Topic: Just-Dice.com : Invest in 1% House Edge Dice Game - page 100. (Read 435357 times)

\
I said improbable, not impossible. And I am talking over the longer term since the site started.

Yeah, it definitely doesn't seem like the 1% house edge is working as intended. The only way investors can make profit is by investing and divesting constantly to take advantage of sharks.

i am also for this

I would move beyond discussion of whether Nakowa is cheating at this point (which would require the OPs help which you have to trust to invest here) and instead more simply ask if the site functioning as intended? It seems we consistently come up with improbable results with yesterday just being the latest example. As an example, gambling for 14 hours straight and mostly spamming max bets should give the house edge an opportunity to work. In effect, this is acting more like a fair coin flip than a weighted dice roll.  That is on the gambling side.

On the investment side, the intention was for player profit or gain  = player's % of bankroll X site profit or gain.   I understand there may be some slight deviations from that, but myself and all the other "passive" investors are experiencing significant deviations from this expectation.  Personally, my personal loss was about 35% greater than expectations for the last session based on the formula above. This is due to meaningful fluctuation in the investment amount at the time of large fluctuations in the house profit.  

If Dooglus is okay with these phenomenon on both the gambling and and investment side, then let him say so and that will be the end of that.  If either of these are deviating far from Dooglus' expectations and plans for J-D, then some sort of corrective action should be considered.

Greed is bad

We should keep this possibility in mind. It would be great if doog could deliver more information about how the server seed is saved.

This sounds like a reasonable idea. You'd also have to set a waiting period for withdrawal to avoid the creation of new accounts to reinvest immediately.

Very good post oda, I agree especially on you analysis on "investment daytrading".

Regarding the competitive advantage of the 1% edge, it looks to me that 1.5% might be less attractive for guys like Nakowa, but nevertheless investors would make more money which is ultimately the goal of the site. I might be wrong and we could lose too much volume, it's difficult to calculate.

Finally, I see the chance of Nakowa cheating almost none. Yesterday he lost two smallish deposits and he did a bigger one to cover that losses, as a degenerate gambler would do. He started with a high bet of 130BTC, after a few hours he went to 170BTC, and the last hours he just went berserk and started to make bets as high as max profit allowed hin. The amount if high bets in relation to the smaller ones also increased dramatically at the end of his session, like he was getting anxious.

Finally, in the first few hours he took the house to - 6k, which meant +4k for him, but nevertheless he didn't stop because his goal was to cover the 5k loss of his previous two sessions. Finally and after 14 extenuating hours he stopped by winning 1k coins, but in the process he risked up to 8k. First, that doesn't seem like an unlikely outcome. Secondly, all the above completely looks to me as the behavior of a deluded compulsive gambler, and not the behavior  of a cheater.

This is wrong. It fits perfectly as there is only two possible outcomes win or loss and when n is large a reasonable approximation to it is given by the normal distribution.

fair enough, i guess we should have defined what we are talking about better

In the original quote stream was the following comment:


If a "TRNG" replaced the hmac-sha512 with server and client seeds scheme it would be impossible to prove if the numbers were fairly generated.

I deposited 2 btc and end up with 15 btc atm
Feeling lucky, but I should stop before loosing everything...

yes, i agree, but this is just a question of using the OS prng, which falls under correct implementation. This may seem like tautology, but i was just trying to say the the system that jd uses is nothing revolutionary or new per se, it is just a collection of known and tested cryptographic methods. As i said, i do not feel that i made a very strong claim, my point was that the relatively simple system that jd uses should provide secure lucky number generation, if the algo guidelines were followed (as opposed to using a new, untested, faster hash algo that did not face rigorous testing)

Okay, this is going to be a longer post, responding to a number of people that wrote interesting comments. But first, can I please make one request:

Can we please try to retain a minimum of respect for each other in this discussion?

If find it incredibly frustrating that it's completely normal to throw around insults to someones intelligence in here because he or she disagrees with you on one or the other matter. If somebody is simply wrong (in an obvious, mathematically provable way), explain why he is wrong. But if a number of people, many of which have an education in a relevant field (CS, math, anything else with statistics) disagree on a non-trivial matter, then please keep in mind that you're poisoning the atmosphere and effectively contribute to reducing the signal/noise ratio if you throw around those insults.





That said, first reply:


The problem, in my opinion: too many simplifying assumptions, therefore the result is somewhat off. For example, a large portion of the current  cumulative site loss was caused by early nakowa wins, when we still used full kelly, and max profit was significantly larger than 200.

So here's a similar calculation that IMO makes slightly less simplyfing assumptions, about the likelihood of nakowa winnings yesterday:

Yesterday's amount wagered: 1.5M, a huge portion of which came from nakowa, whose bets where, by a vast majority, at max profit (180-190). The same calculation as above: (1500000*1%)/(180 * sqrt(0.495*0.515*8333)) = 1.8 standard deviations below expected profit.

Quite lucky? Sure. Enough to reject the null hypothesis (fair RNG, no cheating)? Not quite.

Note that this is also in line with Peter R's results who keeps posting updates on the likelihood of nakowa's cumulative winnings.




Second reply:


and


I personally would prefer if we go on with a 1% house edge for now. It gives the site a competitive advantage which might otherwise be lost.

But about the other point (investment "daytrading") I agree. Which brings me to reply #3:





There's at least one mathematically sound reason why one might disagree with the claim that "daytrading" is entirely random. So, first, please stop throwing around insults against people's intelligence.

That reason, no matter how unlikely, is that if nakowa would actually be cheating (he hacked j-d, found a flaw in the RNG, whatever), then his winning patterns are not completely random either, so "daytrading" in consequence could become (partially) deterministic as well. Agreed?

Please note: I'm not saying that's necessarily the case (that nakowa is cheating), but it's at least conceivable.

More importantly however, whatever you think the reason for this is, please look at nakowa's yesterday winning charts and tell me that there is no "exploitable" pattern to it:



At around bet 6000 or so, some people started a strategy that reversed investing/divesting at around the -2000/2000 mark (I was on j-d chat yesterday). Which turned out to be a rather profitable strategy.

Now, please read the following: Yes, I know. In the long, no such strategy can be profitable (unless nakowa's results are not random either). But yesterday, this sure tweaked the distribution of profits and losses.

So here's my opinion on this matter:

If you want to gamble on j-d, gamble. If you want to invest, make it an investment that lasts longer than a few minutes.

Someone here said he sees very little harm done by investment daytrading: I disagree. In the short run, some successful "daytraders" will skew the distribution of profits/losses among investors, which makes the losing investors rather unhappy, which in the long run hurts the site, because they might take out their investment and not come back.

Additionally, I see absolutely no benefits from investment daytrading. Again: if you want to gamble, you can do so by the gambling mechanism. I don't know where the idea comes from that the investment mechanism has to contain a gambling element as well, especially if (for the psychological reasons described above) it arguably hurts the site overall.

A possible solution:

I think a short, but noticeably invest/divest delay would be helpful.

Here's how it could work: Divesting is always instantenous.  Your initial invest is as well. But after any divest, another invest has to wait 1 hour. That's all.

You can still take out your investment whenever you want. You can still invest/divest several times a day. But "trading" swings when a whale is on the loose should become impossible.

Let me repeat one more time: Assuming the site works as intended (no cheating going on), then I know investment "trading" is -EV,  just like gambling. But in the short term, strategies based on the winning patterns of a whale can fuck up the overall investment situation, and will make many investors understandably unhappy.

Well, the binomial distribution does not really fit here, as it measures the number of successes without placing any weight on losses.

I already know the clientseed and the nonce. If the serverseed is poorly generated you can use the best hashing algorithm available and I'd be able to predict the output.

Please don't feed the troll. Just look away and go on.

the debate was: generation of numbers which determine if a given roll wins or not.

i stipulate that as the sha has a uniform output distribution, the results are evenly distributed. Furthermore, due to the security of the algo, we are unable to predict the next roll without knowing the plaintext (clientseed+serverseed+nonce). Therefore, the generation of lucky numbers is fair, contingent on correct implementation (which i know is not a simple issue)

why do You think it is so unimportant? but it is not possible to determine or predict anything about the output of the function without knowing the server seed (see above).
who could know the server feed? no one? IMHO it should be discussed if it is provably fair only for the player? but how about the investors? is JD provably fair for the investors?

What is nakowa doing?

Its supposed to be standard deviation not variance, edited that mistake.
Using the formula for the standard deviation of a binomial distribution:  √[np(1-p)]
n= 18000 bets, p = 0.495 and (1-p) = 0.505