Our latest entrant to the >100 BTC profit club:

So as johny1976 is refunding, what I'm going to do is hold onto the escrowed funds until johny1976 completes his full refund. After he tells me that it's done, I'll post here -- and people will have 24 hours to contact me if they believe they were not properly refunded. If there's no objections after that point, I'll be releasing the full escrowed funds directly to johny1976.

What does this mean?

Your wording makes it sound like that is in the past, and you are no longer operating at a 3.33x kelly. However:



You realize that is now higher than a 3.33x kelly, right?  Because if the max win is 1% of the bankroll, it means that investors are risking more than 1% (they need to pay 70% of the house edge to MP and the App, win or lose). However their expected value is only 0.3%.  So that establishes a lower bound of a 3.33x kelly (but it's actually a little higher)

Yeah, because of some abuse (and the abusers watching the open source changes to continue to abuse) I made it closed source, but you can still access the old version. Things like the bonuses and core mechanics of the game are unmodified. 

And making the later versions closed source has been a decent revenue stream, I've sold a whole bunch of licenses to sites (especially csgo ones) for 2 BTC a piece =)

Perhaps it makes more sense if you start with the high level:


Originally in bustabit everyone had to bet the same amount of money. When ever the the game busted, 1% of the sum of peoples wagers (the bonus pool) was given out to who ever cashed out last. If multiple people cashed out at the same point, it was divided equally amongst those participants. (and no one cashing out is the same as everyone cashing out at the same point, i.e. everyone gets 1% back).


However, people didn't like to wager the same amount of money and it was introduced the idea of allowing people to vary their wagers. So giving the last person to cashout the entire bonus pot would be pretty unfair. (someone who bets 1 bit, shouldn't get the same bonus as someone who bets 1 bitcoin). So that's how the whole "maxBonusRatio" thing came into play. It figures out the most the largest better could make (assuming he got the entire bonus pot) and caps everyone to the same ratio.  The code simply iterates through all the players and allocates bonuses accordingly, making sure no one can get an unfair amount.

Now the game is just as far for someone who bets 1 bit and 1 bitcoin =)

If you're familiar with javascript, you can see exactly the code:

https://github.com/kungfuant/bustabit-gameserver/blob/master/server/game.js#L543-L627


Basically, after a game ends:

Step 1: https://github.com/kungfuant/bustabit-gameserver/blob/master/server/game.js#L573

Sort everyone in order of their cash out. People who cashed out last are at the start. People who haven't cashed out are at the very end. If two people cash out at the same point, it doesn't matter the ordering.


Step 2: https://github.com/kungfuant/bustabit-gameserver/blob/master/server/game.js#L573

let "bonusPool" bet 1% of what everyone bet


Step 3: https://github.com/kungfuant/bustabit-gameserver/blob/master/server/game.js#L579

let "largestBet"  be the the most anyone bet


Step 4: https://github.com/kungfuant/bustabit-gameserver/blob/master/server/game.js#L595

let "maxWinRatio" be bonusPool / largestBet

Step 5: https://github.com/kungfuant/bustabit-gameserver/blob/master/server/game.js#L597

Iterate through the sorted list (from step 1) grouping by everyone who cashed out at the same point.

Attempt to give out the entire bonus pool to those players based on how much they wagered. However,
never give out more than "maxWinRatio" of their bet (to avoid it being unfair to large players) and never give
out more than the bonus pool.

What ever is given out during that iteration is deducted from "bonusPool" and if more money is still available
continue the iteration.




I'm not sure if that makes it clearer, but I hope so.

The bonus pool is *always* given out on every single game (where the bust isn't 0x). If there's money left over after giving it to people who have cashed out, it is just equally divided amongst the remaining players based on their wager




Yeah, if two people cash out at the same multiplier they are given the same amount of bonuses.


They would each be given exactly 1% of what they wagered. Which means B and C would get 1 bit, and A would get 1.01 bits.




The amount of bonus they will get as a % of their wager

No idea, really. I've actually never chatted with him yet =)

But most gamblers like to use the same style of gambling no matter what site they're on. Did he martingale his way up on PocketDice?

A big welcome to the latest member of the exclusive (and often ephemeral) >100 BTC net profit club:

20% of the house edge represented a 40% reduction in investor expected profits, while also increasing investors absolute risks (by forcing them to pay that on win or loss). Which is totally fine, it's perfectly reasonable to change the commissions structure. However to avoid throwing the whole thing out of wack, investors would need to have been subjected to a 2/3rd reduction in risk compared to what they previously had. Right now, if a whale makes a big bet you are knowingly accepting it despite knowing that it represents negative growth to investors (but you and the app gets a 70% commission).

I'm not sure how that's fair or reasonable to investors, but then again they're adults who can make their own choices and risks  

That's not really an answerable question, because it's irrelevant what someone is betting. What matters is what they are attempting to win.

I'd be interested in though seeing how many bitcoins have been wagered attempting to win > 1x kelly, and > 2x kelly

I'd really like to clear this up, when I designed MoneyPot I structured the risks very differently. The way it is set up now I think is negligent at best and ~8 months ago I had warned them that it would eventually lead them to their investors losing them money. I would've thought the recent events would make them re-evaluate their position, but they still appear dangerously dismissive of the risks.


I worked extremely hard in order to get the maths correct. I think it took me about 3-4 solid days to get the maths right, and I couldn't have done it without the help of blockage  (who is not very active here, but one of the brilliant minds in bitcoin gambling).

As far as I know, MP was the first casino ever that programmatically solved the generalized kelly criterion (no small feat, I must say ) to ensure it never subjected investors to a risk of greater than 1x kelly, designed to ensure the investors would always have optimal expected bankroll growth. Under the system I designed, I believe investors made ~400 (?) bitcoin with the early investors seeing a ~40% (?) return on investment in less than a year.

I'm not sure when or why, but sometime between buying the site and April they changed the logic such that the guarantees I offered no longer applied. I'd encourage you to check out the thread when I found out and did my best to try educate them: https://bitcointalk.org/index.php?topic=1438838.msg14577586#msg14577586  and at least got them to update the site to remove the mentions of system I had designed.

It's a shame that instead of MP admitting "Oh, maybe Ryan that mad dog has a point." they just go on to continue to downplay the risks:


..which isn't even remotely accurate. (For dice games) every bet that a gambler does that tries to win >66.66% of the max profit, investors should expect negative bankroll growth. This isn't some crazy anomaly, it's the common case when a whale is betting! The maths doesn't care about whales having "infinite bankroll" or anything like that, that's silly.

I'm rather puzzled as I can't tell if they still don't understand bankroll management, or it's just a case of an huge incentive mismatch here (because MP and apps) make a constant commission on each bet it is in their financial interests to accept bets even if they're ludicrous for investors.


Since the concept of "bankroll growth" is foreign to most people, I'll give a simple example. Imagine I have a fair coin that I'm going to let you play with, and I am going to give you 2.5x for winning (instead of the fair 2x).  So given this fantastic deal (a 50% chance of 2.5x'ing your money) how are you going to play to make the most money?

If you were naive, you'd care about your per-bet "expected value" and bet 100% of your money on my coin. Now that is fine, but pretty quickly you'll find you'll bet wrong and lose everything. So even though you're maximizing your expected value in each bet, it's a stupid stupid strategy because your expected outcome is to lose all your money.

The same logic exists even if you don't bet 100% of your bankroll, for instance betting 90% of your bankroll would also be risking way too much. So what the kelly defines is the optimal amount you should be in this circumstance. If you bet more than that, you'll make money slower, and if you bet too much more than that you'll expect to lose money, not make money.

The way MP is configured now, is that it accepts bets that it knows are objective a "bad investment" for investors (although is good for them, and good for apps).



Anyway, I don't post here to be a dick. But I don't want people to think I designed such a system or it's something I implicitly am endorsing.

Cool! BTW just to be clear one of the requirements of using the open source one is that all changes you make must be made available and prominently linked from your site (see: https://www.gnu.org/licenses/agpl-3.0.en.html) 

I should probably let Dooglus speak for himself, but you perhaps misread what he wrote. He didn't say that investors make 15% of the EV, he's saying they have made of the EV. I don't see any error or misguidance in his statements.

Perhaps a better response would've been "Hey, our investors made 50% of what they should've. Yours only 30%"  or something. ( I <3 twisting statistics)


But anyway, I think perhaps the bigger concern is most people don't really understand the difference between "Expected Value" and "Expected Bankroll Growth" and just assume that if they're +EV they should expect the bankroll to grow, but the reality is if the bankroll is over leveraged that's not the case.

I'm not sure when MoneyPot switched from the kelly system, but some time after it did I warned of the risks and provided a simulator that showed how investors should expect to lose money. That was posted on April 18, where I recorded the investor profit at 101 BTC, and it's currently 62 BTC and have had substantial dilution in the mean time.

As much as I love to say I was right, it could also be attributed to other factors. But it's probably worth entertaining the possibility that running a bankroll at -BRG is a horrible idea.



That said, I have no problems with letting people do what ever they want (we're all consenting adults) but I suspect if the wording was a bit more clear: "By investing here you get 30% of the EV, but should expect to lose money over the long term if whales play on the site" it might result in less investment. Or maybe not, maybe the investors like to also passively gamble.  

It can be quite difficult to run a business when your income is so volatile, so the service that investors provide is they eat that variance in exchange for some of the expected value. It's pretty neat actually, as a site owner (but not of a MP app) I'd happily trade 50% of my EV for a steady income And presumedly investors are happy with it, by the fact they're still invested.

And if app owners don't mind some of that variance, they're always able to invest themselves.

all 

Yeah, Shiba's database has everything. He just listens to the connection and logs everything he sees and saved it in his own database. If you want play data, this is the way I'd prefer you get it. However, if you really need you can get a JSON dump of a particular game:

e.g. https://www.bustabit.com/game/3260196.json

Just take it easy, try keep it under say 1 request every 5 seconds to avoid getting anti-ddos banned

Thanks, glad you enjoy it =)

Yeah, that's probably the best way to do it.


Back testing with "what bonus you would have gotten" using historical actual user bet data would be pretty cool, but I also would worry it can give pretty misleading results. As you'd expect the very act of playing would change the behavior of other players (especially those also bonus hunting) in a way that's unknowable.

But anyway, I still think it's pretty cool the simulator. It can help debug a script and generate a neat chart of how your profit chart looks (e.g. martingale gives very different looking patterns than flat betting)