Lottery winning techniques

[T'ai Chi]

Say you're interested in a lottery winning technique (who isn't).

Basically in real life they all fail because each draw is independent of the historical draws.

But hypothetically suppose at least some thing is the same between sucessive lottery draws (the balls used, the machines, the people, something). Then what would your technique be?

Some possibilities:
-using the history, keep track of the most frequent picks, and select these next draw.

-using the history, develop a more sophisticated model to predict new pics

[wahrheit]

What do you mean, "suppose at least some thing is the same between successive lottery draws"?

If it's a real lottery, it's random. If it's random, you can't predict it. The ball #23 really doesn't care how it performed in the past.

[baron]

What could there be to make each draw NOT independent of the previous one? It wouldn't make any difference if the same machine were used, or the same set of balls, unless the No. 32 was the size of a grapefuit and didn't fit through the chute maybe, or No. 7 was habitually coated in chewing gum and tended to stick to the sides of the container, but these things would likely be spotted at an early stage.

[T'ai Chi]

What could there be to make each draw NOT independent of the previous one? It wouldn't make any difference if the same machine were used, or the same set of balls,

Well there would probably be an ever-so-slightly higher probability of a certain ball coming up more frequently. Like say the same balls are always used. One day the #32 ball gets a slight knick in it, or some fingerprints, goes unnoticed. It might throw off its 'uniformness' ever so slightly.

It is not too unlikely. This, and minimizing other ways of cheating, is the precise reason why lottery and casinos replace their randomization devices afterall. :)

[Starthinker]

Funny, I was just thinking of the lottery this morning. If my odds are 1 in a million for each draw, and I buy two tickets, for different drawings, can it be said that my odds are now 2 in 2 million?

As for the OP's original question, I think if everything is the same in every single draw and you have a good sample to go by, say, a thousand draws, then you would probably see more straight-line graphs than bell curves so it wouldn't help anyway. I once ran a lottery simulator on an old Compaq Luggable (yes, an 80286) and let it run for a week. At the time it was 6 numbers from 1 to 44. When I finally stopped the program all 44 numbers were picked almost the same across the board. A straight line. I think the numbers were something like 109,XXX with the difference between the highest and lowest being less than 300. Nothing in that little experiment would lead me to pick any one number over any other number.

[Ladewig]

Some possibilities:
-using the history, keep track of the most frequent picks, and select these next draw.


There is only one reason to do that (see wikipedia entry (http://en.wikipedia.org/wiki/1980_Pennsylvania_Lottery_scandal)), but given the advances in security in the past 27 years, that reason has been removed.

A lengthy discussion of the topic can be found here: JREF thread "Netlotto.org - Increase your chances of winning the lottery!' (http://forums.randi.org/showthread.php?t=52044)

[T'ai Chi]

As for the OP's original question, I think if everything is the same in every single draw and you have a good sample to go by, say, a thousand draws, then you would probably see more straight-line graphs than bell curves so it wouldn't help anyway.


In that simulation I agree that would happen and wouldn't be too useful. It is not clear that there is anything in that simulation that accounts for the real-life physical 'erosion' of the balls, dust in the machine, fingerprints on the balls (ok sounds bad), small cracks, stuff like that.

I'm more talking about going either selecting based on frequencies from the real historical draws, or using that data to build a more accurate model.

[Ladewig]

Well there would probably be an ever-so-slightly higher probability of a certain ball coming up more frequently. Like say the same balls are always used. One day the #32 ball gets a slight knick in it, or some fingerprints, goes unnoticed. It might throw off its 'uniformness' ever so slightly.

It is not too unlikely. This is the precise reason why lottery and casinos replace their randomization devices afterall. :)


Except that a knick in that ball would not change its weight and lots of fingerprints would not change its weight. Even more importantly, lotto officals and their regulators routinely weight the balls to ensure no tampering has occurred.

An additional reason is that there are millions of players that track numbers and bet only on the numbers that appear most often. If the number of winners were significantly higher than expected, the officials would immediately examine the balls for evidence that the drawing was not as random as they had hoped and could void the win if they discovered noticeable differences in the weights of the balls.

[Ladewig]

I'm more talking about going either selecting based on frequencies from the real historical draws, or using that data to build a more accurate model.

I don't understand the difference between the two situations that you mention. Would you elaborate?

[Rasmus]

The German Lottery has nearly 14,000,000 different number combinations.

How many drawings would it take until you could notice a significant glitch in the system? (There never were more than ca. 150 drawings per year, so no more than a few thousand games played total!)

And how big would a glitch have to be before it could actually be taken advantage of? Even if I could reliably predict one number in every game, my chances of winning would still be practically zero.

I haven't done the math again, but I think there is a >50% chance of not getting a single number correct. You need 3 numbers to receive any money from the game...

[casebro]

I have a simple system, guaranteed to double your chances of winning. If it works for you, send me 50% of your winnings.

So simple, better inprovements than all those mathematical probabilities that only make a slight improvement.

Here it is:

Buy TWO tickets.

[Mojo]

Say you're interested in a lottery winning technique (who isn't).

...what would your technique be?


Time travel.

[Mojo]

Buy your ticket at the last possible moment, and collect any winnings as soon as possible. In the event that your ticket wins, this minimises your chances of dropping dead before you can collect.

[JoeTheJuggler]

Say you're interested in a lottery winning technique (who isn't).

Basically in real life they all fail because each draw is independent of the historical draws.

I think you're starting off from a false premise. Any lottery winning technique fails because the odds of the game are set up to make picking a winner an extreme long shot.

Each draw being independent of historical draws is part of it being a fair game. If it were otherwise, it would certainly be illegal. If there is a tiny bias that comes right before the numbered ping pong balls are replaced, it would be way too insignificant to matter. Seriously, you're far more likely to die in a car accident than win a big prize lottery.

[BrooklynAndy]

Take lottery money... put in deposit envelope... put in bank

Repeat every week for one year

Look at bank statement only at end of year.

Congratulations, you are a winner.

[baron]

In that simulation I agree that would happen and wouldn't be too useful. It is not clear that there is anything in that simulation that accounts for the real-life physical 'erosion' of the balls, dust in the machine, fingerprints on the balls (ok sounds bad), small cracks, stuff like that.

"Small cracks" sounds worse than "fingerprints on the balls" but that aside, I see what you're saying. In the instance where defects were allowed to go uncorrected then the only reasonable method would be to base future draws on the probability of past draws. It would be very unlikely to provide benefit, but it logically could not harm your chances and there's a fraction of a percent chance it might improve them.

[mijopaalmc]

As I understand it, most lotteries switch out their precision-engineered balls for new one so that the peculiarities in of the balls don't produce long-term patterns.

[ben m]

Please think about what sort of "pattern" you might expect: say, "the 3 ball comes up X% more often than random".

Calculate how much historical data you would need to notice a non-null pattern.

Calculate your advantage over the odds given that you know about the non-null pattern with the confidence possible with the statistics.

If this advantage doesn't exceed the house advantage, you're still losing.

The house edge in state lottos is typically 30-50%. http://articles.moneycentral.msn.com/SavingandDebt/Advice/The10BestStateLotteryBets.aspx. Good luck.

[bruto]

I can see the hypothetical possibility if indeed the same balls and machines are used each time, but even so, it would take a great deal of time to find any bias, and even then, it probably would not result in more than a slight edge, a tendency for certain numbers to come up more or less often. It would still be a gamble.

Tai Chi, if you've never read it, you might enjoy a book called The Eudaemonic Pie, by Thomas Bass. It recounts the semi-successful attempt of a bunch of computer geeks to accomplish a similar result with roulette. Roulette is easier than Lotto because you get more than one spin of the wheel in a week, of course. These guys actually bought a roulette wheel and studied its behavior in order to devise a computer that could then analyze the results of multiple spins. They found that they could increase their odds to a small but significant degree by using the program to place bets within a particular octant of the wheel. It was never precise enough to hit specific numbers, but was able to tip the odds enough to give the long term advantage to the player rather than the house. edit to add, ....and the house advantage is much smaller in roulette than it is in Lotto.

I doubt you could ever accomplish a similar result with lotteries, even if you did find the bias and analyze it, simply because of the frequency of play. You'd need several lifetimes to make it pay off.