DaveW - New Member

If there is a better thread to post this or the answer is found elsewhere, please let me know. Thanks!

Win Rate: Clark County=10.24 Strip=10.26 Boulder 9.91 Downtown= 5.77

Bureaucrats: If played at the full sized table the commission refers to it as Baccarat whereas if played at a Midi or Mini table, its referred to simply as MiniBacc.

Advantage Play: Only if you can memorize eight decks or have paid the dealer to do a false shuffle is there such a thing as real Advantage play. Others milk Baccarat for freebies and enjoy a slow pace.

How does "Win Rate: Clark County=10.24 Strip=10.26 Boulder 9.91 Downtown= 5.77" apply to a sample size? or SD percentage? or even remotely give me a bell curve to guess?

Maybe I should have posted this in the math section....

My second question pertains to the rate of expected return to "normal" as it pertains to sample size per session. Assuming that the higher the sample size the more likely it will represent the norm.

If I play ten hands, I might expect an odd set to turn up once every so many sets...say 8-2 (either way) instead of 5-5. But how many times out of how many sets should I expect a set (count 100) to turn up 80-20? (I've probably played 250k hands - both live and on WoO) and never seen it). Over the life of the recorded sessions the numbers are startling normal, with the banker with barely (insignificantly) the one percent edge over the player (ignoring ties). The largest live set I have played was 1486 hands (19hrs of play), the smallest (live set) 40. The large set "normal", the 40 hand set, a staggering 4-40 (yes I bet against it, and ran out of money!). Unfortunately I did not stay to chart any more hands to find out how long it took to draw that percentage down to even the second standard deviation. The largest shoe difference I have seen is 48 to 10 with bunches of ties (I got on that one after the second set of 5 losses and made up more than I lost on the 4-40!) .

Quote:DaveWThe odds in baccarat of any given hand winning or losing have long been documented.

But, is there a way to determine the reliability of a given sample size?

"reliability" are you talking about online casinos?

I will not go there.

To find out what can be "expected"?

Yes, either with a binomial distribution function or a normal distribution and also a binomial standard deviation formula.

Look them up if you do not understand them.

You can start here:

http://stattrek.com/Lesson2/Binomial.aspx?Tutorial=Stat

Quote:DaveWObviously, the larger the sample size, the more likely the numbers will approach normalcy, so what I am asking is really:

NO,NO,NO

The larger the sample size the more likely the percentages or relative frequencies will approach their expected value(s).

The absoulte difference has a higher probability of actually increasing(the difference between Banker wins and Payer Wins) as "n" also increases.

The Law of Large Numbers deals with percentages, NOT absolute differences. This is where the Gambler's Fallacy comes from, this simple misunderstanding about the Law.

Quote:DaveW1) is there a way to reliably see if the casino is cheating by tipping the odds against someone who always bets only one way or another?

Again, if you are talking about online casinos, I will not go there.

For B&M casinos, why do they need to cheat? They have over a 1% edge on 2 basic Baccarat bets. They do not need to cheat.

Quote:DaveW2) 1.How long and

2. how often does a player have to play, in each session and as combined total sessions to have his own numbers reliably approach the established percentages?

IE. If a player plays 100, 500, or a 1,000 hands in a session what are his chances of approaching the normative numbers?

Again, be careful about numbers and percentages in the same sentence. Talk more about expectation, expected values etc.

This is a binomial standard distribution question.

Formula for BSD is: SQRT(n*p*q)

For 100 hands (no ties included) SQRT (100*.5068*.4932)= ~5.0 (4.999537579)

Banker EV (n*p) = 100 * .5068 = 51 (expected number of wins in 100 hands)

1SD (0.6827) = 46 to 56

2SD (0.9545) = 41 to 61

3SD (0.9973) = 36 to 66

You can now figure out 500, 1,000 hands or any other number of trials.

Quote:DaveWIf he plays 200 of these sessions a year, how many of these sessions will he find numbers within the first standard deviation and what would be the bracketed percentages of the other sessions?

You can now calculate that yourself with the above formulas.

Quote:DaveWMy second question pertains to the rate of expected return to "normal" as it pertains to sample size per session. Assuming that the higher the sample size the more likely it will represent the norm.

Only the percentages or relative frequencies converge to thier expected values, not the actual numbers or absolute differences.

Quote:DaveWIf I play ten hands, I might expect an odd set to turn up once every so many sets...say 8-2 (either way) instead of 5-5.

This to me is a simple binomial distribution question. You will not always get 5-5 in 10 hands, expectation is 5-5 only ~25% of the time.

Google "WinStats". It is a free software program that will do all the calculating and even run simulations for you.

Binomial Distribution table for n=10

Banker

0 wins; 0.09%

1 win; 0.88%

2 wins; 4.05%

3 wins; 11.09%

4 wins; 19.94%

5 wins; 24.59%

6 wins; 21.05%

7 wins; 12.36%

8 wins; 4.76%

9 wins; 1.09%

10 wins; 0.11%

Below are the distribution results from a quick 1 million hand simulation showing the Bankers wins.

`wins freq freq/100`

---------------------------------------------

0.00 808 0.08%

1.00 8682 0.87%

2.00 40392 4.04%

3.00 110655 11.07%

4.00 199278 19.93%

5.00 246400 24.64%

6.00 210206 21.02%

7.00 123560 12.36%

8.00 47922 4.79%

9.00 10888 1.09%

10.00 1209 0.12%

grouped data

items: 1,000,000

minimum value: 0.00

first quartile: 4.00

median: 5.00

third quartile: 6.00

maximum value: 10.00

mean value: 5.07

midrange: 5.00

range: 10.00

interquartile range: 2.00

mean abs deviation: 1.25

sample variance (n): 2.50

sample variance (n-1): 2.50

sample std dev (n): 1.58

sample std dev (n-1): 1.58

One can see how close the actual "percentages" became while there is between 300-500 in the actual(empirical) numbers verses the theoretical numbers. You can again use the BSD formulas to see what 1SD, 2SD etc are.

Quote:DaveWBut how many times out of how many sets should I expect a set (count 100) to turn up 80-20? (I've probably played 250k hands - both live and on WoO) and never seen it).

And do not expect to either.

Exactly 80-20 (in Bankers favor) is another binomial distribution question. In Excel =BINOMDIST(80,100,0.5068,FALSE)

Answer is 0.00000009474170% or 1 in 1,055,501,438 sets of 100 hands.

80-20 or higher (81-19 etc)(in Bankers favor) is 0.000000126106% or 1 in 792,982,821 sets of 100 hands. So you have a better chance of seeing 80 or more than exactly 80.

Quote:DaveWOver the life of the recorded sessions the numbers are startling normal, with the banker with barely (insignificantly) the one percent edge over the player(ignoring ties).

It would be "nomal" for the percentages to be as expected, not exact values.

1.36% difference is a big difference.

1 million hands (not counting ties) expectation is:

506,800 Banker wins

493,200 Player wins

an absolute difference of 13600 hands.

Quote:DaveWThe largest live set I have played was 1486 hands (19hrs of play), the smallest (live set) 40.

The large set "normal", the 40 hand set, a staggering 4-40 (yes I bet against it, and ran out of money!). Unfortunately I did not stay to chart any more hands to find out how long it took to draw that percentage down to even the second standard deviation.

The largest shoe difference I have seen is 48 to 10 with bunches of ties (I

got on that one after the second set of 5 losses and made up more than I lost

on the 4-40!)

Now you can work out the numbers for each shoe, not counting ties.

My sims shows an average of 1.028 as the absolute difference between Banker and Player wins per shoe with a standard deviation of 8.539

The longer one plays, the more trials and the greater chances of seeing something 3, 4 or even 5 SDs from the mean.

I hope you can take what I gave you and run with it. It really is simple math once you know what you are dealing with.

bac shoe that had 50 player

decisions and 10 banker. Not

a single chop in the shoe.

There would be 8 players, 1

banker, 9 players 1 banker,

etc. The bettors were going

nuts with tips and comments.

In a real casino there would

have been a cheer after every

player decision. Everybody

cleaned up, it was amazing.

I'd have emptied the tray as best possible given whatever the limit. I've cleaned up on shoes not even as good as you describe and the one you describe is a dream shoe!

Quote:MDawgI've cleaned up on shoes not even as good as you describe and the one you describe is a dream shoe!

I should have taken a screen shot, it

was solid blue. Normally 6 or 7 new

players join the table every minute.

The last 10 minutes they were joining

so fast the scrolling was non stop.

Constant comments like Holy crap

and OMG and GO GO GO and

I Don't Believe This. The dealer was

all grins as he constantly tapped

out Shave and a Haircut, Two Bits with

his chip on the plastic card tray for

the non stop tipping.. The

casino had to lose thousands. I've

seen hundreds of shoes since 2008,

this was one for the books.

Several times banker was showing an

8 and dealer turned a 9 for player.

In a real casino the yells would have

been deafening.