Topic: Provably Fair vs. Independent Auditing (Read 2934 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.