Just-Dice.com : Invest in 1% House Edge Dice Game

[TheFuneral]

....
I'm quite tempted to just give up on this whole thing.  It looks like the site is simply at nakowa's mercy.  He's promised me in the past that he won't play any more, but never sticks to his word.  So what do I do?  Continue to let the investors take these horrific losses?

I don't think it's fair to take down the site, but I also don't think it's fair for people to be mad at you when they lose. It's not your fault. Sadly I assume the latter will keep happening.

[Peter R]

I'm quite tempted to just give up on this whole thing.  It looks like the site is simply at nakowa's mercy.  He's promised me in the past that he won't play any more, but never sticks to his word.  So what do I do?  Continue to let the investors take these horrific losses?

Firstly, I want to compliment you on your brilliant ideas for just-dice.com.  I discovered it only a few days ago, and have been thinking about it ever since.   I really appreciated the analysis you did for SatoshiDice and I can tell this has given you a great deal of respect in the quickly-growing bitcoin community.   

In regards to taking the site down, well, you are a statistics expert:  Let’s make the null hypothesis “Nakowa is not cheating” and see what the p-value is when we test it.  Obviously, Nakowa is lucky, but if the p-value is, say, no smaller than p = 0.05, then I don’t think we can conclude that he is cheating.   The investors signed up knowing that there were risks involved, and shutting down the site would not be fair. 

Lastly, while I have no doubt that you and your team are completely honest and trustworthy, I suspect that if you take the site down without strong evidence that there is a real problem, it would sort of make you look guilty as per Lohoris’s insight “while I don't think [Nakowa is] cheating, it is worth noticing that if he was cheating, then pretending to be victim of a gambler's fallacy (like he is doing) is actually a great tactic to go unnoticed.”

Personally, I hope Nakowa comes back again soon...

[rigel]

Don't take me wrong dooglus

I's not your fault. Whatever happens I'm with you.

He is still ddosing the site? This is unfair
Why don't you just take down the site during ddos ?

[stslimited]

Some thoughts on the whale's win today (conclusion: he had a 67% chance of pulling it off).

The whale can STOP betting when he chooses, so the probability of what happened today is not as low as it seems.  I haven't dug up the exact numbers, but let's assume:

1. He wanted to win 2500 BTC (after which point he would stop gambling)
2. He was willing to go down as much as 10,000 BTC before giving up
3. He was betting 200 BTC each time and always attempting to 2X his bet (this just simplifies the simulation). 

I ran the following Mathematica code:

In[6]=
bankroll := 10000;
target := 2500;
n := 10000;

bet := If[Random[] > .505, 1, -1]

outcomes = Table[profit = 0;
   While[profit < target && bankroll > -profit, profit += 200 * bet ];
   If[profit > 0, 1., 0], {n}];

100 * (Plus @@ outcomes)/ n

Out[6]= 67.23

So, 67% of the times that he applies this strategy he would earn his 2500BTC target before going bust. 

This is exactly what I've done using martingales at bitzino's betting limits

Their table limit was 100 btc at the time, denominated in 10000 mbtc. My 67% was based on the fact that I could get martingale'd 6-7 times in a row within 150 turns, at any point of course. This was the chance I was willing to take, gambler's fallacy completely in mind, and in that many turns I could expect to make 30-40% gain.

I made 35 btc and was done.

[jonitas]

I'm quite tempted to just give up on this whole thing.  It looks like the site is simply at nakowa's mercy. 

"The temptation to quit is the biggest right before you are about to succeed." - Chinese proverb

Don't give up, there are a lot of investers who still believe in the site. And mathematically we will win in the long run. Just leave everything like it is, you have built a great website! Hopefully Nakowa winning big will draw the attention of other whales and everything will even out.

[jgarzik]

I'm quite tempted to just give up on this whole thing.  It looks like the site is simply at nakowa's mercy.  He's promised me in the past that he won't play any more, but never sticks to his word.  So what do I do?  Continue to let the investors take these horrific losses?

Several people who have run gambling sites chimed in on reddit, bitcointalk.org and IRC with a similar comment:  "1% is too low"

And on CYA front:  make sure the website text frequently notes the "high risk" nature of gambling and investing, because it is a startup business model and a startup website running on a startup currency

The insta-investing feature is nice in general, even if investors lost a shirt or two.  Innovative.

[romerun]

I think he relies a lot on the pattern of current random series, it usually goes like h l h l... and it some points it turns h h h... or l l l.. and that is when he releases his 10folds. If site changes distribution of random number, maybe having like 100 configurations, and randomly pic one for a random amount of time, even he tries to test the water with small bets, but on the real run the series may switch screwing up his strategy. Although I do believe, just lowing max bet will do it, like 50btc or 25 max, and that 1% edge will be useful. Or, if bet is more than 50btc, player will get 1.9% edge.., because it's more risky to the house.

[MSwiza2]

I'm quite tempted to just give up on this whole thing.  It looks like the site is simply at nakowa's mercy.  He's promised me in the past that he won't play any more, but never sticks to his word.  So what do I do?  Continue to let the investors take these horrific losses?

He is going to continue to play and has even said he tried to quit, but can't. 99% of people would do the same since it is basically like printing money. I think investors can make their own decisions, and you are not responsible for their losses, but at the end of the day you are responsible for your business and reducing your own individual risk.

I know the majority of people on here don't like or respect towtoad/Nakowa/or whatever he wants his name to be, but I think you do have to respect his winnings. Doing it once or twice might have been luck, but it appears he has found a way to consistently beat the game and that is not going to change any time soon IMO.    

[drawingthesun]

Dooglus, I am of the opinion that he will lose everything to the site, eventually. Taking the site down will rob the investors of that opportunity. Also we have close to 30,000 bitcoin invested so there is still confidence.

[jgarzik]

I look at it like beta software:  just iterate through the problems.   There are always variables to play with:  house edge, further limiting max bets, ...  before contemplating shutdown, in my self-interested and biased opinion.  

Also we have close to 30,000 bitcoin invested so there is still confidence.

+1    When users may vote with their wallets, you always know where they stand.

[Bugpowder]

Here is an analysis of cake's betting today using his real bets.

This same sequence of bets was performed 20,000 times with a 49.5% chance of winning on each bet.

He had a 17.6% chance to finish as well or better than what he did.

His average loss was expected to be 3,750BTC.

Of course with a random walk and a large bankroll, as long as he is disciplined to stop when up, he will probably finish up.  However, one day he may just keep losing and losing and losing.

[drawingthesun]

He had a 17.6% chance to finish as well or better than what he did.

When that percentage is constantly single digits then we worry.

[Bugpowder]

He had a 17.6% chance to finish as well or better than what he did.

When that percentage is constantly single digits then we worry.

It is probably in the high single digits for his combined sessions.  Waiting for that data to be provided.

[drawingthesun]

It is probably in the high single digits for his combined sessions.  Waiting for that data to be provided.

Dooglus just posted a tar file to the JD chat bug.

[spartanmorning]

Can someone explain to me why a site that allows the house to have both seeds/keys and can use that information against player money 'invested' in the bank's position, wouldn't be suspected of being a scam? Per your SD and other statistical analysis, it could easily manipulated by the con in several ways. One way would be to intentionally lose a number of bets to create a distribution to their liking and simply betting more less times.

To me it seems like a perfect setup for a con. I'm not saying it is a con, but if I were a betting man....

[drawingthesun]

Perhaps lowering the max bet would be a good idea. How many real life casinos allow for 1% of bankroll bets?

[dooglus]

Here is an analysis of cake's betting today using his real bets.

This same sequence of bets was performed 20,000 times with a 49.5% chance of winning on each bet.

He had a 17.6% chance to finish as well or better than what he did.

To be clear, is this what you are saying?

"If he had made that sequence of bets *with his eyes closed* he would have had a 17.6% chance of doing at least that well"

He didn't stop after those bets at random.  He stopped immediately after hitting a balance of 10k BTC for the first time.

So a better question, rather than "what is the chance he was doing well after exactly those bets" is "what is the chance he would eventually get to be 4430 BTC up if he bet that sequence of bets over indefinitely until he either got there or lost his entire bankroll (which is unknown, but at least 10k)"?

[dooglus]

It is probably in the high single digits for his combined sessions.  Waiting for that data to be provided.

Dooglus just posted a tar file to the JD chat bug.

Here: http://just-dice.com/nakowa.txt.bz2

I may have missed some of his accounts - he tends to disown the ones that lose I think (like gigad and alison who lost a total of 420 BTC between them when he was first coming back around the time that LD launched).

[Bugpowder]

Here is the expected outcome distribution of all of Nakowa's bets, (excluding the 3% that were of non 49.5% odds of winning).

1.9% chance of doing at least this well if he bet the same sequence again.

Expected profit of -8,522 BTC

[Bugpowder]

Here is an analysis of cake's betting today using his real bets.

This same sequence of bets was performed 20,000 times with a 49.5% chance of winning on each bet.

He had a 17.6% chance to finish as well or better than what he did.

To be clear, is this what you are saying?

"If he had made that sequence of bets *with his eyes closed* he would have had a 17.6% chance of doing at least that well"

He didn't stop after those bets at random.  He stopped immediately after hitting a balance of 10k BTC for the first time.

So a better question, rather than "what is the chance he was doing well after exactly those bets" is "what is the chance he would eventually get to be 4430 BTC up if he bet that sequence of bets over indefinitely until he either got there or lost his entire bankroll (which is unknown, but at least 10k)"?

Here it is with a 10,000BTC initial bankroll and if he stops if he touches +15,000BTC

5.8% chance of massive success, 66.7% chance of blowing up.

Code:

bets = []
outcome = []
nSessions = 10000

    bets = nakowaCrop(:,4)*1e-8;
    outcome = nakowaCrop(:,6)*1e-8;

shuffledResult = sign(rand(length(outcome),nSessions)-.505);
shuffeldOutcome = repmat(abs(outcome),1,nSessions).*shuffledResult;

%%


cso = cumsum(shuffeldOutcome,1);

%%
bustedIdx = repmat(50000,10000,1)
winIdx = repmat(50000,10000,1)
for i = 1:10000
    try
bustedIdx(i) = find(cso(:,i)<-10000,1);

    end
    try
winIdx(i)    = find(cso(:,i)>15000,1);
    end
end

finalOutcome = sum(shuffeldOutcome);

finalOutcome(find(bustedIdx < winIdx)) = -10000
finalOutcome(find(winIdx <  bustedIdx)) = 15000


%%
figure(1);clf;hold on
hist(finalOutcome,100)
%plot([15000 15000],[0 1800],'r--')

title('Outcome distribution if Nakowa stops at +15k with 10k starting bankroll')
ylabel('# of outcomes')
xlabel('Nakowa''s total profit (BTC)')

probProfit = 1-sum(finalOutcome<15000)/10000;
probBust = 1-sum(finalOutcome>-10000)/10000;
text(12000,1000,['p = ' num2str(probProfit)],'color','r')
text(-9000,1000,['p = ' num2str(probBust)],'color','r')
set(gca,'xlim',[-10500 15500])
print(gcf,'-depsc','NakowaStopping')