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

[stripykitteh]

I agree, as long as we keep the tinfoil hats off (celeste has a special edge/cheat), then this is normal variance due to an abnormally large bankroll.  However, it is normal variance which can scare off investors and if celeste does stop playing or continue staying lucky, can dig J-D investors in a hole too deep to earn our way out from in a sane amount of time.  I just know that something needs to be changed - we do not need to have so much larger a max profit per bet than our competitors.  We not even in the same universe with them - they are 3 to 5 BTC and we are 270 BTC?

Apparently dooglus derived 1% of the bankroll as the maximum bet by trial and error but as it turns out it's been mathematically proved that this is the most efficient maximum bet to optimally grow the bankroll. The proof is called the Kelly Criterion and investors have known about it for years:

http://www.investopedia.com/articles/trading/04/091504.asp

Kelly % = W – [(1 – W) / R]

Where:
W = Winning probability
R = Win/loss ratio

W = 0.505 (bank's edge)
R = 1:1 (profit and loss are both equal for a 2x bet)

Kelly % = 0.505 - [(1 - 0.505) / 1]

= 0.505 - 0.495 = 0.01, or 1% of the bankroll.

Note that a full Kelly Criterion doesn't try to minimize volatility; rather it maximizes growth in the long term.

If you are a short-term investor J-D may not be the right investment for you.

[1713879127]

1713879127

[1713879127]

1713879127

[1713879127]

1713879127

[1713879127]

1713879127

[HurtK]

^ exactly, keep it as it is

[Bugpowder]

My initial analysis of celeste's betting, based on the total betting results generously provided by Dooglus and celeste here https://just-dice.com/whale.zip.

I took the bet sizes and randomly assigned the win/loss outcome assuming a 1% edge on 50/50 (I neglected the fact that some were 49.5 bets and others were 50, it would make almost no difference).  

After 20,000 shuffles, this is the outcome distribution, with celeste's actual outcome highlighted.

In short, it is lucky (top 7% of expected outcomes), but not truly exceptional nor significantly different than the null hypothesis (which is celeste's system is = to random ordering of betting).

 

Audit the MATLAB code

Code:

bets = []
outcome = []
nSessions = 20000

for i = 1:length(celeste)
    bets = cat(1,bets, repmat(celeste(i,3),celeste(i,1),1));
    outcome = cat(1,outcome, repmat(celeste(i,4),celeste(i,1),1));
    
end

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

figure(1);clf;hold on
hist(sum(shuffeldOutcome,1),100)
plot([sum(outcome) sum(outcome)],[0 650],'r--')

title('Outcome probability of celeste betting')
ylabel('# of outcomes')
xlabel('celeste''s total profit (BTC)')

probProfit = 1-sum(sum(shuffeldOutcome,1)<sum(outcome))/20000;
text(2000,680,['p = ' num2str(probProfit)],'color','r')
print(gcf,'-depsc','Celeste')

[mechs]

I agree, as long as we keep the tinfoil hats off (celeste has a special edge/cheat), then this is normal variance due to an abnormally large bankroll.  However, it is normal variance which can scare off investors and if celeste does stop playing or continue staying lucky, can dig J-D investors in a hole too deep to earn our way out from in a sane amount of time.  I just know that something needs to be changed - we do not need to have so much larger a max profit per bet than our competitors.  We not even in the same universe with them - they are 3 to 5 BTC and we are 270 BTC?

Apparently dooglus derived 1% of the bankroll as the maximum bet by trial and error but as it turns out it's been mathematically proved that this is the most efficient maximum bet to optimally grow the bankroll. The proof is called the Kelly Criterion and investors have known about it for years:

http://www.investopedia.com/articles/trading/04/091504.asp

Kelly % = W – [(1 – W) / R]

Where:
W = Winning probability
R = Win/loss ratio

W = 0.505 (bank's edge)
R = 1:1 (profit and loss are both equal for a 2x bet)

Kelly % = 0.505 - [(1 - 0.505) / 1]

= 0.505 - 0.495 = 0.01, or 1% of the bankroll.

Note that a full Kelly Criterion doesn't try to minimize volatility; rather it maximizes growth in the long term.

If you are a short-term investor J-D may not be the right investment for you.

You might be right - my expectations were for up days and down days but not this level of volatility.  What I wonder though is our bankroll large enough to support the Kelly Criteria or can we be bankrupt (or lose enough investors flee) before the Kelly Criteria can assert itself.

[wolverine.ks]

My initial analysis of celeste's betting, based on the total betting results generously provided by Dooglus and celeste here https://just-dice.com/whale.zip.

thanks bugpowder, this was very informative. keep up the good work. id love to see more stuff like this.

[Bugpowder]

I agree, as long as we keep the tinfoil hats off (celeste has a special edge/cheat), then this is normal variance due to an abnormally large bankroll.  However, it is normal variance which can scare off investors and if celeste does stop playing or continue staying lucky, can dig J-D investors in a hole too deep to earn our way out from in a sane amount of time.  I just know that something needs to be changed - we do not need to have so much larger a max profit per bet than our competitors.  We not even in the same universe with them - they are 3 to 5 BTC and we are 270 BTC?

Apparently dooglus derived 1% of the bankroll as the maximum bet by trial and error but as it turns out it's been mathematically proved that this is the most efficient maximum bet to optimally grow the bankroll. The proof is called the Kelly Criterion and investors have known about it for years:

http://www.investopedia.com/articles/trading/04/091504.asp

Kelly % = W – [(1 – W) / R]

Where:
W = Winning probability
R = Win/loss ratio

W = 0.505 (bank's edge)
R = 1:1 (profit and loss are both equal for a 2x bet)

Kelly % = 0.505 - [(1 - 0.505) / 1]

= 0.505 - 0.495 = 0.01, or 1% of the bankroll.

Note that a full Kelly Criterion doesn't try to minimize volatility; rather it maximizes growth in the long term.

If you are a short-term investor J-D may not be the right investment for you.

You might be right - my expectations were for up days and down days but not this level of volatility.  What I wonder though is our bankroll large enough to support the Kelly Criteria or can we be bankrupt (or lose enough investors flee) before the Kelly Criteria can assert itself.

It's asserting itself right now...  Site is back to flat.  For the moment.

[mechs]

My initial analysis of celeste's betting, based on the total betting results generously provided by Dooglus and celeste here https://just-dice.com/whale.zip.

I took the bet sizes and randomly assigned the win/loss outcome assuming a 1% edge on 50/50 (I neglected the fact that some were 49.5 bets and others were 50, it would make almost no difference).  

After 20,000 shuffles, this is the outcome distribution, with celeste's actual outcome highlighted.

In short, it is lucky (top 7% of expected outcomes), but not truly exceptional nor significantly different than the null hypothesis (which is celeste's system is = to random ordering of betting).

 

Hey Bug,
   That is a great analysis.  You actually convinced me to stick with it for a while longer.  I think I'll stick with it until my investment is down 25% at which point I will need to reassess.

[shawshankinmate37927]

As of this moment, JD is profitable again....thanks celeste!

[Bugpowder]

My initial analysis of celeste's betting, based on the total betting results generously provided by Dooglus and celeste here https://just-dice.com/whale.zip.

I took the bet sizes and randomly assigned the win/loss outcome assuming a 1% edge on 50/50 (I neglected the fact that some were 49.5 bets and others were 50, it would make almost no difference).  

After 20,000 shuffles, this is the outcome distribution, with celeste's actual outcome highlighted.

In short, it is lucky (top 7% of expected outcomes), but not truly exceptional nor significantly different than the null hypothesis (which is celeste's system is = to random ordering of betting).

Hey Bug,
   That is a great analysis.  You actually convinced me to stick with it for a while longer.  I think I'll stick with it until my investment is down 25% at which point I will need to reassess.

In the last 20 minutes he has now shifted into the top 14% percentile, with a 1500 BTC loss...

Oh crap.  Now he just won 800 coins :-P

[willphase]

As of this moment, JD is profitable again....thanks celeste!

[stripykitteh]

You might be right - my expectations were for up days and down days but not this level of volatility.  What I wonder though is our bankroll large enough to support the Kelly Criteria or can we be bankrupt (or lose enough investors flee) before the Kelly Criteria can assert itself.

The whole point is that J-D *always* has a 1% max bet. If the roll drops 50%, the max bet drops 50%. If the roll doubles, the max bet doubles. The max bet is *always* tuned to optimal efficiency.

Whether investors come or go doesn't really make much difference. My own stake on J-D is about 20BTC. Even if I was the only investor, I'd have to lose 1/2 my bankroll 24 times in a row for it to be reduced to under 100 satoshis when I couldn't offer a 1% bet anymore (I don't know what the chances are but I think I'll die of something first so it's a non-issue).

[HurtK]

celeste is on monkey tilt right now, betting 111 or 222 BTC like every 5 seconds

[HurtK]

lol, he just won 2200 BTC in 11 straight wins in like 20 seconds. sick life

[stripykitteh]

lol, he just won 2200 BTC in 11 straight wins in like 20 seconds. sick life

I publicly dare him to go for 20 straight wins. Just like the A's.

[willphase]

lol, he just won 2200 BTC in 11 straight wins in like 20 seconds. sick life

2220 BTC won in 27 seconds:



Will

[filharvey]

lol, he just won 2200 BTC in 11 straight wins in like 20 seconds. sick life

2220 BTC won in 27 seconds:



Will

That is beyond luckly.

Phil

[Bugpowder]

lol, he just won 2200 BTC in 11 straight wins in like 20 seconds. sick life

2220 BTC won in 27 seconds:



Will

That is beyond luckly.

Phil

Put it in context though... He lost 10/11 IMMEDIATELY before that.

[shawshankinmate37927]

lol, he just won 2200 BTC in 11 straight wins in like 20 seconds. sick life

2220 BTC won in 27 seconds:



Will

That is beyond luckly.

Phil

Put it in context though... He lost 10/11 IMMEDIATELY before that.

I had my finger on the trigger and fortunately, I was able to divest 20 BTC at or near the end of celeste's losing streak while JD was up about 500 BTC.  It's now been reinvested.

[TheFuneral]

Thanks BugPowder for doing the work! Great to see that going on around here.

[romerun]

wadda