Quote from: Bitsinmyhead on September 30, 2013, 11:47:31 AM

Can't believe you smart math guys keep arguing against idiots about this invest/divest stuff instead of using your smarts to do something like this. (The irony is that those idiots are actually correct at the moment, because right now being invested is -EV by itself).

Some quick back of the envelope calculations to show why you should probably all just divest:

Wagered on site: 3.6M
Expected profit: 36 000
Actual profit: -3 000

Site is 39 000BTC under expectations. Now let us simplify and assume all bets made were 200BTC bets. (The actual situation will have less variance than this because most of nakowas volume has actually been on bets less than this.)

Total bets made 3.6M/200 = 18 000

Variance of 18 000 bets of 200BTC at 49.5%: 200 * sqrt(0.495*0.515*18000) = 13 500

Site currently running 39 000 / 13 500 = 2.88 st devations under expected.

Conclusion chance of the site running this poorly by chance: Way less than 1%.

You guys keep saying trust the math, but your conclusion is all messed up. The math is telling you to GTFO, chances of something being wrong here is so high that staying invested currently is -EV!

The calculation is incorrect I believe, the probability of the profit being worse than it is is around 7% according to my calculations.

Assume each bet is 200BTC, the probability of casino winning is 0.505.
Let X be the random variable representing the casino profit on such a bet.
E[X] is 2, and Var[X] is 39996.
From the central limit theorem, the difference between the observed profit over the 18 000 bets and the expected profit, divided by the square root of the number of bets (18000) is close to normally distributed with mean 0 and variance the variance of X, 39996.
-39 000 / sqrt(18 000) ~= -291.
And the probability of a normally distributed variable with mean 0 and variance 39996 being less than or equal to -291, is (according to wolframalpha) around 0.073, i.e 7.3%.