Topic: Provably Fair vs. Independent Auditing (Read 2939 times)

Sure, if adding a seed will help convince people we can't predict their actions no problem. I'll work on that soon.

That's pretty much become the point of this whole thread:

Provably fair is a show stopper for all other, non-Bitcoin gambling establishments. Coupled with the much lower friction Bitcoin allows, this is the end of any.other.casino.

As such, I wouldn't be surprised they'd try and pretend this isn't happening.

Provably fair casinos are not supposed to predict a player's actions. You can ensure the results are random by using a client seed.

No, I din't mean any joke. I don't think predictable means not random.

For example, husbands and wives often become accustomed to each other's behavior. A husband might "predict" that his wife shortly after returning from work will go to the bathroom. Now, if the husband waits for his prediction to come true, and it does, does that mean his wife is definitely not random? She is doomed to stay in that pattern? Or that the husband could never be wrong? I don't think so. Yet, her behavior can certainly become predictable. However, that doesn't mean it can't change for some reason.

This is a joke, right? "Predictable" necessarily implies "not random." I'm not going to respond to the rest of your quote either, because it's clear you've never taken a statistics course.

But why/how does the rock-paper-scissors AI win more than 1/3 of its games? It does it by studying what many many players do against it. That doesn't say to me people are not random as much as it says they often employ predictable strategies.

In other words, if the AI robot shows rock, rock, rock, three times in a row, and does this to a large enough sampling of people it's probably near certain it can conclude most people will not expect it to show rock a 4th time. That's simply because people expect things to be mixed up a bit, which makes sense. However, someone like me, knowing the AI was trying to build an analysis on me would put a higher probability on a 4th rock exactly because I would think it would think I wouldn't expect it.

It's kind of like that game Family Feud, where they go out and survey 100 people on a question like "name a reason someone might make fun of your car". The number one answer might be "dirty". That doesn't mean those people are not random, just that they think the same way in terms of that question (actually the number one answer here is "old", dirty was two).

That still doesn't mean (to me) you standing at my table don't have equal 50/50 chance to win. You don't have to go with your instinct, which might be 'left' because of the left-to-right reading thing I mentioned. You might get a 'gut' feeling to choose right, which might be correct. Who knows why. Some people call this psychic ability, some luck, some fate, whatever. The point is since you are able to think freely you have equal chance to get it correct (at least to me).

Again, we're talking about a hypothetically dishonest game that claims to be random but is actually employing some sort of statistical method (like the RPS AI uses) to win more than 50% of the games. Just as the RPS AI wins more than 1/3 of its games (putting the chance of a loss for a person standing at the AI's table >33%), and we say the person, since he or she doesn't make random decisions, has less than 33% chance to win, a left-or-right-card AI could win more than 50% of its games. Then a RNG would still have a 50% chance of winning, but a person would have less than 50% chance of winning.

Yeah, I only re-launched a few days ago and haven't done any advertising. But I don't see why it's necessary to have game stats (any amount) for you to tell me if you're standing at my table why you'd have a greater or less than 50% chance to win.

Oh geez, I didn't realize your volume was that small.

Alright, let's pause for a year or so.

Games: 1981
Choice 'left': 1041
Percent choice left: 52.55%

Do you think this means anything? If so, what?

Very well. Let's say this is a debate



Okay, so if you're standing there at my table and don't have a 50/50 chance, what percentage chance to win do you have?

Debate != argument. This is the former. In my high school we had a debate team that had as much fun as the sports teams.

No. It was the simplest way of asking the question. I think there will be a statistically significant bias, but I only asked it that way because other phrasings would take more letters and words.

Read up on "statistical significance" and "statistical power."

Yes. Yes I am. Because, like I said, I think there will be a bias towards one of the cards. If you would just post the stats (do you need help with bash or Python or R?) then we wouldn't need to be having this circular hypothetical discussion.


Whether your house edge is most conspicuous in the payout or the odds has no bearing on the existence of a house edge.

This is a strawman which assumes people only play the 2x multiplier on Just-Dice. Which doesn't even make sense, because your site doesn't have a 2x multiplier option.

Assuming Just-Dice and DoubleMyCoins have the same house edge, it is always possible, for any bet on DoubleMyCoins, to reconstruct the same bet scenario on Just-Dice. In fact, I even put the equivalent one in the quote you were responding to: a 1.98x bet (50% odds) on Just-Dice doesn't have the "no way you could have won" scenario. In fact, any win odds over 50% don't have it.

Well, yes I do know exactly the game I'm providing. Maybe not so much about what others do, though -- you're right I just checked and Just Dice does have a 1.98 setting. I'm not a big gambler so I didn't know they had that (their default is 2x), but yes that would result in a dice roll target of 49.9999 to 50, or basically 50%.


Did I run over your dog unknowingly? You seem to have some antagonistic grudge against me. Yet I'm pretty sure this thread has been our first ever contact.

Doesn't even know the game you're providing, gosh...

Since you really can't see how wrong you are, here it is: you are providing a bet with payout 1.98x, therefore if you go to any site with 1% and bet with payout 1.98x you get an exact 50/50 chance to win. Obviously, if you set the payout to 2x, you don't have an exact 50/50 chance to win, because of the edge... You don't give a 2x payout, so I can't even begin to comprehend how you can say such things.

nimda, it's kind of fun arguing with you

I have issues with most of the points you posted, but I only have time to respond to what I think are the most important, the last two:


Yes, every game is recorded. I could check but I don't need to: I'm going to go out on a limb and say you think there will be a bias toward picking the left card, correct? Why would there be a bias? Off the top of my head I would say because people in Western countries read from left to right, so that would make the left card slightly more favored (my unscientific theory).

If I were to run a check would I find that's true? I might, but that wouldn't prove anything. Not enough games have been played. It could be that given the same time into the future the bias shifts to 'right'. It would take a large enough time period to be relatively sure any bias really was there.

But all that doesn't get to the core issue -- which is walking up to a table and having a 50/50 chance at winning. I think the "online" aspect is your sticking point, so let me put it this way: Say I'm like those people on the street in New York that shuffle three cards and ask you to find the Queen. Of course, their sleight of hand makes that difficult. But imagine you walk up to MY table and there are only 2 cards there. I tell you one of them is a Red Ace and all you have to do is flip it over to win.

Now, you're honestly going to tell me you don't have a 50/50 chance of winning? Even if I have some personal theory about people reading left to right?


It's not the same. I believe my game favors the bettor more. I pay out 99% of the win (each win is less 1% house fee).

This is actually better than a game playing "edge", if you play only 1 game. If you play Just Dice, for example, which has a 1% edge, on 1 single roll for 10 bitcoins (about 10K) if that roll falls between 49.5 and 50.5 you lose. You would lose 10 BTC (10K) on 1 game no matter what action you take. Whether you would have chosen High or Low makes no difference. That tiny 1 percentage point in the middle of 1-100 can eat your entire bet, regardless what was chosen, no way you could have won.

That means you're most likely to only win or lose by action, but very likely to lose at least 1% of the time over many games. The average amount you should lose should be 1% of your total wins/losses. The problem is for playing just 1 time your bad luck may hit you all at once, because that 1% is built into each pick.

Losing regardless of your action is impossible with DoubleMyCoins. You are guaranteed to either win or lose by your action. No in between. If you win, you win pretty much the full 10 BTC (9.9 BTC or about $9,900). Think of it like paying insurance, you're paying insurance to be sure your action will result in only a win or loss. (same problem with Roulette -- 1 game could come up 00, green)

Been reading Dostoyevsky, have we?

The rock paper scissors AI is simpler in concept than your site's HTML. And, like I said, I bet it works, for some value of "works."

Nope. It would need to have a higher than expected profit ratio, which is not the same as winning more than 50% of games. In other words, it could win the large bets and lose the small bets with >50% and <50% frequency, respectively, and still have a win/loss ratio of 50%.

That doesn't matter though, because, again, I'm not making any allegations, but rather discussing the "provably fair" concept in abstract.

---

Actually, I can. Do you keep statistics or past bet data? If yes, can you quickly check what percentage of bets in the top 1% of value were placed on the left card? If no, can you start?
Thanks.

But you don't pay out 2x the bet, right? A 2x bet on 49.5% odds has the same house edge as a 1.98x bet on 50% odds. That's a bit offtopic though.

Thanks a lot marvolous

Provably fair is a great topic. People like myself try to build entertaining games and add value to the community. Math and secure hashing algorithms provide a way for us to prove we're not cheating with our games. People don't seem to realize, we don't need to! Casinos are a great business because math is on the house's side. All it takes is to stay in business. That's why I also want to give a shout out to dooglus for coming up with the Invest option. I think that shows true innovation in the space.

One more note about DoubleMyCoins is people seem to forget you can also play 1 game. The real idea behind my site is for people to walk up to the table, plonk down one massive bet, and sweat it out before making a choice. That's really the experience I tried to create. No one can say that's not a 50/50 chance, which is provably fair.

If you do that in Las Vegas at the Roulette wheel, betting on Red or Black, it's not 50/50 because 00 (green) can still beat you. My game is exactly 50/50.

Excellent replies DoubleMyCoins, Nimda and Light.  I am just trying to fully understand all aspects of provably fair vs. traditional online casinos which I enjoy.  Don't even let individuals that don't contribute constructively to the conversation waste your time.  Perhaps it would be a good idea to have some of the provably fair systems audited to see precisely what their ratios, edges and percentages are. 

moderate, you've seemed to have a chip on your shoulder from the beginning of your replies, although I've only responded civilly and done nothing to you. I guess that's part of being in the Casino business, dealing with people probably angry for being on a losing streak or who knows what. That's okay, I still enjoy it.

Hate to tell you, but being retard doesn't make your game provably fair.

That's irrelevant.


Your words are irrelevant, if you don't know why try thinking on it some more and maybe you'll see.

See, two can play the game of talking while saying nothing.

Irrelevant.


Irrelevant, too. Maybe this one will take a bit longer for you to understand why it is irrelevant.


The thing is, your friend dooglus pointed you to something that is flawed and you cannot see by yourself why is that. It has been pointed already, but you can't understand it.

It can? This is something you know? Or something you're just thinking, because I can tell you my site can't. Further, I'd be interested in you telling me any computer on earth that can predict what I'm going to do accurately. If that was possible why should anyone live life? They could just go to this super smart computer and have it tell them what actions they would take before they need to -- live their life for them in other words.

Even if you could convince me that was possible you think I could build something like that myself? Further, that I would make such a fancy contraption to try increasing the house edge? By how much? At some point the complexity of trying to make such a thing isn't worth it for how much extra money would be gained, especially if users proved such existed and tarnished the site's reputation. It's simply not worth it.


Sometime even things that seem random are not. Supposedly, when you go to Las Vegas and a dealer spins the roulette ball he can't control where it falls. It's supposed to be random. But are you really sure no amount of his skill can influence it? I'm not. Anyway, I've set my site up to be random, to what I believe is actually random. I even play myself. Anyone is welcome to try and discover a pattern or analysis algorithm and if they do exploit it.

EDIT: and one more thing, you can see here that the win/loss ratio is almost exactly 50%, just as would be expected. If my site was influencing winning, wouldn't it need to win more than 50%? http://doublemycoins.com/games

No, I can't, and your destruction of that strawman proves nothing.

What percentage of your users, would you guess, use coin flips or other reliable sources of random data to base their decisions on? (I have a guess: "low.")

Like I said, though, the point is the site can predict what the user will do with enough accuracy to effectively increase the house edge. This would be fine, and even fun, but not if the game were doing it under the pretense of being random. (I'm not making any allegations here; it's a hypothetical.)


I'll bite:

Then the site could use a stop-loss and go back to being random. With enough laypersons playing, being dishonest would still be profitable.

Man this thing is pretty good fun, finally got there on my second try. I'm not exactly sure what the whole purpose of the exercise proves, just because thought patterns can be "predicted" over a set range doesn't mean that statistical probability doesn't approximate the expected for extremely large cases.

I've seen and played that before

I can't remember my score, but I believe I was successful making the computer "think" I was going to do something I wasn't, allowing me to win -- in other words behave irrationally on purpose.

But that's my point, and why like I said this topic interest me, why I built the game in the first place. It's actually very hard to know when something is truly random. Often a pattern can function well enough to be random, even though it's a pattern. This is known as pseudo-random, and is how computers usually generate random numbers (computers are not random by nature).

Like I said, though, the point is the site can never control what the user will do. Let's say my site tried your statistical analysis (which it doesn't). If anyone discovered the tendency of the algorithm they could use that to their advantage, and win reliably which would be costly. Second, according to that link humans are not very random. Let's say that's true (and I think it mostly is). Find a nickel and flip it. If it's heads choose right, tails choose left. Flip that coin over and over to make your choice if you don't trust yourself. You can't tell me my computer could analyze and accurately predict the result of your random coin flip...  

The day computers can accurately predict the result of random coin flips we should all just pack up and go home.

Some of the BTC online Blackjack sites actually hit on soft 17 which is not usually the case with real world casino. It adds about 0.2% house edge which is quite large compared to the standard house edge of ~0.4%. So I'm not too sure that there are that many sites with better odds.

I'm not forgetting anything.

Do me a favor and beat this in 30 rounds:

http://www.nytimes.com/interactive/science/rock-paper-scissors.html

That's true, but setting a pattern would be a disadvantage. Any user that discovered the pattern could then accurately predict the answer and break the bank. I've done a lot of thinking on the topic of random numbers, Martingale systems, roulette, etc., because the subject interests me

The fact is we show the player exactly what the answer is at the bottom of the screen before they choose. How we arrive at that answer is irrelevant, because we can't control what the player will do. If they pick the correct side they win and we can't change that without providing a different game hash to what we showed.



Yes, you're correct that we could try to pre-guess what the player will do, but you're forgetting any player can imagine we're doing that, just as you're imagining. They can then know with higher probability what the computer will select and alter their choice to win -- in other words "fake out" any prediction algorithm, choose "left" four times in a row to get the computer to select "right" as you say.

As I describe above, there are only 2 choices, and one will win, which is shown. The player always has a 50/50 chance to choose correctly. There is nothing we can do to stop that.

That happens to be exactly what moderate is referring to. With some statistical analysis, you can make educated guesses as to which card the player will pick next. Then you generate a hash of the opposite one.

1. User picks "left" four times in a row
2. Server creates string "right:54fb18c1a..."
3. Server hashes the string ("0dd38ef...") and shows the hash to the player
4. The user loses, and checks the hash

With a client seed, this becomes more difficult. With some nice client-side Javascript which suggests a random client seed, this becomes much more difficult.

Yes, you missed the point entirely. I'm not interested in what someone else told you, the fact is that with the current scheme you are free to set whatever pattern you want as the user has no input into this.

No, not for how our game works. You're thinking about games like Dice sites or Lottery sites where the winning answer is generated after a player chooses. For example, if you choose a "high roll" on Just Dice you will take that action before their server knows the answer. You click "high" and the server then makes a calculation based on the seed etc. to see if you win.

In our game the winning solution is known by the server before a user picks. So we actually show them the answer at the bottom of the screen. Here is an example of an answer:

left:18845f1e40ab0a8a2c0568f51e44b4aa834b6105e1b3d83f152172fcc35ae57d

We take that message and add a secret key to it to make a game hash which we show the user:

15063ea21c5a75d013008877740faa96ecc16b6f0b2c850e6e54104a2bdd4348

Now, after the user picks we show them the message above (showing left in this case) and the secret key used to make that shown hash, which is 530090401876695576969539 in this case. You can check these values with our Online Check tool.

With SHA-256 changing even one character of a message or secret key results in a completely different hash. So there is no way we could make the message say "right" in any way and get the same game hash we showed the user. Dooglus actually gave me the idea for streamlining my provably fair system to what it is now from when I first started out.

There is also another problem, where people that think they're providing something provably fair are not in fact doing so. Let me give an example with DoubleMyCoins, they do the following:



The issue with that is that there is no user input involved, you're free to always generate a key that results in the right card being the winning one if you check that the new players always pick the left card. The user must always contribute to the process, and it must be done after the server picked whatever seeds it wanted to pick.

marvolous: it's possible for some casinos to market themselves as "provably fair" but actually not be so.

Part of the problem is a misunderstanding by players of what provably fair actually means. In the Bitcoin community provably fair is supposed to mean players can't be cheated by a crafty house, at least not without being potentially found out and exposed.

Let me give an example with DoubleMyCoins, which has an easy to verify Provably Fair system. We have a 50/50 game, where a player picks the left or right card. One card wins, one card loses, simple. Now at the bottom of the screen before a player picks we actually show the winning answer - but it's in SHA-256 hashed form, so the player can't read it.

After they pick we show the revealed winning message and secret key used to make the hash shown previously. By using these values anyone can use any SHA-256 script to verify the values equal the hash, so it's known we couldn't change anything (we wouldn't get that hash).

That's the way math and hash algorithms work. Now, one way we could attempt to cheat is realize users won't always pay attention to the hash we show. We might then swap in a similar looking hash to make the player lose when we want. As long as players don't notice the switch we could successfully cheat them.

To prevent that, however, we also record every game with hash information automatically for the player and give them an online check tool which calculates any SHA-256 hash. We make it super easy for any player, even if they don't program, or pay attention to hashes while playing, to verify any of their games at any time. That means it would be very risky for us to try and swap in a fake hash, because we've empowered players to check games easily. The damage to our reputation wouldn't be worth it to risk swapping a few games, especially since we have a 1% house edge already. I hope that make sense for truly Provably Fair  

Completely unrelated comparison. Just because you paid someone to make a shiny crap unprovably fair casino for you, it doesn't make it better than the other ones.

Provably fair is not a marketing tactic. You either provide it, or you don't and scam your players.

Even so if it is Provably fair, still chances of winning is more likely less then the chance of being stuck by a lightning.

A property of cryptographic hashes is: "it is infeasible to generate a message that has a given hash". If results are generated properly using a server seed and client seed, neither the house nor the player can find the result before both are revealed.

The article you linked to shows how the provably fair system could be circumvented by the house. Games that are truly provably fair are (almost) impossible to manipulate.

A truly provably fair casino is always better than one that is independently audited. Individual results from an RNG could be manipulated, while provably fair results cannot be feasibly modified.

Read this if you're looking for a detailed explanation of provably fair.

Very interesting response, but that is to be expected.  We actually understand provably fair very well and understand how it can be manipulated.  We are not sure what bitwincasino uses for its casino software however we also play on traditional online casinos that license software from larger companies which have their software audited.  Are you sure you are not the one being scammed?   

Dear lord, thorough research is not reading paid articles or copy & paste from reddit and other dumb sources. Come back when you actually understand the concept, till there feel free to get scammed by bitwincasino and others providing what you are after.

I am an avid gambler and have done thorough research on "provably fair" casinos and it seems that every article I read affirming it there are just as many articles describing how provably fair isn't necessary provably fair.  Here is one of the articles:  http://coinbits.com/2013/06/06/provably-fair-by-bitzino-not-provable-with-proof-of-concepts/  

Many of the traditional casinos I play on have independent auditing services frequently reviewing and certifying their games.  Also, these websites license their games from larger companies and thus do not even have the ability to manipulate the games, whereas the "provably fair" casinos have full control over the game.  I am not saying that any of the "provable fair" casinos are cheating anybody, but is the bitcoin community over relying on the provably fair concept?  After reading the above article it is easy to see that the provably fair casinos can be manipulated, yet those with games licensed from larger, reputable companies can't be manipulated in anyway by the Licensor.  Just my thoughts.

Either way, I perform just as well with the traditional online casinos and find my winnings / performance to be the same and experience better.  Therefore, I guess if it makes people feel better that is a good thing.  Or is it?