lexicographic clustering

Reply #43

time*treat — Wed, 27 Feb 2008 04:51:10 GMT

45 numbers have plenty of room to spread out in a space of ~288K nums. A space that large with a data set so small won't tell you much unless there is a software glitch that sends everything to one side or the other.

There are probably better filters to work with such small data sets. 90 points in a field of 575,757 is not enough to tell you much, unless you devise a filter that makes them clump near each other (like a number or pair that shows up relatively often).

Reply #42

GASMETERGUY — Wed, 27 Feb 2008 02:13:19 GMT

After checking the last 90 draws for the Tennessee P5 game, there are 45 draws in the Lower Half and 45 draws in the Upper Half.

This means something but I don't have a clue as to what it may signify.

Any thoughts?

Any one

Reply #41

GASMETERGUY — Wed, 27 Feb 2008 01:21:46 GMT

In case there is someone who cares, the mid-point for a Pick 5 out of 39 game is:

5 - 16 - 22 - 26 - 29 (lex # 287,878)

or

5 - 16 - 22 - 26 - 28 (lex # 287,877)

depending on where you wish to draw the line between Upper Half and Lower Half.

Reply #40

GASMETERGUY — Tue, 26 Feb 2008 01:36:01 GMT

I begin with the RNG numbers. I feel there is a vast difference between RNG and ball drawn numbers.

The only proof I have is the fact that with ball drawn numbers, there were many times when a number went more than 40 draws before coming out. With RNG numbers, that is not the case. All the numbers will be drawn within 35 draws. This makes me feel there is a vast difference between the two systems.

Some of my other patterns have also been skewed because of RNG numbers.

Reply #39

time*treat — Mon, 25 Feb 2008 21:52:12 GMT

TN may have changed their method of drawings seven months ago, but their pick-5 game goes all the way back to Sept 1, 2004. Maybe you are using a different source?

I just checked their site. It's all there, over 500 drawings for pick-5.

Reply #38

GASMETERGUY — Mon, 25 Feb 2008 21:02:03 GMT

I agree that 50 is way too small of a sample size. 1000 draws would be much better. Tennessee's P5 game is seven months old and there have been approxiately 84 drawings to date. I have to work with what I have.

And I agree that adding a number or two should not change anyone's method. What has been developed for 35 numbers should ( I repeat, should) work with 39, 40, or 44 numbers. It should also work whether the game is P5 or P6. But I play patterns. I do not use mathemetics except to determ... [ More ]

Reply #37

time*treat — Mon, 25 Feb 2008 16:24:12 GMT

GASMETERGUY,

For a 5/39 game, this is how I divided it (10 groups) of ~57,600. You could use 20 groups of half the size.

As you can see from the earlier replies, we each used a sample size of over 1000 drawings. I think 50 is way too small a sample size for pick-5 type games. As for my own game, even with a sample size over 1000, I didn't find much variation (<5%). RJOh's lottery is also 5/39 like yours, and since he was the only one who understood tested what I was talking about, his re... [ More ]

Reply #36

GASMETERGUY — Mon, 25 Feb 2008 01:23:19 GMT

So what conclusion might one draw from the information I have posted above?

Where would you draw the line assuming there is a line to draw

Reply #35

time*treat — Sun, 24 Feb 2008 20:59:14 GMT

Yep, that's what we were checking, GASMETERGUY. I've noticed that some things are common among pick-5 games even if they have different pool sizes (e.g. 39 vs. 36).

Reply #34

GASMETERGUY — Sun, 24 Feb 2008 18:29:38 GMT

I have taken a look at Tennessee's P5 game and found the lex number for 49 drawings.

The first column gives you the date of the drawing; the second column gives you the lex number for that data; and the third column has the lex numbers organized in ascending order.

I think this is what you are looking for.

02.20.08

568,818

02.18.08 173,955 4,298

02.15.08 546,811 56,370

02.13.08 323,733 64,555

02.11.08 516,370 74,561

02.08.08 472,524 102,448... [ More ]

Reply #33

bobby623 — Sat, 23 Feb 2008 16:44:06 GMT

I wonder what these numbers would look like if you reduced the field using Alphanumeric Substitution.

I know for a fact that the total number of combinations would be drastically reduced.

In AS, the numbers are converted to letters using a fixed code. I.E. All single digits become As,

the teens (10-19) become Bs, the 20 (20-29) become Cs and (for a Pick 5 with less than 40 numbers) the 30s would become Ds.

Once converted, the letter combination AAAAA would represent 'x' number of num... [ More ]

Reply #32

time*treat — Wed, 20 Feb 2008 20:30:05 GMT

We were looking to see if the number combinations were bunching up anywhere. Such a trend might be more noticeable if the combinations were expressed as a decimal representation, rather than a group of 5 numbers. That decimal representation is the 'lex number' and will depend on how many total numbers are in the draw range.

Now for 6/49 there are 13,983,816 combinations, so that's how many lex values there are.

The combination 01,02,03,04,05,06::49 would have a lex value of 1

The combin... [ More ]

Reply #31

crystaltips — Wed, 20 Feb 2008 19:53:45 GMT

I have no idea what you are all talking about! Sounds interesting though.

What about a 'lex' for 6/49?

Thank you

Reply #30

time*treat — Tue, 29 Jan 2008 16:08:23 GMT

... that's if the combos are graphed as draw history. If they are graphed in a natural generated order, and consolidated, the sums create a bell curve.

Reply #29

time*treat — Sun, 20 Jan 2008 11:21:49 GMT

I'm not sure how someone would graph a list of 5-point groups without consolidating them into some single value like a sum or a lex value. A quick glance at my own numbers leads me to believe that a graph of either of those sets will look much like the ones in LottoVantage's blog.

Reply #28

RJOh — Sat, 19 Jan 2008 22:15:24 GMT

possible combos of 5 with 49 numbers are 1,906,884

MATCH ODDS WINNING COMBOS

5/5 1 : 1906884 1

4/5 1 : 8668 220

3/5 1 : 202 9460

2/5 1 : 14 132440

1/5 1 : 3 678755

You might want to change that ~20 millions to ~2 millions

Reply #27

time*treat — Sun, 13 Jan 2008 20:07:37 GMT

XL has 65536 rows, 256 cols, and 32 worksheets, IIRC.

If you were willing to use multiple columns, or worksheets, you could show as many combos as you could stand. It could be done. I've never needed to look at all of the combos at once. For huge blocks of data like that, I would create a file, then use that file like a random access database.

Reply #26

LottoFan — Sun, 13 Jan 2008 15:23:44 GMT

time*treat:

Have you been able to display all of the 5/39 combinations in a way that you can view them... such as a spreadsheet?

hypersoniq helped me with the VB coding a couple of years ago.

I was working on a 5/49 back then, which I remember generated about 20 million combinations. I used excel to display them.

I displayed each combination from lowest to highest number. I then sorted the combinations.

I noticed the ~20 million combinations I was working with were finite. T... [ More ]

Reply #25

time*treat — Sun, 13 Jan 2008 05:21:50 GMT

@LottoFan

Ah, you mean low numbers in the early positions, and high numbers in the later positions.

Yes, of course, that is understood. It is used with the Delta system, so that's a whole other method.

Reply #24

psykomo — Sun, 13 Jan 2008 04:50:26 GMT

YES>>>WE >>>NOTICE.................HOW

the lottery can screw-UP your PICK's with ###########'sssss

the LOTTERY mechanic's>>>>throw-into-GAME!!

??can U-BLOCK (THEM???)

ask the lottoarchitech????????

YOU may have a friend in the WHEELING BUSINESS>>>>>$$$???

TELL HIM WHAT UR PROBLEM'$$$$$$$$$$$$$$$$$>>>>>>ARE??

HE MAY KNOW HOW TO development software to HELP>>>HIM!!

LOL

PSYKOMO... [ More ]

Reply #23

LottoFan — Sun, 13 Jan 2008 04:28:42 GMT

In reference to a 5/39 lotto:

I don't even look at individual numbers. I think looking at lottery combinations (such as 1-2-3-4-5) has almost nothing to do with individual numbers characteristics. (Your mileage may vary.)

In my previous post, I never mean't to make a distinction between individual numbers and their relationship with the overall lottery.

I did try to indicate that there are more combinations leading with the number1 than any other combinations.

I also tried to in... [ More ]

Reply #22

RJOh — Fri, 11 Jan 2008 01:53:33 GMT

I solve my problem by adding 5 lines and changing one in the routine in my earlier post. If time is not wasted writing the index numbers and combinations to a file or the screen, the routine takes about 2.5 seconds to run. I tested it by entering 35 36 37 38 39 and it took 2.5 seconds to match the combination and print out index number 575757.

INPUT #1 ,N1

INPUT #2 ,N2

INPUT #3 ,N3

INPUT #4 ,N4

INPUT #5 ,N5

N=0

FOR A=1 TO 35

FOR B=A+1 TO 36

FOR C=B+1 TO 37

FOR D=C+1

Reply #21

RJOh — Thu, 10 Jan 2008 22:39:44 GMT

Ion Saliu's program works for any lottery once the combination and number pool sizes are entered. That's more along the line of what I'm thinking off, I just have to figure out the number of loops needed to reach a certain combination.

Reply #20

time*treat — Thu, 10 Jan 2008 20:25:55 GMT

I know of two methods, but you can't mix them.

Easy to code, slow to run vs. hard to code, fast to run.

What you have above is the easy to code version (presuming your 'PRINT' goes to a file) for creating the file.

Once the file is created, if you know the A, B, C, D, E and are looking for N, what you would do is something like.

FOR N = 1 to 575,757

IF val(col1) = A then

IF val(col2) = B then

IF val(col3) = C then

IF val(col4) = D t... [ More ]

Reply #19

RJOh — Thu, 10 Jan 2008 18:08:30 GMT

N=0

FOR A=1 TO 35

FOR B=A+1 TO 36

FOR C=B+1 TO 37

FOR D=C+1 TO 38

FOR E=D+1 TO 39

N=N+1:PRINT N; = ;A; - ;B; - ;C; - ;D; - ;E

NEXT E

NEXT D

NEXT C

NEXT B

NEXT A

That's a big loop to run and store all the results. I'm sure there's a better way to find N when you know A,B,C,D and E

Reply #18

RJOh — Thu, 10 Jan 2008 01:32:03 GMT

Ion Saliu had a program on his website that would find the lexicographic number of any combination and the combination of any lexicographic number. Years ago I down loaded it when his website was free but I didn't transfer it when I bought my last computer and now his site requires a paid membership to down load his free programs. So now I giving some thought to writing a routine for my program that will do the same thing.

Reply #17

time*treat — Thu, 10 Jan 2008 00:58:00 GMT

Since these are constants, you just generate each combo, in order, and have a file contain them.

combo 1 (the first record in the file) will be 1,2,3,4,5

combo 575,757 (the last record in the file) will be (35,36,37,38,39)

The combo value (record number) is in effect the lex val

It's a big file, but you only need to make it once. Then you just read from it. Probably find other uses for it, down the line.

I made a routine that would create the lex val from the 5 numbers, and ano... [ More ]

Reply #16

RJOh — Thu, 10 Jan 2008 00:36:15 GMT

I tried something similar on the 10 lines I posted on the prediction board for tonight's Ohio Classic Lotto and the best I could do is match 3. It still needs some work.

Reply #15

RJOh — Thu, 10 Jan 2008 00:26:43 GMT

Rather than looking for clustering of combinations with all the winning numbers, I'm look for clustering that won the most prizes or the most money. When looking at combinations with all the winning numbers the distributions of combinations are spread out fairly even.

This should be fairly easy to check once I figure out a routine to assign the lexicographic numbers to each combination as I check for matches with other combinations in the data file. It too big of a task to do by hand.... [ More ]

Reply #14

time*treat — Thu, 10 Jan 2008 00:08:59 GMT

In the total combination set, no number appears more or less often than any other.

In a 39choose5 ball game, each number appears in exactly 73,815 combinations.

Stated another way: Pick any number X.

Out of 575,757 possible combinations: 73,815 combos will contain number X. 501,942 combos will not contain number X.

Reply #13

LottoFan — Wed, 09 Jan 2008 18:19:35 GMT

I look at total combinations, which can be expressed as a part of lex.

For the 5/39-

The largest amount of combinations contain the number 1.

The second largest amount of cominations contain the number 2, and so on, leading to:

The least amount of combinations contain the number 39.

The combinations with low leading numbers appear to have odds favoring the lottery (such as 1-12-22-31-38).

The combinations with high leading numbers appear to favor the player (such as 35-36-37-38... [ More ]

Reply #12

time*treat — Mon, 07 Jan 2008 18:11:23 GMT

(07,08,09,10,11:35) = lex val 205,878

(08,32,33,34,35:35) = lex val 243,902

That block contains 38,025 combos, ~11.7% of the total 324,632.

Reply #11

time*treat — Mon, 07 Jan 2008 18:05:05 GMT

Connecticut (CT) Cash 5

Game Type: 5/35 (324,632 combos)

group lex start lex end comb start comb end members

1 1 32,463 01 02 03 04 05 01 10 15 19 32 32463

2 32,464 64,926 01 10 15 19 33 02 07 12 19 31 32463

3 64,927 97,389 02 07 12 19 32 03 06 11 21 34 32463

4 97,390 129,852 03 06 11 21 35 04 06 16 18 22 32463

5 129,853 162,315 04 06 16 18 23 0... [ More ]

Reply #10

THEGUY — Mon, 07 Jan 2008 17:31:51 GMT

Interesting, so my answer should be yes, that in connecticut the lex value that stands for the group which may contain 7 or 8, as its lowest combination value would be the most repetitive.

Reply #9

benmas — Mon, 07 Jan 2008 01:31:25 GMT

the drawings of the games are random ..yea they do cluster sometime somewhere but so what ...and many times people tend to notice patterns that they want to see... dont want to knock your method or your way of thinking but it will only work once in weeks apart...there are no fancy ways about lotto lex number, magic number all that crap ..i think one should keep it simple whatever method it is ...I pick my numbers based on observations and intuitions... many times i'm wrong (as i should be ba... [ More ]

Reply #8

RJOh — Mon, 07 Jan 2008 01:01:32 GMT

I recently combined my old BigGame and MegaMillions files and the only value I could find in the combined file was no combinations of five had ever repeated so I now have 1109 combinations of five to avoid.

The same thing probably would be true if I added all the PowerBall past results to the same file.

Reply #7

time*treat — Mon, 07 Jan 2008 00:27:50 GMT

It would take a few hundred years for a good chunk of the combos to fall. I'm in a bit more hurry than that.

I just checked 1,096 combos on our 5/36. The difference between the largest group and the smallest group was 48 combos. A not-so-whopping 4.38% difference, but it tells me that these p-5 games aren't so different even if the matrix size changes. Looks like 'old' data and theories need not be tossed after a matrix change.

Folks may want to dig up their pre-matrix change stuff for P... [ More ]

Reply #6

RJOh — Sun, 06 Jan 2008 23:48:05 GMT

I just considered it for about 10 minutes after reading your post since it wasn't that hard to check. Depends on what you consider fairly even, high and low varied by about 40 combinations. Since less than 0.2% of the possible combinations were included, it hard to draw any conclusions. Using an old Buckeye5(5/37) with almost 5,000 drawings might yield different conclusions.

Reply #5

time*treat — Sun, 06 Jan 2008 23:23:24 GMT

Heh, yeah, I know it can be done, but I just gave it any serious consideration yesterday.

I figured I'd see if anyone else had tried it, and found anything.

From your 1,050 results, was it fairly evenly spaced

Reply #4

RJOh — Sun, 06 Jan 2008 23:05:28 GMT

Since you know the the start and end combo of each group, all you have to do is sort the combinations in the lottery file from lowest to highest and count the combos in each group. If you don't have a sort routine in your lottery program and your lottery files have random access records then you can use the sort command in the command window.

I sorted the 1050 combos in Ohio Rolling Cash5 and the results were:

#1-#97 combos were between 01 02 03 04 05 and 01 13 19 22 25 (the lowest was 01... [ More ]

Reply #3

sheba88 — Sun, 06 Jan 2008 22:02:51 GMT

Is there anyone out there who has a system as to what number relates to the other? for example does the following numbers

1= either 9-3-7-5

2 = 4, 6, 8 or 0

3= 1-7-5-9

In other words 908 do we have 2 evens and 1 odd?

I play the california midday and daily three

Reply #2

time*treat — Sun, 06 Jan 2008 21:07:02 GMT

What I'm referring to is the clustering of the lex values, not groups of 2 or 3 numbers.

Example: In a 5/39 game, there are 575,757 total combos. These can be represented as combo 1 through combo 575,757. These can be divided into 9 groups of 57,575 combos and 1 group of 57,582 combos.

grouplex vals start comboend combo

1 1 ~ 57,575

01,02,03,04,05:3901,13,19,22,25:39

2 57576 ~ 115150

01,13,19,22,26:3902,10,21,22,32:39

... ... ......

10 518176 ~ 575,757

14,20,23,33,39... [ More ]

Reply #1

THEGUY — Sun, 06 Jan 2008 19:24:09 GMT

I'm in conneticut, where it's 1-35, I'm noticing that the cluster is odd and that the difference between the numbers is about 2, 6, or 8. Two numbers with a difference of four not a likely happening. I'm on the 7 combo matrix now to see how perhaps on a standard basis a lower prize, could be gained consistently, which will involve clustering to some degree. Right now cluster gambling looks like this

7,8,/12-15/22,23,26/32,33

The same perhaps type of cluster only appears with the twenties... [ More ]

lexicographic clustering

time*treat — Sun, 06 Jan 2008 07:08:23 GMT

lex num combination

101,02,03,04,05:36

376,99232,33,34,35,36:36

-- --

101,02,03,04,05:39

376,99208,09,18,24,25:39

575,75735,36,37,38,39:39

Most of us would call the 'lex' the combination number .

I'm wondering if anyone has noticed their p5 game tending to cluster somewhat anywhere? Or are the combos spread out evenly? Not looking for anything spectacular, maybe 15~20% of the combos falling in a 10% range, or 40% falling in a certain third of the range. May... [ More ]