Throwing the dice too hard.

Quote: Buzzard

Ahigh may be on to sometimes. I saw a guy roll back to back 6-1, There was proof enough for me.

Which casino and what color were the dice? Hurry, I'm on my way.

Quote: Keyser

Ahigh,

Your data set isn't significant in size. You merely simulated a larger sample from only 375 or so tosses. You guys seriously need to stop posting the combined totals of both dice. It serves no purpose other than to curve fit your samples.

One moment, and I'll post a real sample

Here's a look at your sample's real values.

Total Ratio St. dev.
130 5.76 0.51
125 5.99 0.02
108 6.94 -1.65
125 5.99 0.02
119 6.29 -0.57
142 5.27 1.68

Chance of randomness is roughly only 1 in 2.4

Now here's what the bets look like for the theoretical if we let the computer recombine the dice.

edge on
casino
on lay
odds bet


4__2.28%
5__1.92%
6__2.51%

8__0.72%
9__1.31%
10_-1.97%

Now here's the field bet.
-2.6%

I don't know where you get -2.6% from those face samples, but here's what I get with just straight math instead of simulation: 1.8758% edge to the player's favor



Also, I'm curious if you are saying those player edges on the lay consider vig up front as is the case in every casino I know of besides Fiesta Henderson.

Using the same spreadsheet, here are the field advantages per math on the field for various face weights of theoretically biased dice:

7,6,6,6,6,7: 1.869806094%
8,7,7,7,7,8: 1.239669421%
9,8,8,8,8,9: 0.7600000%
10,9,9,9,9,10: 0.382653061%
11,10,10,10,10,11: 0.078043704%
12,11,11,11,11,12: -0.173010381%

In a nutshell, there will be exposure to the house in the field for any heavy 6-1 face dice as soon as the ratio exceed 12/11ths .. IE 10% or more on the six one, and it's no fun for the house in the field.

As far as "combining" the dice, the samples are being made on the stick, not on the dice. These are 749 valid samples from the stick, not the dice themselves. When the assumption is the the entire stick is fair, this is 100% valid way to look at the fair dice as they are all identical in theory! Any significant variance, no matter what the cause, is what is being tested. Not a specific cube. The house doesn't get to force any cube on you, and that's the game. The details aren't anything that can be controlled and are not part of the experiment. Just the two outcomes per roll are the two samples per roll. I have no idea what else you could possibly say about the fact that there are two samples per roll, but there ya go: that's my side. There are 749 valid samples no matter what you have to say about it! And if that's not significant in your opinion, that is just your opinion!

I will agree that it would be much more significant if I had twice as many samples, but that's where I say you go and do the work yourself!

I'm still not grinding that field even for $5 myself because it's damn boring! That's not how I play and probably never will be.

All of this is just work following up on Harley's theories. This is not showing how I play the game or saying that I am out there exploiting the field.

It is just work, plain and simple. Does it prove anything? No! Is it evidence that supports Harley's theories? Yes. Is it significant? I think it is significant, even though it doesn't prove anything. Just having heavier six one faces is a 1 in 15 chance right there along from 6 choose 2. So it's not just the fact that the p-value is 0.40 .. but let's just say 15 times less likely from the 6 choose 2.

Alrighty then! Significant to moi! S'il vous plaît

WinCraps calculates the edge and standard deviation for every bet
once you enter the values for each die if different from a fair die.
(Plug: The program is for Windows and is free to have and use. It is shareware.
To register so you can run simulations is less than $20)

Probability Tab tab

then the Advantage tab
I get the same 1.88% edge for the Field


For the Lay bets I had the 5% vig paid on wins only

Quote: SanchoPanza

Which casino and what color were the dice? Hurry, I'm on my way.

I ain't telling. Building my team to take advantage. Just as soon as Ace gets back from AC.

I said this up front that I think the whole idea of making an edge on what should be free bets and increasing the edge on the passline is total bull crap, and that's why I went and modeled these theoretically bad dice to say, "if this were true there would be a player advantage in the field."

The whole idea was hilarious until I went out and saw these samples matching what Harley was saying!

I am not going to even say Harley doesn't sound like a nut job, because I think he knows he comes across that way. He gets it all the time about how he's a conspiracy theory guy. He's used to it in fact!

But when I meet with the guy privately in his room (a very nice room at that) and then he goes on to tell me stuff that sounds ridiculous, and then I go out from what he said and I start collecting samples, and it matches up, you better believe I'm sharing what I found.

All of you guys who say this is a bunch of bull crap should go and collect your own samples at the Fiesta or Gold Coast or any other casinos that Harley suggests are using suspicious dice on weekends and come back and show us the samples saying that the dice are fair. It's hard work, and we would all appreciate it. Even me!

Just stomping your feet and making fun of Harley is really easy to do.

But this guy, no matter what you say about him, knows his dice better than anyone out there.

Period.

I know all 6 numbers that are on a die. Any knowledge beyond that is simply superfluous.

Quote: Buzzard

I know all 6 numbers that are on a die. Any knowledge beyond that is simply superfluous.

Like a fortune cookie .. IN BED.

Quote: Buzzard

I know all 6 numbers that are on a die. Any knowledge beyond that is simply superfluous.

Nonsense. You need to know that a) The die produces a random result and b) that each side is equally likely to show up.

Both may be trivial facts, but without those axioms, you can't tell anything, and are not superfluous.

----

I would assume (a) and (b) until proven otherwise.

I disagree with Ahigh that the data presented is even a matter of 'opinion' for proof.

Quote: thecesspit

Nonsense. You need to know that a) The die produces a random result and b) that each side is equally likely to show up.

Both may be trivial facts, but without those axioms, you can't tell anything, and are not superfluous.

----

I would assume (a) and (b) until proven otherwise.

I disagree with Ahigh that the data presented is even a matter of 'opinion' for proof.

I say significant and you say proof. Potay-toe Potah-toe let's call the whole thing off!

Quote: Ahigh

I say significant and you say proof. Potay-toe Potah-toe let's call the whole thing off!

But with what degree of significance?
You need to tell that to the judge.
The only data that failed at the 5% significance level was Harleys 30 count sample
(10 times smaller than the recommended size to use with the chi-squared test)

But the test only returns a yes/no for the null hypothesis.

the p-value again just tells us if we can accept the null hypothesis,
in the first case below I compared the face counts to a 9,8,8,8,8,9 die (18% and 16%)
The NH (H_o) being the observed data [1] 130 125 108 125 119 142
could from a die with a bias of 9,8,8,8,8,9

I use a significance level of 1% because I know the probabilities of the die. It is not unknown.
So to reject the H_o I want the p-value to be less than 1%.
The alternative hypothesis would be that the observed data could not have come from a 9,8,8,8,8,9 die
R code
any one can use this code online here
http://aleph.sagemath.org/

freq = c(130,125,108,125,119,142)
probs = c(9,8,8,8,8,9)/50
chisq.test(freq,p=probs)

Chi-squared test for given probabilities

data: freq
X-squared = 2.1747, df = 5, p-value = 0.8245
> freq
[1] 130 125 108 125 119 142

So the p-value is .8245 meaning that if we took many samples of 749 freqs we would find
82% of them to have a HIGHER critical value number (the difference squared/expected summed)
so the observed distributions can be even worse or more wonky than the one observed

***I accept the H_o that the observed freqs could have come from a 9,8,8,8,8,9 die.

Now, would this still hold with a larger sample size??
Who knows unless we have a larger sample size.
Any reason to look into this further?
I say no because we are dealing with a multinomial distribution.
Lots and lots of possible results.

Here are a few critical values (X-squared or x²) and the p-value for each
example: (5,5) = critical value = 5
and the degrees of freedom = 5

> 1-pchisq(5,5)
[1] 0.4158802
> 1-pchisq(7.5,5)
[1] 0.1860298
> 1-pchisq(10,5)
[1] 0.07523525

So even at a 10 critical value, over 7.5% of more distributions would have even a higher critical value

Compared to a fair die
freq = c(130,125,108,125,119,142)
probs = c(1,1,1,1,1,1)/6
chisq.test(freq,p=probs)

Chi-squared test for given probabilities

data: freq
X-squared = 5.1175, df = 5, p-value = 0.4017

> freq
[1] 130 125 108 125 119 142
> mean (freq)
[1] 124.8333
>
***I accept the H_o that the observed freqs came from a fair die.

Looks like the sample could have come from either die, the fair or biased. OK.

The p-values do not say that one die is more likely to return the observed data than another.
We could compare two different observed data sets to each other and
have the null hypothesis be
could both samples come from the same population or distribution? (the same die)
But we need different samples, not the same one.
That is a different test (Chi-squared tests for homogeneity)

Consistently low p-values over many samples may throw a red flag, it still means we need to look further.

But if you are at a Craps table and you see more 1s and 6s than expectation by your tracking
Go for it.
Bet the Field, Hop the 6,1
It adds a different flavor to the game.

At any time that trend may go to other numbers.
Now our chi-squared value changes, did it change too much??
Where is my binomial table instead?

But if the dice are really such poor quality and biased to be seen in 200 counts or less
and will keep a strong bias as they continue to wear during a shift of play,
I say go for it.
Bet the Field, Hop the 6,1
I say Hi, you say Lo

Have fun on Monday.
Try the Hard Throw test too.
More (better) random results from the harder you throw the dice.

The word "significant" is a specific word relative to testing a null hypothesis. So just to be more succinct, I was not saying the results were significant in that context.

So let me state what I was trying to say using other words to be more clear.

A year ago, I wrote Harley off as a total nut job. I watched Koga's videos and I saw a guy who didn't know how to use a balance. I read Harley's blog and I connected these two guys together and I thought, "oh my freaking god how in the world could they expect anyone to believe any of this bull crap!"

So then I made the die balance to put this whole issue to bed. Then I made a video effectively saying "these guys are bonkers! Look at this: a die looks bad in the poor balance and it looks fine in my balance! Case is closed!"

And it was for a long time.

Then fast forward to the Ahigh show and all the Koga Ninja videos and now more claimed proof that biased dice exist and popped up on my show!

So reluctantly (VERY RELUCTANTLY) I reopened the case and I gave Harley a fair re-trial.

I met with him and I listened to him and I let him tell me what kind of observations he had.

While he did all of this, I still had (and even HAVE) my skeptics hat on. And I still think he sounds like a total conspiracy theorist. And I absolutely believe he is holding a false belief or two up there in his conspiracy theorist's mind. And I tell him this to his face, and he understands and respects that.

But when I use "significant" .. let me reword that: the evidence that Harley has provided to me and that I have collected on my own means enough to me personally for this issue NOT to be put to bed any longer.

The case is back open, and I will continue to dig up evidence. What would be great would be if other people would HELP to dig up evidence and keep an open mind about distributions coming from dice that fit known theoretical profiles of distributions that could favor the hold on the table, especially during the busiest times when people bet with more common and/or expected bets.

And here is an assertion that I will make: the game itself is designed to be a fair game. There is a don't side, and if the game were played equally by as many don't betters as do betters, this would be less of a concern.

But the fact is that fewer people play the don'ts. And therein lies a big opportunity for the house to be selected about what dice have statistically had better holds during certain times and opportunity to re-introduce dice that have had favorable holds in the past.

I personally like to feel razor sharp corners on a fresh stick of dice any time I show up to a table. I like being there at the beginning of a shift with an extra security of knowing these are brand new dice!

So all I'm saying is that this can of worms is back open. I'm not saying anything has been proven. And I'll be honest about this: I think that the fact that this can of worms is open makes the game more entertaining even if it is the kind of thing where more suckers (including suckers making free odds bets) are getting taken out there without knowing everything that can be known about the game.

Now I know it might rub some people the wrong way having their belief systems challenged who aren't used to being wrong about things like the cost of a bet that should be free.

But this is what I am saying: pointing out fallacies in those who are not used to it is a lot more fun than making fun of the easy targets.

And the truth is that there is this possibility: the possibility that brand new sticks of dice have enough of a bias to have an edge on what should be a free bet. A possibility that milks the common "knowledge" of maximizing odds to be more likely to win could in fact be making intelligent readers of this forum suckers for the casino from making free odds bets thinking it's not costing them anything in the real of millions of throws.

And they would be right, if only they knew with certainty that the dice were 100% fair out there.

Here's the truth: the dice are not 100% fair. We can argue 99.99999% fair versus 97.00% fair, but nobody can tell me the dice are 100% fair because we should all know that there is either a player edge or a casino edge on a free bet depending on the bias of the dice. And every single die ever manufactured has a bias, and that is where the edge is going to go whether it goes for the player or for the house depends on the dice not on the math. And so if you want to say all the dice out there are not set up to take money from what should be free bets, that is what I would like to believe too.

But as long as the holds are being profiled for sticks of dice that are out in play and then they are re-introduced, there is a chance that players are doing battle without the sufficient accounting that the casino is doing. They are just making "free bets" and expecting everything to be fair.

The casino is doing their homework looking at holds and writing down serial numbers and the players are not.

So we will see, I guess. But in the short term, I think that all players' guards should be going up on any free bets. And if you want to keep it easy, absolutely switch to the dark side until the cloud passes. But I wouldn't be going to a 100x odds place and going for it myself without looking a little further into this issue myself. Not if it were my money.

Add to this two additional facts: The LVH and the Cosmopolitan both switched from paying triple in the field to paying double in the field in 2012. Sure, it's very possible this is a coincidence. But if more Chinese dice are being made with bias that negatively affects the field (even INFREQUENTLY) ... why does a 2.78% edge PER ROLL BET .. REALLY need to be 5.56% all of a sudden? Go to these places and compare the number of field bets being made to a place that has 2.78% field bet. The number of bets on the field gets cut MORE THAN IN HALF when they do this! They aren't making more money in the field from my view; they could possibly just be looking to lose less to the lucky people who inadvertently got luckier than they should have SOMEHOW.

" why does a 2.78% edge PER ROLL BET .. REALLY need to be 5.56% all of a sudden?"

gee, why have 0 and 00 in roulette ? DUH !