i have created 4 more files.
"lotto649frequency.html" which reads "649freq.txt" (result file of number occurrences since games's inception)
"lottomaxfrequency.html" which reads "maxfrequency.txt" (result file of number occurrences since lottoMax's inception)
all these files are included in the lotto.zip (above).

More details on the proof of concept.
Since our theoryTest.txt has 7 numbers defined and we're only choosing 2 numbers the odds of getting a match of both numbers is 1/(7!/2!/5!) = 1/(5040/120/2) = 1/21 ~ 0.047619
0.047619 * 32000 = 1523.808
So out of 32000 draws we should have about 1524 matches.
Here's an excel sheet of actual test results.

By using the defined probability model in theoryTest.txt as an example, we've made a 4.65% increase in our odds of having a match.

I have observed that the more biased (the more uneven the distribution is), the better the probability model quick pick works while the even distribution quick pick stays about the same. Here's an observation on a probability model that is much more biased.


update:
bought another 6 lines of 6/49 (using past year's data) and won nothing so my winning odds are now 2/18

Great news,
The Lottery Corp called me at 4:00am (probably 7am their time since they didn't know where i was from).
They told me they re-painted their machines in 2009, and they changed balls since 2008. They weigh the balls once every 6 months making sure they're the same weight. They have 2 sets of balls for 6/49 and 2 sets of balls for lottoMax.

So i did a javascript test to prove that randomly selecting from a probability model (that is the sum of 2 different probability models to represent 2 sets of balls) givevs me better odds than selecting from an even distribution model.
So now, i just gotta manually grab data from from 2009 up until now and i should have a pretty good probablity model.

I added LOTTOMAX.csv to be used with lottomax.html (data since Sept 2009)
and 649.csv to be used with lotto649.html (data since Jan 2009).
also changed the theorytest.html to read 2 probability models (theorytest.txt separated by the string "set2").

I also created a "theory649Test.html" (which uses "649.csv" as input) and simulates 660,000 draws each run and returns winnings (any prize) of the "Past Result based model" and the "Even Distribution model". results shown below

You can download the files in the below lottoSim649.zip below and try for yourself.

I am unsure of why i am getting different ratios than the posted odds of 1/6.6 by the Lottery Corp.

I found out where the 6/49 got the 6.6 odds.
By having all the balls perfectly equal.
So if all the balls are perfectly equal, their result comes from an unbiased probability model and so if you choose your numbers randomly from an unbiased probability model, you'll get the 6.6 odds. Here are the results, if the balls are perfectly equal, selecting from a probability model based on past results will actually give you worse odds than from an evenly distributed probability model.

The 6.6 is the selected cell, well actually closer to 6.625
it all makes sense now.
The real decision to be made here is: if you believe that the balls are always perfect and all equal then you should buy quick pick from the store since i am guessing they're picking from an even distribution model. If however, you believe that the balls are slightly flawed in the real world and some balls will occur more often than others, then using this the past results probability model is better.

You are so mathematical, I have no idea how you can figure this all out.

I didn't figure it out,
I just ran some simulations and let it randomly pick numbers, and if you run it enough times it gives you stats that you can be very confident with. that's all.