http://www.linkedin.com/...il=&ut=3HLgYAPv75ERQ1

The answer is no. Even if some balls come up slightly more often than others the answer is still no.
Some of the responses to the thread show the danger of using an "engineering" approach ie. applying theory broadly without really understanding it, which is what "data analytics" is in a way.

Google "sub martingale theory". There's your answer. Essentially for any game where the odds are not in your favor there is no strategy that will change the odds to be in your favor (barring cheating) and over a long enough time you *will* lose an infinite amount of money. That doesn't preclude one from being lucky and winning, but no matter what, if you play long enough you *will not* come out ahead. For example no combination of betting scheme in roulette in a casino will help you come out ahead. Same with craps.

Flip this around. This is why "risk diversification" works and why on average people do reasonably well in blue chip stocks over the long term. Same idea.

A good example of blindly using formulas without understanding them is the 2008 financial crises and credit default swaps. The formula/theory hasn't been proven wrong based on the assumptions stated. The problem was that the assumptions stated were ignored (ie that can never happen).
:)

*runs off*

Err the article linked didn't talk about why some numbers come up more than others, but talked about people betting non-randomly, so in order to maximize EV. But yes, it's probably still a negative EV proposition.

Also CDS is a slightly more complicated issue. CDOs for sure were mispriced with correlation and P(default) being estimated incorrectly as well as the copula used to create joint P(default) distributions not really working for the distributions in practice, and I guess to some extent the same assumptions were used for CDS but normally when people think of CDS being mispriced it's because the risk on the underlying was mispriced so no-arbitrage arguments led to CDS being mispriced as well. I haven't heard that much about CDS written on non-CDO assets blowing up, except in cases like Greece where the model itself was not flawed by the P(default) was "wrong" (or default just happened and people thought that a low probability event was a 0 probability event)

but i spent all last week converting my lottery spreadsheets into a database. though technically, i just got tired of copying the sheet. i am curious as to if the quick picks are bent. seem to see a lot of repeat numbers.

This was a good read. Thanks martian

is this related to that guy who won the the lottery seven times because he never changed his numbers and bought a lot of tickets?

ok martian, explain this to me as I'm not taking the math courses for my PhD until this fall:

if you have a game whereby the prize is divided up from the money each player pays to enter the game, minus a fee for the organizer, you've got a game where the odds are against the player, correct?

let's say this is a lottery with 2 numbers (0,1) and those players that pick the winning number automatically get 80% of the entry money pool, divided equally among them

if there is some sort of bias causing players to choose 0, but the numbers are equally likely to come up, wouldn't it be a very positive NPV strategy to pick 1?

so apply this same but on a less obvious scale to people picking the same numbers in a lottery, how does that work out?

bonus

@Pontius: Depends on how strong the bias is. There would be assuming that you were the only one who picked up on it and no one else knew (unlikely to happen irl over a long period). In that case the real odds are actually not against you but in your favor (along with the house).
What you are hitting on here is that probability can be subjective which complicates things.
The thing with the lottery is that even though you might be able maximize your odds, you haven't made a big enough difference to actually increase your expected loss to an expected overall gain. Losing less than everyone else is still losing.

IRL: this is why in some cases arbitrage works. The only problem is that people catch on really fast thus shifting the odds back to normal (although in the stock market they are not stacked for you to lose for the most part).

yeah I was just posting that to try to question whether the lottery is "unwinnable" theoretically, I understand that if there was an obvious arbitrage it would get arbitraged out, even if the game was full of completely uninformed players

Using advanced data analysis to determine if a RNG is truly random is pretty clever actually. The thread is about using analytics to determine if a game is truly not in your favor, it is not about using analytics to win a game which is not in your favor.

unlike the people who sell books on how to win at slots?:P