Random Number Generators (revisited)

[winhunter]

Colin,

You referenced the Sums post on L-Post...
/threads82447p1.htm
(complete the link as you see fit)

BTW, a good coder doesn't mind sharing his code to be peer reviewed. But since you didnt post your code, no one can prove your claims. How convenient for you to make an unsubstantiated claim of:

A> Reducing my code to 15 functional lines
B> Producing a true RNG from using a PRNG and a Timer via several calls to the Randomize Function

Since you brought it up (and I never bothered to work with the code after that, what I had served it's purpose...)
The code can be optimized to 9 lines...

Dim intChar%, strFinal$
Do
Randomize (Timer)
intChar = Int((122 - 48 + 1) * Rnd + 48)
If (intChar < 58 Or intChar > 64) And (intChar < 91 Or intChar > 96) Then
strFinal$ = strFinal$ & Chr$(intChar)
End If
Loop Until Len(strFinal$) >= 6
strFinal$ = Chr$(Int((57 - 48 + 1) * Rnd + 48)) & strFinal$

Now whether or not it generates truly random sequences, has yet to be seen or tested. So whether or not 9 lines suits the need, I dont know. I dont have your system for testing randomness, nor do I have your 15 lines of code to test against.

Come on Colin. People here like to see tangible results. I did exactly that with WINHunter. I provided sourcecode and all to be peer reviewed, user tested, etc. Check the Sourceforge statistics yourself, the app is still being downloaded on a regular basis. You failed to see the ultimate goal of WINHunter and the possibilities it opened up to users.

A change needs to occur in the Lotto software world, and sitting on your lazy bum poiting out flaws at all the other coders out there taking risks is not getting squat done to support the effort. Your right, there is tons of computing power out there, and software should be able to churn out numbers at blazing speeds, and back test that information against previous draws. The WINHunter concept, was to begin to explore the concept of software optimization where the filter design would generate the least amount of numbers and match the most amount of numbers possible over the entire draw history.

Oh, and I will forgive you in advance for any attempts at character assasination in your reply (although, I don't see their purpose here, as most of us here are searching for the same goal cooperatively).


Thanks,

Andrew

 

[CMF]

Sensitive lad ...

Andy

I come here to relax.

Think what you like.

I'm doing my thing, you're doing ...

I'm glad you worked out to use the ASC11 numbers. The nights I spent tossing and turning - why can't he see it?

I shall continue my mission - just like I said on that other forum - like the Blues Bros, I am on a mission from God. Now, if you have a problem with that - talk to my Boss.

Bye
Andy

Oh, and regarding using History to determine numbers to play - I do have a quite sophisticated program for extracting the best possible outcome. Is it any better than Random Selections? No, the results fall within the ambit achieved by Random Selections. This of course is to be expected as there is no relationship between any history and any future draw, anywhere. But if you can be on par with Random Selections, that's an achievement - although my 9 year old daughter has achieved better than that with her design.

When eventually you settle on a preferred method for using your software, instead of offloading that decision to users - well, I'm quite prepared to test it.

 

[tomtom]

Re: Sensitive lad ...

But if you can be on par with Random Selections, that's an achievement - although my 9 year old daughter has achieved better than that with her design.

Hey Colin,

You are absolutely wrong…Your quicke pickie – "stick to the random" lottery theory is wrong, and your opinion about winhunter is wrong . Andrew made a first class lottery program. That’s the simple true fact

 

[Brad]

bit of history review is in order here ...

Tom, don't sweat it cuz it's probably not worth the aggravation. You may want to know that Colin already got tossed from another lotto forum for pushing ... ehm, excuse me ... presenting his brand of lotto wisdom.

Then he set up his own site, which looked very vacant last time I glanced possibly because no one is buying into his ideas and theories

What he's looking for here I don't really know. He says relaxation. I already speculated some time ago that he may be after converts to bring to his lonely site, look at his signature and see if I'm close.
Not sure why LT has not warned him about it so allow me to speculate again: he possibly finds it nonthreatening for the low occupancy reason I stated above, hence he chose to overlook it

So other than having a questionable sig Colin has done nothing wrong here, he's entitled to his opinion like anyone else. I don't really follow his reasoning much anymore because as Andrew said he has not offered anything tangible that I could see so far.


Cheers


PS Colin, them are just some facts and observations old boy, I ain't looking for a duel with pistols at dawn.

 

[tomtom]

Re: bit of history review is in order here ...

Then he set up his own site, which looked very vacant last time I glanced possibly because no one is buying into his ideas and theories

His lottery ideas and theories ...which ones ?

Having a quite rare and rather outstanding opportunity of playing a 6/40, or even less than 40 lotteries at quite fair ticket prices , at the same time undervaluing anything else than random doesn’t look like an idea or theory…it should be rather identified a mistake….

 

[winhunter]

Randomness

In some cases during our trials of WINHunter, some noticed a slight edge over random odds, while I myself experienced a decrease in edge over odds.

But, there was a primary difference in what I was targeting with WINHunter, and what some of the other users were targeting.

WINHunter can be configured in such a way that it targets mainly 6of6, or jackpot wins. In one particular configuration, it predicted 6of6 from 11 numbers out of 53 with a single design. While in some cases it's predictions are way off, it does tend to swing back in line with it's predictions.

The discussion regarding randomness, had everything to do with machines and what the output was from the machine. That discussion has two different belief sets, either the machine produces totally random output, or it produces pseudo-random output. In otherwords, given the fixed manner in which balls are drawn from, will certain sequences ever occur, or will other sequences occur more frequently. In the case of the Florida lottery, I want to say that no 6of6 pattern has ever hit again, but already some 5of6 patters have. With such a small history, and so many possibilities, why already then do we have matching sequences from a totally "random" system?

Although I have read somewhere that an actual repeatability test was performed scientifically via computer control, I don't seem to be able to find a link to any paper written about it.

Anyway...

I still believe there are alot of aspects of Lottery analysis that have not been explored. WINHunter is just another exploration tool.

 

[winhunter]

Stay on topic M8

When eventually you settle on a preferred method for using your software, instead of offloading that decision to users - well, I'm quite prepared to test it.

There is NO prefered method, as the methods you are able to produce in some cases mimic the same methods of other software available.

As has been discussed here before, WINHunter gives the user the ability to test the method against draw history.

To prove your point wrong regarding randomness, a simple method can be devised within WINHunter, and will yield different results when tested against different lottery histories (as you put it, they are all random). When the draw history (again, random) is used to predict results, some lotteries fair better than others. If they were truly "random", then wouldn't expectation dictate that since you are using the history (Random) that you would still end with the same results?

Why then do different methods yield different results, when fed with the Totally Random history of that lottery?

Why do some WINHunter filter designs perform better when specific amounts of history are bypassed, and yet other filter designs perform better when based immediately upon previous draws?

Any system that generates numbers based upon a totally random system will generate no winning numbers without luck, What is worse, is there is no way to re-test any given set of results. At least with WINHunter you are not only able to back-test your design, but you are also able to Optimize a design via draw history.

Oh, and regarding using History to determine numbers to play - I do have a quite sophisticated program for extracting the best possible outcome. Is it any better than Random Selections? No, the results fall within the ambit achieved by Random Selections. This of course is to be expected as there is no relationship between any history and any future draw, anywhere.

Oh, so the machine itself has no bearing on possible outcomes? So a simple equation (fractal set) has no noticably predictable outcome either?

Also, in previous discussions regarding WINHunter. If you are able to reliably predict WORSE than random expectation, then you simply invert your selection set (use the numbers not predicted), and you might have better than random expectation in some cases. This is a simple selectable option within WINHunter, not sure about other software.

You are absolutely wrong…Your quicke pickie – "stick to the random" lottery theory is wrong, and your opinion about winhunter is wrong . Andrew made a first class lottery program. That’s the simple true fact

In my decision to make WINHunter freeware, I have learned that most users are not interested in investing time into learning how to use free software. Only a select few users here took the time to learn it, and give positive feedback which resulted in the Initial Release of WINHunter.

What makes WINHunter different than most other Open Source projects, is that almost no one wants to share in their discovery of a truly winning method. So why bother helping to improve a tool that may as well enable any user the ability to Win Big? Some people just aren't willing to share what they know. But I guess your contributions to the effort here and elsewhere speak for themselves...

 

[Rob50]

I don't use WINHunter, neither other "predictive - prophetic" softwares, as I am of the opinion that RANDOM = UNPREDICTABLE. I am assuming that the lottery draws are truly random. This is very close to 100% if the draws are performed with a mechanical system and all the measures to avoid any non random influence are properly taken. Not quite true if the draws are carried out through a computer algorithm. In this case the numbers are called "pseudorandom", meaning they are not random because they are produced with a deterministic algorithm, but they behave like random numbers. Actually, if one does not know the algorithm he can't predict the coming number no matter how many previous numbers (history) he knows. There is no randomness in the sequence if you know the algorithm. (As an example, you can perfectly predict the sequence of random numbers produced by the random generator used by the VBA of Windows, I don't want to be too long in this post. Whoever is curious I can provide a few code lines to prove this). Just a question for whoever wants to think about it and give an answer:

Based on what winhunter has posted here: Does the quickpick system used to produce random numbers, therefore lotto tickets, is the best lotto predictor system, since very often it hits the jackpot?

 

[tomtom]

Well Rob,

If you play 14.000.000 QPs you may hit a JP ....or may not ...

 

[CMF]

Things still going on here ... I got bored when the mud slinging started ...

Andrew

Here's an interesting scenario to consider ...

If I play the immediate past draws for a 649 Lotto Game against the next draw for 8 Combs, 26 Combs and 49 Combs respectively, what results would you expect?

Given that the next draw knows nothing about the previous draws then any immediate past draw set is equivalent to a good set of random numbers and should produce results within the range for Random Selections.

What do you think?

Colin

 

[bloubul]

Hi All.

Would you please stop this "WAR" OF WORDS.

DOES ANY OF YOU WAR LORDS HAVE A REAL WINNING METHOD FOR 4PLUS NUMBERS, IF YOU DO PLEASE POST IT THAN.


Bloubul

 

[Bertil]

Random draws

Well Rob,

If you play 14.000.000 QPs you may hit a JP ....or may not ...

It you buy 14 million tickets by quick pick you are likely to
have 1/3 repeats, or exactly36.8%, because the picks
are made by replacement.

Bertil

 

[winhunter]

Re: Things still going on here ... I got bored when the mud slinging started ...

If I play the immediate past draws for a 649 Lotto Game against the next draw for 8 Combs, 26 Combs and 49 Combs respectively, what results would you expect?

I have stated my opinion here and elsewhere before on this same issue. It depends on the "signature" of the draw machine, which is based upon past draw performance.

Depending on the history you use, and how much draw history mat alter your stack design inside WINHunter.

What is interesting, is you state that a lottery is totally random. Then why is uding the history a "bad" idea? Isn't the history an excellent source of random data? Case in point, is the Redbus filter. It is based upon the notion of number position, and locating a number relative to a found number in that position and moving via row/column motions (with wrap around). With enough history, this method yields interesting results.

The primary difference, is that my random number generator is always the same, thus I can adjust how I use it and benefit from it which then makes my prediction results repeatable over the entire history range. I can then backtest my filtering methods against the entire history and gauge my results.

Given that the next draw knows nothing about the previous draws then any immediate past draw set is equivalent to a good set of random numbers and should produce results within the range for Random Selections.

For starters, your not "given" anything. The foundation for your statement rests on the belief that the draw is totally random, yet the machine is loaded the same every time, balls are drawn using the same method, mixed using the same method and drawn in the same relatively short time frame.

There is still too little historical data for any lottery to make the claim that it is totally random. Unless you believe .01% (or less in most cases) of data is enough to make that judgement (1600 draws out of 16 million combinations, roughly 3 draws a week, 52 weeks a year for 10 years).

Since WINHunter's primary design was to target a win, how often would you expect any lotter method to make a jackpot winning match?

For the Florida Lotto, 2 stacks generated the following results over the entire history... (586 draws)

FLLottoTest6.xml
3: 20, 5 of 6
180: 21, 5 of 6
425: 18, 5 of 6
427: 17, 5 of 6
482: 18, 5 of 6
497: 20, 5 of 6
499: 19, 5 of 6
577: 19, 6 of 6 ***
582: 19, 5 of 6
585: 19, 5 of 6


FLLottoSept4A.xml
153: 25, 6 of 6 ***
164: 23, 5 of 6
170: 33, 6 of 6 ***
199: 17, 5 of 6
290: 11, 6 of 6 ***
323: 21, 5 of 6
391: 21, 5 of 6
446: 23, 5 of 6
467: 25, 6 of 6 ***
484: 25, 5 of 6
551: 16, 5 of 6

FLLottoTest2.xml
16: 12, 4 of 6
131: 15, 4 of 6
199: 14, 4 of 6
290: 13, 4 of 6
311: 11, 4 of 6
335: 11, 4 of 6
348: 13, 4 of 6
364: 13, 4 of 6
413: 12, 4 of 6
448: 17, 4 of 6
501: 16, 4 of 6
582: 13, 4 of 6
584: 13, 4 of 6
585: 13, 4 of 6

FLLottoTest6 & FLLottoTest2 are both based upon a simple stack with a single Filter group with a single Perkisize Processor. The FLLottoSept4A stack is a 3 group-tier stack.


Andrew

 

[winhunter]

FLLottoTest2.xml

<?xml version="1.0"?>
<STACK>
<INITIAL_HISTORY_FILE></INITIAL_HISTORY_FILE><GROUPS><GROUP usegroup="-1"><GROUP_NAME></GROUP_NAME><FILTERS><FILTER usefilter="-1"><FILTER_NAME></FILTER_NAME><PROPERTIES><PROPERTY keyname="skip">4</PROPERTY><PROPERTY keyname="use">16</PROPERTY><PROPERTY keyname="oncycle">0</PROPERTY><PROPERTY keyname="offcycle">0</PROPERTY><PROPERTY keyname="startcycle">0</PROPERTY></PROPERTIES><PROCESSORS><PROCESSOR useprocessor="-1" keyname="perkisize"><PROCESSOR_NAME>Perkisize</PROCESSOR_NAME><PROPERTIES><PROPERTY keyname="groupa">3</PROPERTY><PROPERTY keyname="groupb">3</PROPERTY><PROPERTY keyname="groupc">4</PROPERTY><PROPERTY keyname="groupd">3</PROPERTY><PROPERTY keyname="groupe">3</PROPERTY><PROPERTY keyname="perkisizeweight">1</PROPERTY></PROPERTIES></PROCESSOR></PROCESSORS></FILTER></FILTERS><SELECTOR keyname="highavglow"><SELECTOR_NAME>High/Avg/Low (Inversions)</SELECTOR_NAME><PROPERTIES><PROPERTY keyname="selector">1</PROPERTY><PROPERTY keyname="type">0</PROPERTY><PROPERTY keyname="selectmin">10</PROPERTY><PROPERTY keyname="keeps"></PROPERTY><PROPERTY keyname="deletes"></PROPERTY></PROPERTIES></SELECTOR></GROUP></GROUPS></STACK>

 

[winhunter]

Cut/paste

Just cut/paste the stack code into notepad, then save as the filename you choose.



Andrew

 

[winhunter]

FLLottoTest6.xml

<?xml version="1.0"?>
<STACK>
<INITIAL_HISTORY_FILE>C:\WINHunter\history\FLLotto.txt</INITIAL_HISTORY_FILE><GROUPS><GROUP usegroup="-1"><GROUP_NAME></GROUP_NAME><FILTERS><FILTER usefilter="0"><FILTER_NAME></FILTER_NAME><PROPERTIES><PROPERTY keyname="skip">14</PROPERTY><PROPERTY keyname="use">1</PROPERTY><PROPERTY keyname="oncycle">1</PROPERTY><PROPERTY keyname="offcycle">3</PROPERTY><PROPERTY keyname="startcycle">0</PROPERTY><PROPERTY keyname="startoffset">1</PROPERTY></PROPERTIES><PROCESSORS><PROCESSOR useprocessor="-1" keyname="compoundhit"><PROCESSOR_NAME>CompoundHit</PROCESSOR_NAME><PROPERTIES><PROPERTY keyname="overdueweight">0</PROPERTY><PROPERTY keyname="notrecentweight">1</PROPERTY><PROPERTY keyname="hitspan">0</PROPERTY><PROPERTY keyname="repeatabilityweight">0</PROPERTY><PROPERTY keyname="undertotalrepeats">1</PROPERTY><PROPERTY keyname="underrepeattest">0</PROPERTY><PROPERTY keyname="undermaxrepeatsweight">0</PROPERTY></PROPERTIES></PROCESSOR></PROCESSORS></FILTER><FILTER usefilter="0"><FILTER_NAME></FILTER_NAME><PROPERTIES><PROPERTY keyname="skip">22</PROPERTY><PROPERTY keyname="use">5</PROPERTY><PROPERTY keyname="oncycle">1</PROPERTY><PROPERTY keyname="offcycle">3</PROPERTY><PROPERTY keyname="startcycle">0</PROPERTY><PROPERTY keyname="startoffset">0</PROPERTY></PROPERTIES><PROCESSORS><PROCESSOR useprocessor="-1" keyname="compoundhit"><PROCESSOR_NAME>CompoundHit</PROCESSOR_NAME><PROPERTIES><PROPERTY keyname="overdueweight">0</PROPERTY><PROPERTY keyname="notrecentweight">1</PROPERTY><PROPERTY keyname="hitspan">0</PROPERTY><PROPERTY keyname="repeatabilityweight">0</PROPERTY><PROPERTY keyname="undertotalrepeats">1</PROPERTY><PROPERTY keyname="underrepeattest">0</PROPERTY><PROPERTY keyname="undermaxrepeatsweight">0</PROPERTY></PROPERTIES></PROCESSOR></PROCESSORS></FILTER><FILTER usefilter="-1"><FILTER_NAME></FILTER_NAME><PROPERTIES><PROPERTY keyname="skip">0</PROPERTY><PROPERTY keyname="use">30</PROPERTY><PROPERTY keyname="oncycle">0</PROPERTY><PROPERTY keyname="offcycle">0</PROPERTY><PROPERTY keyname="startcycle">0</PROPERTY><PROPERTY keyname="startoffset">2</PROPERTY></PROPERTIES><PROCESSORS><PROCESSOR useprocessor="-1" keyname="perkisize"><PROCESSOR_NAME>Perkisize</PROCESSOR_NAME><PROPERTIES><PROPERTY keyname="groupa">1</PROPERTY><PROPERTY keyname="groupb">1</PROPERTY><PROPERTY keyname="groupc">2</PROPERTY><PROPERTY keyname="groupd">3</PROPERTY><PROPERTY keyname="groupe">4</PROPERTY><PROPERTY keyname="perkisizeweight">1</PROPERTY></PROPERTIES></PROCESSOR></PROCESSORS></FILTER></FILTERS><SELECTOR keyname="highavglow"><SELECTOR_NAME>High/Avg/Low (Inversions)</SELECTOR_NAME><PROPERTIES><PROPERTY keyname="selector">1</PROPERTY><PROPERTY keyname="type">0</PROPERTY><PROPERTY keyname="selectmin">10</PROPERTY><PROPERTY keyname="keeps"></PROPERTY><PROPERTY keyname="deletes"></PROPERTY></PROPERTIES></SELECTOR></GROUP></GROUPS></STACK>

 

[winhunter]

Leave them off

The CompoundHit processors in the 2nd and 3rd filters are not used in the FLLottoTest6.xml file, and were not enabled for the test above.


Andrew

 

[CMF]

Good health to all ...

Andrew

Get well soon.

Colin

 

[winhunter]

Blind Faith

Based on what winhunter has posted here: Does the quickpick system used to produce random numbers, therefore lotto tickets, is the best lotto predictor system, since very often it hits the jackpot?

Quickpicks seem to win more often simply because more quickpicks are purchased than manually picked numbers. In the case of major jackpots, there are enough tickets sold to basically guarentee a hit by a quickpick.

BTW, Id rather take my chances with WINHunter, than to blindly purchase from a machine that might produce the following sequence:

1-2-3-4-5-6


At least with WINHunter, you can develop a "fomula" and see how well it performed against the history. What is interesting to note, is that the history is the "signature" of that game itself. Soem formulas (stacks) in WINhunter produce no results at all, where other stacks produce results.

Any questions?




Andrew

 

[Rob50]

I don't have anything against the sequence 1-2-3-4-5-6. If Winhunter thinks that this is a bad or wrong sequence he is free to not play it. Can Winhunter explain with sound mathematical arguments why the sequence 1-2-3-4-5-6 is not as good as any other combination 6/49? The probability to hit the jackpot for whatever combination is about 1/14000000.

I wonder when this simple fact will be understood?

This is the question for now winhunter.

Other question, even simpler than that might come in the future.

You can hunt and trap as much as you want, your odds will never be neither better, nor worse than with QPs! Do you know why?
Just because this is a mathematical truth. As simple as that.