Quote from: Deprived on June 01, 2013, 04:19:10 PM

Quote from: greatscott on June 01, 2013, 04:02:47 PM

Quote

Asicminer is not setting any sort of precedent. Dividends used to be weekly back in the old Global Scam Exchange days (with some daily). MPEx set a precedent by getting rid of the sort of bullshit. The only surviving corp off the Global Scam Exchange hasn't yet gotten the memo, they're stubborn. They'll learn. So will you.

Did I strike a nerve? Anyways, I didn't know - but even if it's not a precedent, I think moving to a weekly dividend pay-out could make confidence in the stock return more quickly...

Quote

You need to read up on Bitcoin. Don't start with the concept of provably fair, start with the very basics.

Quote

You need to read up on cryptography. Don't start with "why does a hashing function have to yield uniform distributions", start with the very basics.

I think you misinterpret what I meant. Provably fair means that the game isn't rigged (i.e. the house isn't cheating the gambler), not that the statistics of the game itself are good. I know Satoshi Dice is provably fair: the public keys (that represent the secret daily message) were put into the blockchain before the site's launched. I've gone through the whole verification process of both validating that the secret matches its public key from the SHA256 hash, and that the "lucky" number is properly produced, by doing the keyed SHA512 on the txid+secret message and looking at the top 4 digits. This part of the system certainly work.

I'm talking about the casino statistics of the game itself, and setting the game parameters (such as allowable min/max bet) so that the profit/loss variance is smaller (i.e. you're less likely to report a loss for the month). It is probably uniform, but if the distribution of "lucky" numbers happen to not be uniform, then this could be probabilistically exploited. E.G. If numbers < 32,000 tend to lie more in the 1:16k range than in the 16k:32k range or vice-versa. Something like this could contribute to a bigger profit/loss variance or bias it.

Weekly dividends would increase the apparent variance not reduce it.

In terms of reducing the variance in general there's a few ways to do it :

1. Increase the house edge. Can't see this being a flyer. A fairly low house edge is one of the best means SD has to make it hard for significant competition to arise.
2. Reduce max bet size (variance is much lower on 100 bets of 1 BTC than 1 bet of 100 BTC). This isn't viable for very similar reasons. People who want to make a 100 BTC bet won't settle for making 100 1 BTC bets.
3. Remove the higher-odds wagers. Not a good idea - if SD doesn't provide them, other sites will.

Which leaves the ONLY viable way to reduce variance without losing market share - increase volume. That's what efforts need to be focussed on, not analysis to establish very basic facts already known (e.g. to minimise variance you'd ony allow bets of exactly the same size on a single bet type). You have work within the range of what those betting actually want - and, unfortunately for SD, people don't want to bet a fixed amount on a single bet. They want to be able to try to design variations of Martingale and other betting systems. Whilst none of those systems can work some DO add a lot of variance. Variance is the price the house has to pay to get people to wager on an unfair game (by which I mean no more than a game where one party has an edge). You beat variance with volume - and over the long-term - not by driving away customers to try to avoid short-term losses.

Great counter, thanks. I'd just like to see the perception of SD itself look more like a business with a steady revenue stream, rather than a gambler itself. I think you're right - the long-term key is volume. Nonetheless, it does seem that SDice might be due for a good optimization.