hypersoniq's Blog

because I hold each position in a pick 3 or 4 game as a separate entity,the OOP in LoopRS does not apply, hence the abbreviated LRS designation.

Since I must do each position separately, I am probably going to go back 10 draws at a time (side by side) to speed things up a bit. Just have to modify the formulas a bit so I can have everything on one sheet for each game.

LRS3 and LRS4 will be one ticket straightshooter systems, not backtesting box hits this time. no run-downs or workouts... I get what I get.

The LRS system will be backtested then put to "paperplay" on the LP predictions page for awhile. My biggest time killer is having to reconstruct the pick3 sheet from scratch. I have from 1977 to 2005 so far.

Based on the performance of the Cash5 backtest, I kind of have higher hopes for this pick3/4 system than normal. Heck I might even put this into actual play if it tests well...

BTW my LoopRS sheet is calling for #30 to fall as the red ball tonight in PB, let's see how that goes.

Also, sometimes in an OOP system,you get duplicate numbers in a pick... I resolve this by filling inthe bet slip with the unique numbers I have and also checking the QP box,it fills in the rest... 

My LoopRS excel project gets it's numbers from previous draws. I chose this method because it incorporates the same randomness as the game itself... you're just using the most frequent positions.

Ever notice how great deals of data for a single game kind of show trends year to year? regardless of the machine changes or ballset changes? I believe it has to do with that particular game... Each game has it's own "random identity". For example,the #'s 20, 31 and 35 in PB... in 10 years they remain the hottest... and it is spread out over the 10 years, even though they change machines and ballsets every draw!

I have traditionally dismissed coincidence and looked upon randomness as a problem to be solved... this time I am embracing randomness AND coincidence. Instead of trying to beat the game with some formula or equation, I'm simply letting the past draw history pick my next draw. It is a VERY tedious process to keep going back draw after draw for each position, but I don't want to automate it, because I feel the need to SEE the system unfold one step at a time (like stepping thru a debugger).

Seeing this unfold for some reason makes me think that it really doesn't matter what machine the balls are drawn from, somehow each game has it's own identity and keeps it consistent... like how 2 states can have such widely varied pick3 draw histories.

I have tried for years to BEAT randomness, now I'm gonna relax and JOIN it for awhile ;-)

Everything in nature is trying to find balance, from us as people to the cells inside of us to the atoms everywhere...

I'm stuck on LoopRS development now, but the next one on the back burner will be based on the concept of Homeostasis

perhaps instead of looking for the "most frequent" or "least likely" we should be seeing what draw results will tend to balance out the draw frequency...

I didn't ditch PROfile, because there is much valuable data there that I just haven't figured out how to use (or backtest).

I haven't yet envisioned the form this "balance" system will take (other than an excel file), but the possibility of truly understanding the nature of the numbers intrigues me!

comments? ideas?

LoopRS backtest of PA cash5 would've paid $1 had I spent one to play... still have several thousand draws to sift thru before live tests.

Powerball RB should be sifted by this weekend, hopefully giving me something to play regularly besides the QP,that's not workin' out these days :-(

White balls should be done by summer... till then it's 1QP+PowerPlay every draw.

I'm looking forward to the test on the PA evening pick3, although the draw count is over 10K, the process for non-replacement picks (like the red ball in PB) is much quicker (no OOPs to sift thru)

now if only I could automate the sifting process...

Next system in full swing

LoopRS=

Latent out-of-position Repeating Sequences (or schedule, or any other S word that makes sense)

What this does is hunts for repeats, even out-of-position ones. For example, 12 is WB1 in one draw then comes back as WB2 in the next draw. My old format excel programs would miss this. When I tire of LoopRS, I shall keep the engine and revisit my old stat systems with this new backtest ability.

right now, I have a test sheet for Powerball, and a new one for PA cash 5 (a game I don't really play, but one with 4000+ draws and no matrix change in decades)

the hunt for the best fit must be done by hand, but the backtest is finally 100% automatic.

PB status - almost back to 2 years ago in WBs and almost 4 years ago in RBs

C5 status - back 11 draws so far, that excel file is almost 30 megs! (SLOW recalcs)

of interest is the subset LRS (Latent Repeating Sequence) for the red ball. Based on my belief that a pick 3 game is 3 sequentially unrelated draws of 1 in 10, I can take this new system to the daunting 10,000+ draw history of the venerable PA pick3 evening...

did I mention that this is my first system capable of adapting to any game?That's definitely a first for me. From now on I am dropping all the old systems I tried that cannot be backtested. I was faced with a challenge in the LoopRS backtest sheet because sometimes the picks would have doubles (or even triples), like 11 twice in one pick. A pretty clever use of a "uniqueness" check and some simple math helped me verify genuine hits only for backtest results.

I am hoping to have the PB history thoroughly checked by may, the C5 could take 'till the end of the summer! anyway, the live fire tests won't happen until I have completed the preparatory detective work currently underway.

as always the system will spit out just one number to play for the planed games....

1 PB ticket, 1 C5 ticket, 1 pick3 ticket (and if I feel particularly ambitious) 1 pick 4 ticket (evening only!, I don't do computer generated games like PA midday 3/4)

to paraphrase "Rage against the machine"

"Rally 'Round the Spreadsheet... Pocket full of Pens"

Still plugging away here...

got an access DB built with all PB sort order data. Now to build the queries...

As for the Excel sheets, big leaps in summary screen automation and updating is getting to be a quicker process... system finally generates the "quick" pick just by entering the results from the last draw.

think I'll keep 'em both!

hot indicators are definitely direction and divisibility, still awaiting inspiration on narrowing them down however.

New deadline... 1/1/2008

I'm giving myself until then to get repeatable results or I'm wiping out the files and giving up. (systems that is, will always have that 1 QP)

Ahh the challenges of designing a database.

After much thinking and a quick review of the "Designing a database" book I used to study for my MCDBA, I have come to the realization that my entire suite of excel workbooks can be reduced to 2 tables and a few carefully crafted queries... Added bonuses,

1) new filters can be added as easily as dropping in a new query.

2) once complete, a few minor changes can be made to convert to any game

3) I can backtest by simply selecting certain date ranges

I have to refresh my memory on the use of Access forms to make it feel more like an app than a db. but at least I have a plan now.

look out MUSL ;-)

The massive collection of worksheets and workbooks now occupying my "PROfile" folder will eventually become an Access Database App. By converting the summary sheets into queries, I will finally get the power to backtest using date ranges (or draw numbers).

Updating the sheets is already a formidable challenge, but I am essentially looking at an exploded database with multiple queries... I didn't know what I wanted from Access before, but it's all becoming clearer now...

There was good in the first test...

Direction... 5 of 6 went in the indicated directions

Divisibility... 5 of 6 went the indicated way (Even or Odd)

HEXade Groupings... 3 of 6 went where the clusters indicated

Next Step...

Time to add follower ranges... This will be one decent eliminator for some numbers and useless for others.

Also time to make a color coded "elimination sheet"

marking downn what reasons numbers were eliminated. I will keep a workbook full of such sheets to go back and review. Some may be eliminated on more than one filter... I should note this

Also time to continue the PROfiling... Including consistent o/e mixes and the long saught after (by me at least) "out of position repeat" counter. Most of this will not be ready for saturday, but I am giving myself no deadlines, only want one win... ;-)

The complexity of this "suite" of spreadsheets is daunting, but the ease of readability has improved 100 fold over prior attempts. Some stats update from multiple workbooks (usually I use multi-sheet formulas).

Using percentages as confidence indicators is great when deciding what to play, but even better at deciding what NOT to play, hopefully accuracy will get better over time.

Finally done sheets for red ball and the 5 white balls...

Now picks can be made from the data by filling in a "quiz" for each number.

Side-by-side, decisions can be made...

I must modify the Failure analysis form to include the white balls.

Here are the questions I fill in for each number...

Last number Drawn
Expected Direction (Higher/Lower/Repeat)
Confidence Index
Hot Follower
Hot Follower Count
Hot Cluster (HEXades)
Cluster Count Composite
Cluster Count LOW
Cluster Count MED
Cluster Count HI
Frequency of drawn number
Follows Raw Odds? (Y/N)
Odd/Even Count
Lower Numbers
Higher Numbers
Observed Higher
Observed Lower
Observed Repeats
Hot HEXade Subdivision ODD
Hot HEXade Subdivision EVEN

Once they are all side by side, I can make a more informed decision. Doesn't always work out (except for direction which is right about 75% of the time), but it IS the basis for all current work. I updated my sheets and got a pick within 1 hour. (a personal record considering how long it took before).

Still can't get down to just one line to play tho, but since I am getting about 10WB/2PB I think I'll get into the "Powerball Challenge" for wed.

Not waiting for Beta, this alpha system's goin for a LIVE fire test next draw!

Red Ball #29 is my guess.

here's how I arrived at it...

the "magic ratio" seems to be 2 to 1

Direction... in the 25 appearances of #15, the next powerball was higher 16 times and lower 8 times. It never repeated. (2 to 1 Higher)

Divisibility... my category name for odd/even. the 24 followers so far produced 16 odd and 8 even, advantage ODD (2 to 1)

the less than 2 to 1 events leading up to my pick are as follows...

HEXade Clustering... the highest number of followers of 15 end up in Hx5 (25 to 30)

based on that, we have a choice of 25,27 and 29

the oddball event is a "palindrome" potential in the follower list.

following the 15 in the last 4 appearances are 29,5,23,5,??

palindromes are all over the PB data, some in even longer chains than 5.

what I ignored...

"Hot Follower" the number 39 followed 15 3 times (12%)

Hot Follower has proven to be an unreliable indicator. but so has the HEXade cluster... but Hx5 has a higher percentage of hits than hot follower.

I bought one ticket for the draw, filled in the QP box AND 29 as PB so I have a "Q5" ticket. In 5 tries, the direction almost always holds out, the divisibility is right behind it. last week I caught 1+1,if I catch 0+1 or better this week, I will expand this to the sort order white balls (then eventually to draw order) for wed. draw.

good luck all :-)

Got 1 1/2 more workbooks in excel to complete the full first draft of PROfile.

This has got to be the most massive excel project I have ever undertaken. There are, for example, one worksheet for each powerball number, plus 3 additional sheets including the finally fully automated summary sheet.

All this one does is present data, it makes no guesses... beauty is that it seems applicable to any game. If PB tests go ok, I might just create one for the PA cash5.

Backtesting so far is non existent, but once the whole process is worked out, then it should be at least do-able. I have learned a great deal about the powerball data just in the creation of the excel files... The logging system for recording what worked or didn't is so far the best byproduct.

Cashing in a 1+1 PB winner today, hopefully much more to come.

As I delve back into Excel system creation, I am trying to get better results than the dozens of previous attempts by keeping track of more details. Thistime I am using dated Post-Draw notes to get in depth as to where things went right and wrong. Mostly it's a way to capture a copy of ideas before I forget them.

Below is my first entry for the PROfile Powerball system I am working on...

PRO-file Predictive Decision System
Failure Analysis for 2/7/2007 (wed.) Powerball

Drawn Red Ball = 10 (21st appearance)
Previous Drawn Ball = 37 (34th appearance)

Product of Decision Process = 29

What worked...

1) Direction (Higher/Lower/Repeat)

The Confidence Index of a follower of 37 being lower than 37
was 83. LOWER was chosen and the drawn ball was indeed lower.

what didn't work...

1.) Divisibility (Odd/Even)

There were 21 followers that were odd
and 13 followers that were even.
This was not an accurate aid in elimination.
Could D.O.W. facor in the distribution of O/E?

2.) Range (HxN L/M/H)

Stong apparent indicators excluded the HEXade containing the
right number (10), It could be possible that the system may
be trying to create balance rather than clear-cut indicators...

3.) Follower Frequency

Just because the 29 followed the 37 more than any other number, did

NOT make it reliable as a predictor.

Follow Up and Corrective Action Plan
A). Take apparent trends with a grain of salt. Perhaps a confidence

index is required for each possible follower number? or for each metric...

B). Incorporate Deltas (Drawn/Follower, Precursor/Drawn,

Precursor/Follower). Maybe this can create data that can be

manipulated with different tools (Trigonometry perhaps?)

C.) Continue to automate the PROfile sheet, This file needs to be

complete BEFORE even attempting to expand to the white balls!

D.) Attempt what works for Saturday's guess. PB 10 only has a Confidence Index of 58 to go higher... (Confidence Index Ranges from 24 to 98)

E.) Perhaps a "Backup" pick of the deviant data path (e.g. goes low

when expected to go high) might be in order later in the live trials.

I have kept track of decades in the PB for some time, the data is just as wild as the draw data.

I decided to group the PB data into full matrices...

the powerball numbers 1-42 fit into a 6x7 grid I call HEXades (Groups of 6)

Hx1 = 01,02,03,04,05,06

...

Hx7 = 37,38,39,40,41,42

The white Balls fit nicely in a 5x11 grid I call PENTades (Groups of 5)

Pn1 = 01,02,03,04,05

...

Pn11 = 51,52,53,54,55

So far this yields no great truths or hidden connections, but I'm glad to be back into working a system out.

Since I'm working right now on just the red balls, the 6x7 grid can be nicely broken down further by Low Medium and High

Hx1 Low = 01,02

Hx1 Medium = 03,04

Hx1 High = 05,06

Each subdivision yields one even and one odd number. Still looking for some way to apply this info...

Before moving in a positive direction with my excel work against the PB, I gotta let out the negativity, so here goes...

First, let's start on the same page, with the simplest of all games (nice round numbers), the straight pick 3. (not worth the time to get into box play)

In pick 3, you are playing 3 sequential games of 1:10, you MUST match all 3 games correctly to have a straight hit.

odds of first correct ball = 1:10

odds of also selecting the second correct ball = 1:100 (or 10 possible first game combos times 10 possible second game combos)

odds of ALSO selecting the third correct ball = 1:1,000 (10x10x10)

the range is 0-9 for each game, 000 to 999 for the entire process. To further prove the math I'm about to lay down, most serious pick 3 players know there are 720 combos that contain NO doubles or triples... the derivation is to imagine 10 balls and they pick 3 of them

first ball 1:10

2nd ball 1:9 (one ball is gone)

3rd ball 1:8 (two balls are gone)

10 x 9 = 90,  90 * 8 = 720

All in agreement so far?

now let's look at what the MUSL advertises as the odds for a powerball jackpot win... 1:146,107,962    ... and it's flat out wrong. False on purpose? who knows.  NOW let's do the actual math...

the game starts with one group of 55 white balls and one group of 42 red balls. the odds of matching the first number...

1:55  simple enough. on to two numbers

55x54= 1:2,970 (55 for the first ball, then 54 remain) 

three numbers (a prize they claim has odds of 1:290) 55x54x53 = 1:157,410

four numbers (they say 1:14,254 ... I say) 55x54x53x52 = 1:8,185,320

five numbers 55x54x53x52x51 = 1:417,451,320 (wait, it gets better...)

odds of winning the powerball jackpot, 55x54x53x52x51x42 =

1:17,532,955,440

you read correctly, that is 17 BILLION to one!!! Makes that 15 million minimum look a bit low, yes?

you may slice/dice/analyze this as you see fit... but it's true.

you start with 55 balls, the first one could be any of the 55... the second could be any of the 54 remaining...

This is what we're all up against.... and I STILL accept the challenge!

the bright side is the 42:1 odds of matching a red ball... that is where I choose to begin anew. Also, all 17 billion are not unique, many sets contain the same numbers. BUT trying to create a system that uses draw order data does require you realize the TRUE odds.