hypersoniq's Blog

So, we will be looking at a 6/49 game, the PA Match 6. Needing 2 number sets to start with...

1. The most frequent by position

2. The most frequent by any position

3. (The dumb part). For each position select the number that is between them... that would be for example: IF the most frequent by position is a 1 in the first column, and the most frequent overall (after sorting) is a 5, then select a 3 because that is the same distance. BUT... what if the numbers do not have a middle? Such as 1 and 4? Easy... for the first 3 positions that the one higher... eg 1, 2, 3, 4 would be 3. For the highest 3 take the one lower, such that 45, 44, 43, 42 you play the 43. IF the number is the same, play that number... also if they are one apart, such as 39 and 40 in the 4th position, play the lower number (39)

Sounds too simple, and it surely has no scientific basis, you could use any 2 lines, such as the last 2 draws...

In my case, I am using numbers derived from frequency, but since we are forgetting the math and making up a coincidental system from scratch... why not?

I saw something on a Netflix spy show recently (Night Agent) where they were planning a break in and the advice given was to start with the stupidest plan and refine from there...

It is no more or less valid than any other workout system, and could be applied to any game with 6 drawn balls, such as the PB or MM.

Let's hope for dumb luck!

Updated the Match 6 file, last entry was October 10th, 2024...

Mid and eve PA pick 3, last updated Dec. 9th 2024...

Even though I am not actively working with these files, every now and again I will update them.

Working today with the graphic part of the QP idea, plotting a grid, putting numbers on that grid and making the numbers selected plot a point in the center of their respective grid spaces. After that, will draw the triangle (or line, in the case of the same number twice), plot the center point of the triangle formed, and connect it to the actual draw result grid space. Thinking blue for the triangle and green for the connection line.

Once I have gathered as much data as possible (line lengths, angles, etc) in a summary report printout, then I will begin a brainstorming session with AI... only with actual example data in hand.

The data entered could easily come from game history as the last 3 draws. Not only would no changes be required, but it could then be back tested... minimal code redesign to read from a .csv rather than enter manually.

I believe that I had mentioned looking at trigonometry earlier... this is the implementation of that.

The first idea that pops into my mind when trying to imagine how to display the collected data is to draw a grid to mark the 3 quick pick values in each game.

When deciding a grid layout, the 7 x 10 grid from the PA lottery kiosk seems sufficient.

The bonus balls could be contained in a 7 x 4 grid, also as used on the PA lottery kiosk, but it might be more useful as a 6 x 5 grid.

Here is the initial goal in Python...

1. Draw the grid containing all of the game numbers, lowest in the top left, highest in the bottom right. The grid target values are the dead center of the squares, and the dimensions are 1 x 1.

2. Place a visible dot in the center of the 3 quick pick values.

3. Calculate and draw a triangle (or a line if two of the same number are given) from the center of the grid squares containing the QP values... line 1 is value 1 to value 2, line 2 is value 1 to value 3, and line 3 is value 2 to value 3.

For this triangle we want to record angles and line lengths, and the point representing the dead center of the formed triangle

4. Plot the point in a different color for the winning number in that position. Grabbing any relevant measurements, such as line length from the center of the winning grid to the center of the triangle.

5. With this data collected, begin the conversation with generative AI to begin constructing a formula to describe both the relationship of each triangle to the winning point, and also the process of creating a generalized formula to solve all data points simultaneously. This will be the time soak...

6. Once there is something to work with (good or bad), grab 3 QPs for a draw and attempt to apply the generalized solution to obtaining the values of the next draw.

Certainly not ideal, as step 5 may never be truly completed, but at least it is a starting point!

Where to begin to use 3 QPs to reverse engineer the winning combo?

The first idea is to solve simultaneous equations. There are 3 well established methods for this

Elimination

Substitution

Graphing.

We will start with graphing, as this gives a visual as well. With 3 data points, we can plot these in 3d space for each set, then where lines intersect we should find a solution... IF lines intersect...

Our sample set is limited to one week of data for both games, so we have 5 sets. It is with this small sample size and huge possible outcome list where we start.

What is known is the 3 QPs and the result of each of the draws.

Remember, there may be absolutely NOTHING here, but it seems like a good exercise in trying to solve an impossible problem.

After collecting QP data for a week's worth of Power Ball (Monday/Wednesday/Saturday) and the Mega Millions (Tuesday/Friday) from 3 different locations, I can see how truly bad QPs are. Not one set of 3 QPs managed to pull in more than one white ball match, and zero bonus ball matches...

Obviously One Arizona player got lucky in the MM draw last night, and looking at the payout, only 9 PA players managed a 4 out of 5 without the bonus ball...

And so it begins, the analysis part...

Step one, use the 3 QPs for each draw to "reverse engineer" matching the winning numbers for that draw... whatever that looks like. Repeat the process for each of the 5 draws for which I have collected data.

Step 2 will be to look at the ways in which the data points were manipulated to produce the winners and look for any common factors so that these steps might be applied to a more general process or formula. Here is where generative AI will be used. To both explore the concept and to generate starter code for a program that takes as input 3 QP lines and applies processing to generate one "best guess" pick.

Step 3 will be to test... getting 3 cheap QPs for the draw early enough to apply the process and then play a 4th ticket for the same draw with the processed result.

I do not have high hopes for this method, and it falls primarily under the category of "no stone unturned". This could, in theory, provide a seed for the vertical horizon system that works better than using the most frequent overall numbers... or not.

The up side is that all other play is suspended, so there is a small benefit financially.

Happy Coding!

First draw with 3 QPs on the Mega Millions and the result was just as bad. Only one white ball matched, #49, but it was on 2 of the 3 tickets.

Last gathering of QPs for Friday's MM draw then the work begins.

Because this is the first dive into the concept that the QPs are less than fair, expectations are low.

Because the end result system would require purchasing 3 QPs before selecting one line, this already violates my long held belief that one ticket is enough... it is however somewhat cheaper because either game still has an option to NOT play a multiplier, therefore either game can be played for less than $10 (3× $2 QPs and one $3 "loaded" ticket).

It may take a long time to develop something that results in a playable line.

I am not questioning the randomness of the results, let's be clear on that. The results of all draw games analyzed so far are close enough to expected distributions to be considered random enough, even though they all fail Chi Square tests. That has more to do with the number of combinations drawn with a matrix vs the number of un-drawn combinations that will never be seen in a million years...

I am questioning the randomness of the quick picks! As I purchased my last set of QPs for Monday's draw, I looked to see how truly poorly the QPs performed on Wednesday and Saturday. Wednesday's 3 QP tickets had 1 match (43, a white ball) and Saturday matched zero. I expect similar for Monday. These quick picks did so bad, they would have had the sole 43  match if counting all 6 lines against BOTH draws.

So, what could the premise be here? The Quick Picks are intentionally bad! But how could they all be so far off if they had no idea of what the winning combo was going to be? Could their greed be so great as to send out intentionally bad QPs to force the draws to roll over and increase sales? Because of the randomness of the draw, they would not need to do this but greed is an all powerful drug that nobody is immune to... if humans are involved, the chance of interfering to maximize profits is non zero.

So the premise is, can you buy 3 quick picks and use that data to reverse solve the draw? Can how bad the QPs are be the guide to a better pick?

Why 3 QPs? Why not 1 or 10? Triangulation! Using the data points that are possibly intentionally so far away from the winner to figure out if there is any repeatable truths hidden in the data. 

Notice how the QPs get better when the jackpot gets hit... or in the weeks leading up to the big win...

The low jackpots seem to bring in bad QPs because somebody has to pay for the roll overs... The premise is not as far fetched as it sounds.

I bought 3 QPs for Wednesday Night, matched 1 white ball. I bought 3 QPs for Saturday, will wait until that draw goes off, then will be buying 3 more for monday. QPs don't need the multiplier or anything, so the whole data gathering phase will cost $18 using $2 tickets.

After gathering the data, it will be time for analysis...

The goal? Reverse solve each draw based on the data points from the QP tickets. Will it be impossible? Probably. Will it be repeatable? Most likely will be non repeatable. However long the analysis phase takes (weeks, months, the rest of 2025) will represent savings because all other systems will be suspended.

Why try it? Well, lack of any better ideas, boredom and the challenge... what if? What if it is possible to use 3 QP lines to "triangulate" your best guess? Any jackpot system needs only to work 1 time... the thought of coming up with a system for once that does NOT require studying past draws is enough to green light the project. With a total initial data acquisition cost of under $20,  why not?

May you all find something that works this year, this is my "big idea" for 2025... and it is not even that good... ever forward.

No need to even retool the software, all I need to do is add the combo column to the .csv files!

Only the pick 3 evening has enough data for this, too many zeroes in the pick 5 games... a little over 5% of the possible combos have been drawn, but in pick 3, the minimum repeat is 5 for all 1,000 combos.

Which ever of the holidays you celebrate, may it be a joyous one!

Since I put all that time into creating follower programs for MM and PB... and they each generate just one pick... why not try to play the generated lines for BOTH games? It makes no sense and has no statistical foundation, but then again NOTHING appears to work at any level, so why not? May not have time to run everything until the weekend...

Staying on the 0 1 5 for both PA mid and PA eve. Will do a weekly update to see if that combo changes, but it is pretty far ahead of the nearest competitor.

Results so far less than stellar, but a box hit almost pays for 3 weeks of continued play, and if they slip up and let it out straight, that will allow a live test on one of the the jackpot games...

Otherwise out of ideas for the rest of the year...

Here it is, the most drawn combo across PA pick 3 day and night... 0 1 5

That is what I am going with for the rest of the month. With permutations counted, this combo came up 182 times across day and night draws.

0.50 straight / 0.50 box

With 22 days left in the year, that is 44 draws at $1 per ($2/day), so total cost of $44... that is the plan for closing out 2024. One single box hit would still represent a $4 loss, but that seems worth the effort... and it gives me 22 days to come up with a new direction for 2025.

So, the weekly combo for the PA pick 3 eve and PA pick 3 day both expire today. I still need to sit down and update both to see if anything changed. So here is the plan for the upcoming 7 days (starting tomorrow). When updating, I will be extracting the top 5 combos for each game and setting them aside for paper play for the week. The combo that will be played will be one for both mid and eve that results in combining both p3 data sets. The cost is the same ($14) and paper playing the pick 3 is not an issue should those combos start showing up, at most I would miss out on a $290 hit.

This will allow testing 2 different approaches in the same week, without incurring any additional expense.

The hope is that over 26,000 data points will be better for picking a combo. But then that is it, one combo, both mid and eve covered.

Attempting proof of concept on the pick 3, with the hopes to extend to the pick 5 if promise is shown.

The only additional work would be a combined spreadsheet and .csv file to handle the merge, the python script will require zero modifications.

The fact that I am at the point of merging 2 separate history files might just be an indication that I can finally admit to myself that, though it might be possible, I lack the skill to crack the code...

A best guess is still just a guess, but the idea that what happened the most should indicate what is most probable to happen again is a difficult notion to shake... and that is what these state lottery administrations are counting on.

So I am still here, still innovating and still waiting for them to slip up and catch a win... one that will pay for my final statistical tactic to have a shot at a jackpot game. I just have the patience to wait until they fund it!