High-Probability Computer Program for Florida Lottery “Cash 3” Game

This thread is an experiment based on high-probability software that I write. I’m currently a hedge fund manager and use an extremely accurate AI program for my subscribers.

That aside; Just for fun, last night I wrote a much simpler computer program that figures out the highest probability of lottery numbers. This isn’t an original idea, but there IS something I do differently than what most other people would think of. I only use the lottery results since the last time the plastic ball sets were changed.

In Florida, and for the Cash 3 game that this thread concerns, the ball sets were last changed in April of 2009. This not only gives us a large past data set to calculate, but it also gives us at least 3 solid months of time to play until the ball sets will be changed again in April of 2011. After that time, it would only be logical to wait several months to play Florida’s Cash 3 game again with a new high-probability number, based on at least a few months of that new post-April 2011 data.

But between now and April 2011, I’ll be constantly running the program with every day’s drawing results added... just in case the highest-probability number changes. As of right now it’s...

033 - Played Boxed and Straight Every Day

Over the last 19 months, the number has won at an average of once every month and has more than doubled the hypothetical amount of money played on it because it comes out straight a lot. ($2.00 a day, $1200 spent and about $2500 won)

I’ll personally be playing the number for 25 days with a $50 investment, and we’ll see if it wins frequently enough to build up profits by using winnings to increase the amount being bet each day. Remember, Florida has TWO Cash 3 drawings per day. So the figures above represent .50/.50 Straight/Boxed for the morning drawing, and .50/.50 Straight/Boxed for the evening drawing... 25 days @ $2.00 per day.

So I’ll update this thread with the winning/losing results, and also with the latest high-probability number if it changes. Results can be found here. This thread will be starting with the Dec. 21st Evening Drawing...

fllottery.com...

Edward Slayton

P.S. Here’s the code, minus the one thousand past drawing results. Sorry about the smileys!

---------------------------------------------------------------


OpenConsole()
EnableGraphicalConsole(1)

cash=0:highest=0ldhighest=0:high$="xxx"

For d=0 To 9
For e=0 To 9
For f=0 To 9
cash=0

a=3:b=0:c=4:Gosub lota:a=3:b=1:c=1:Gosub lota:a=6:b=0:c=6:Gosub lota:a=1:b=9:c=6:Gosub lota
a=6:b=4:c=4:Gosub lota:a=4:b=2:c=7:Gosub lota:a=5:b=0:c=4:Gosub lota:a=0:b=0:c=2:Gosub lota

ONE THOUSAND PAST DRAWING RESULTS WOULD BE HERE, AND LOOK LIKE THE FEW LINES ABOVE AND BELOW

a=5:b=4:c=9:Gosub lota:a=7:b=0:c=0:Gosub lota:a=4:b=1:c=2:Gosub lota:a=1:b=3:c=3:Gosub lota
a=4:b=2:c=7:Gosub lota:a=5:b=5:c=8:Gosub lota:a=4:b=7:c=4:Gosub lota:a=2:b=2:c=1:Gosub lota
a=8:b=8:c=0:Gosub lota:a=6:b=6:c=1:Gosub lota:a=8:b=7:c=1:Gosub lota:a=6:b=7:c=0:Gosub lota


If cash>1 And cash>highest: highest=cash: high$=Str(d)+Str(e)+Str(f)rint("Highest : ")rint(high$)rint(" - $")rint(Str(cash))rint(" # of wins:")rintN(Str(ww))
EndIf

Next f
Next e
Next d

PrintN(" ")rintN("Done")

Lota: ; Boxed

won=0

If a=b And bc: wintype=2 ; 662
ElseIf b=c And ab:wintype=2 ; 266
ElseIf a=c And ab:wintype=2 ; 626
ElseIf ab And bc And ac:wintype=1
EndIf

If d=a And e=b And f=c : won=1:cash=cash+250:ww=ww+1
ElseIf d=a And e=c And f=b :won=1
ElseIf d=b And e=a And f=c :won=1
ElseIf d=b And e=c And f=a :won=1
ElseIf d=c And e=a And f=b :won=1
ElseIf d=c And e=b And f=a :won=1
EndIf

If won=1 And wintype=1:ww=ww+1:cash=cash+40
ElseIf won=1 And wintype=2:ww=ww+1:cash=cash+80
EndIf

Gambler's Fallacy
2nd line

For anyone curious, I can obviously do other things with the Cash 3. I simply chose .50/.50 Straight/Boxed for this thread because it has the highest probability of winning the most money AND coming up more frequently. (Which is important for the $50 lasting until the number hits so the money can grow)

Here's the numbers that come up in position one, two and three most often...

[atsimg]http://files.abovetopsecret.com/images/member/ab920d6b66ba.jpg[/atsimg]

Position one is: 4
Position two is: 8
Psotion three is: 9

This 489 number should be played STRAIGHT because Florida Cash 3 uses three ball boxes with plastic balls 0-9 in each.

The third thing that can be done is to find the number that comes up the most often straight. It's 033, which is a coincidence and isn't always the case wrt the experiement in the first post in this thread.

Edward Slayton

Extremely interesting OP!

and a question for both of you.

You said that the set of balls had to be changed. What, if anything, effect could there be from minute differences in the qualities of each ball; such as weight and distribution.

At some level, they must vary to some degree, and one could assume that this variation could also represent itself in repeated trials as a biased outcome.

As a student in computers i find this very interesting. I created a thread that talks about this type of stuff.
www.abovetopsecret.com...

OP, are you involved with creating evolutionary algorithms and neural nets then?

Extremely interesting OP!

and a question for both of you.

You said that the set of balls had to be changed. What, if anything, effect could there be from minute differences in the qualities of each ball; such as weight and distribution.

At some level, they must vary to some degree, and one could assume that this variation could also represent itself in repeated trials as a biased outcome.

As a student in computers i find this very interesting. I created a thread that talks about this type of stuff.
www.abovetopsecret.com...

OP, are you involved with creating evolutionary algorithms and neural nets then?

Originally posted by john_bmth
Gambler's Fallacy
2nd line

I'm aware of that. However, there are two important reasons why this has a better than 51% chance of winning more than is spent. Otherwise I wouldn't waste my time...

One; The plastic balls are the same for a long period of time with this lottery experiment. (i.e. Casinos change card deck)

Two; Gamblers can't POSSIBLY count or calculate something in a casino that would be comparable to this lottery experiment. In other words, a gambler would have to keep track of thousands of roulette spins on the SAME wheel with the SAME ball. Not only would he be thrown out after standing in the same spot for three days, but I'm sure they relocate the wheels and switch the metal balls around.

The AI I use for my hedge fund is similar. It's actually measuring trader behavior and not something truly random. Truly random (or close to it) would be what happens when the card deck is constantly changed at a Black Jack table.

Edward Slayton

reply to post by EdwardSlayton


i looked at the same numbers, boxed or straight...for the SC drawings:

Digit1 Digit2 Digit3 Draw Date
3 0 3 Evening 10/26/10
3 3 0 Midday 07/31/10
3 0 3 Evening 06/04/10
3 3 0 Midday 02/06/10
3 0 3 Evening 07/25/09
0 3 3 Midday 01/20/09
3 3 0 Midday 07/01/08
3 0 3 Evening 01/13/08
3 3 0 Evening 01/03/08
0 3 3 Evening 04/28/07
3 0 3 Evening 04/03/07
3 3 0 Evening 07/04/06
0 3 3 Evening 03/15/06
0 3 3 Midday 12/12/05
3 3 0 Evening 12/05/05
0 3 3 Midday 06/13/05
3 3 0 Evening 05/09/04


result -> the payouts aren't worth the investment -> the norm is a hit 3X per year, but in 2010 the combo hit 4X already



next... you have to really trust the source as reliable... that the numberd balls are changed once a year,
then you have to believe the #'ed balls are never touched or cleaned by human hands


i dismiss your theory, its certain that the balls are cleaned of the residual human hand oils, from being touched at each drawing (off the air) to replace into the mixing bin.


your thread sounds impressive... but thats the difference between theory & practice
?(& how hedge fund promises/rationale make the manager the most money)

Then you might be talking about,
Bayesian inference or maybe
Frequentist inference
?

Originally posted by VonDoomen
Extremely interesting OP!

and a question for both of you.

You said that the set of balls had to be changed. What, if anything, effect could there be from minute differences in the qualities of each ball; such as weight and distribution.

At some level, they must vary to some degree, and one could assume that this variation could also represent itself in repeated trials as a biased outcome.

As a student in computers i find this very interesting. I created a thread that talks about this type of stuff.
www.abovetopsecret.com...

OP, are you involved with creating evolutionary algorithms and neural nets then?

Your question about the balls is answered in a post above I made while you were making this one. That IS the theory. Minute differences in the balls causing repeated results.

No to your second question. I spent two years working on my stock market algorithm, and because of a combination of they way I think, a little luck, and using 4 computer processor cores every night for 16 hours straight, I nailed it. Other than that, I don't really feel the motivation to do other complex experiments other than this lottery thing... which took me about 2 hours to "Format" all of that past winning number data, and about 10 minutes to write the computer program. (I used a Macro... a keystroke recorder/repeater to automate most of the formatting of the past winning numbers)

I could do the lottery thing with "Lotto" type games, but the amount of money that needs to be spent in the first trial before the first or second win, would be way too high because you would have to play about 5 to 10 different groups of lotto-type numbers for each drawing.

Edward Slayton

You used to be able to improve your payout by NOT picking mega lottery numbers in date range EG 1-31.

Many people play birthdays and other sentimental dates so there are more winners that share those low number choices.

Originally posted by St Udio
reply to post by EdwardSlayton

 

i looked at the same numbers, boxed or straight...for the SC drawings:

Digit1 Digit2 Digit3 Draw Date
3 0 3 Evening 10/26/10
3 3 0 Midday 07/31/10
3 0 3 Evening 06/04/10
3 3 0 Midday 02/06/10
3 0 3 Evening 07/25/09
0 3 3 Midday 01/20/09
3 3 0 Midday 07/01/08
3 0 3 Evening 01/13/08
3 3 0 Evening 01/03/08
0 3 3 Evening 04/28/07
3 0 3 Evening 04/03/07
3 3 0 Evening 07/04/06
0 3 3 Evening 03/15/06
0 3 3 Midday 12/12/05
3 3 0 Evening 12/05/05
0 3 3 Midday 06/13/05
3 3 0 Evening 05/09/04


result -> the payouts aren't worth the investment -> the norm is a hit 3X per year, but in 2010 the combo hit 4X already



next... you have to really trust the source as reliable... that the numberd balls are changed once a year,
then you have to believe the #'ed balls are never touched or cleaned by human hands


i dismiss your theory, its certain that the balls are cleaned of the residual human hand oils, from being touched at each drawing (off the air) to replace into the mixing bin.


your thread sounds impressive... but thats the difference

You missed something in my first post. 033 hit boxed and/or straight 11 times in 19 months. Now look at the years for what YOU just did for SC and how many hits. 5 times within the same period. 033 comes out straight a lot.

We'll see if it works. The ball sets ARE really changed and used as stated. It's the law and people could sue the POOP off the State of Florida for shenanigans. It's more regulated than most people think.

Edward Slayton

Originally posted by Epsillion70
Then you might be talking about,
Bayesian inference or maybe
Frequentist inference
?

edit on 21-12-2010 by Epsillion70 because: Edit spelling

That's it!

Edward Slayton

I tend to think the OP is using a Heuristic approach to something that is essentially a random event on each lottery draw and each numbered ball having the odds of 0.5 each time. Am I wrong?

Originally posted by Epsillion70
I tend to think the OP is using a Heuristic approach to something that is essentially a random event on each lottery draw and each numbered ball having the odds of 0.5 each time.

Well it's "Supposed" to be a random event. But my theory is that, with enough data from the same balls, it's NOT random.

And I'm aware of the other, sitting-at-the-water-bong coffee table fact that NOTHING is truly random. But I've done pretty well so far with my stock market AI algorithm. I don't have to work any more. So I would be confident enough to say that the odds at least lean in my favor with this experiment, and at least as much that is spent will be won (break even)... given the 2 to 1 win ratio on past results of 033 Str/Bx.

Edward Slayton

reply to post by EdwardSlayton


the topic of random in computing is very interesting.
at one level of reality, the newtonian level, things do seemed to be governed with to much randomness.
however, i feel the smaller we try to perceive reality, the more randomness is inherent in the system. Or atleast, random to our newtonian perception.

however, we cant simulate pure randomness with software alone correct?

reply to post by EdwardSlayton

But my theory is that, with enough data from the same balls, it's NOT random.

Years ago I played around with a system that mimicked random coin tosses. If you toss a coin a trillion times you will almost certainly find a run of coin flips in your data where the coin comes up heads 28 times in a row.

It did seem that these longer run events showed up more predictably in the data. The counts for shorter runs of less than 8 consecutive heads or tails vacillated as the data built, taking random walks before returning to the statistical mean expected. Probably just an illusion unless randomness has a temporal quirk that has not been widely published.

Originally posted by VonDoomen
reply to post by EdwardSlayton

 

the topic of random in computing is very interesting.
at one level of reality, the newtonian level, things do seemed to be governed with to much randomness.
however, i feel the smaller we try to perceive reality, the more randomness is inherent in the system. Or atleast, random to our newtonian perception.

however, we cant simulate pure randomness with software alone correct?

That is correct. That's exactly how my secret hedge fund software works. It uses random numbers to look for candle patterns that are too small for the human eye to see. However, the secret of getting it to work was to run the previous market data through approximately 4 times the number of cycles that one would THINK should be necessary to find all relevant data points. And it still isn't perfect, but you don't need perfection when using AI to find something the human eye or human mind is incapable of finding.

Originally posted by Bordon81

Years ago I played around with a system that mimicked random coin tosses. If you toss a coin a trillion times you will almost certainly find a run of coin flips in your data where the coin comes up heads 28 times in a row.

It did seem that these longer run events showed up more predictably in the data. The counts for shorter runs of less than 8 consecutive heads or tails vacillated as the data built, taking random walks before returning to the statistical mean expected. Probably just an illusion unless randomness has a temporal quirk that has not been widely published.

I can do that right now with a DIGITAL coin. Shouldn't even take a few minutes. But I'll do that LATER.

For NOW though!... the subject at hand!!!!!!!!!!!!!!!

My AI number:
033

Today's daytime number:
073

Won't be long.

Edward Slayton

Here's the digital coin flipping experiment. It took 139 million flips to get 31 in a row. Other records are shown below on the only run I did of the program...

0's in a row: 6 Flip# 11
0's in a row: 9 Flip# 395
0's in a row: 10 Flip# 829
0's in a row: 11 Flip# 2,307
0's in a row: 13 Flip# 8,281
0's in a row: 17 Flip# 12,465
0's in a row: 19 Flip# 253,158
0's in a row: 23 Flip# 5,238,673
0's in a row: 24 Flip# 49,394,324
0's in a row: 25 Flip# 72,642,439
0's in a row: 31 Flip# 139,200,968

It only took about one minute to get this high. I stopped the program after 31 because this old internet-only computer I use has a tendency for the processor to overheat. And the type of program above, and also full-screen YouTube or full-screen Hulu TV, will max the processor to 100% and overheat it.

Edward Slayton

reply to post by EdwardSlayton


My brother wrote a program that works backwards from a random sampling and finds the source of a the Random Number Generator. It uses things found in nature, simply put it uses something similar to the Golden Ration, it extrapolates backwards 1000's of times over, until it finds a set of generation points, and then it generates forward from there and compares to the real data set, until it gets a very near match. Then, it "predicts" the next point.

Using his system, and 25 recent drawings, I was able to win the $500 prize 3 times in 2 weeks, and I spent less than $100. It was amazing!!

Then, it stopped working. I spent about another $250 trying different variations, but it never hit again?

I have no idea why it was so accurate for such a short time. I can't believe it was just dumb luck, but then again I can't see any possible way that it could have stopped so abruptly?

Anyhow, Good Luck to you!

reply to post by getreadyalready


Did you keep checking the numbers even though you stopped betting on them? I would think the real test would be to see if the initial $100 (plus all winnings) would ever dry up. That's what I'm attempting in THIS thread. And as I said, I'll keep running the program and if the highest likely number ever changes, I'll change the number too. I don't think it's going to change though, because the only thing that makes 033 so profitable is the fact that it's comes out Straight so many times.

Maybe they changed the ball sets if he and you didn't know anything about it.

The number that comes out more often in Florida's Cash 3 as a 6-way boxed is 135. But it would only have won half the money spent over the 19 months.

033 is it though. And today's day number of 073 is pretty darn close, and only took 2 drawings from the time I started this thread. The odds hitting two of the numbers within 2 drawings from thread start time are 1 in 50.

Edward Slayton