REVERSE ENGINEERED BACCARAT

First I would like to thank the Wizard for all of the extremely valuable information and data. I have reviewed the web sites frequently for several years. (Had I started earlier, I would have saved a lot of time and money)

I have been working on gaming systems for quite a while now especially for roulette and baccarat. I have an engineering background and enjoy solving problems. I thought I had finally developed a very profitable baccarat sytem. My basic data has always been the Wizard's 1000 shoes. My system results were 1,080 net wins, 1,164 units won with a stake of only 78 units, and a winning percentage of 4.7% on bets made.

I thought that these were fantastic results. I also tested the system against 137 actual casino shoes I had played, and the end results were similar, but with a bit more variation along the way.

Then I thought I would check the system against the Zumma 1000 shoes and disaster resulted. The end result was 32 net losses for a total 134 units lost with a stake of 450 units needed.

What I don't understand is how I could have developed a formula which would work with so well with over 80,000 hands and not work with other data?? i can certainly understand creating a system which is fine tuned on a very limited number of trials, but not for such a large data set. And as I said, the maximum drop from the maximum balance at any one time is only 78 units out of 1,164 so there is a fairly consitant pattern.

Any ideas on this?? Thanks in advance.

Any system is effectively a bet against something that could in fact happen. However unlikely what you are betting against is, it will eventually happen.

It sounds like you just had more of a match for your unplanned scenarios with the newer data.

There was a first time the New Orleans got filled up with water! Before that, there was a system of betting that this wouldn't happen. It worked until that storm came through, and OOPS.

Just because nobody realized they were betting against that from happening, mother nature didn't say "oh so sorry you didn't prepare." The disaster was the result.

You cannot design any system to cover every possible thing that can happen.

Just some of the more likely things that can happen.

Betting against the longest long-shot yields the smallest profit leveraging the most amount of money.

When eventually even the longest long-shot that you are betting against happens.

There is your answer.

DISASTER. There is a word for it! The unprepared suffer!

There is no mercy for those in denial. Your world crumbles.

the Black Swan is lurking just below the surface unseen to all.....

How did you develop a formula that beats a game in which every single wager favors the house? What combination of wagers did you find that overcomes that significant disadvantage?

Quote: treetopbuddy

the Black Swan is lurking just below the surface unseen to all.....

Quote: BDMWA

Had I started earlier, I would have saved a lot of time and money.

If you stop now, including visiting the messages boards for actual gambling advice, you'll save a whole lot more.

Quote: BDMWA

What I don't understand is how I could have developed a formula which would work with so well with over 80,000 hands and not work with other data??
i can certainly understand creating a system which is fine tuned on a very limited number of trials, but not for such a large data set. Any ideas on this?? Thanks in advance.

So you beat 80,000 hands (1,000 shoes) ... big deal.
I bet you did not wager on every hand.
IMO, Your sample size is too small.

1,000 shoes is a such LARGE data set.
Who says?
Ain't so.

It is a small data set compared to the sample space.
Your 1,000 shoes you worked with (a sample) does not represent the whole population.
Or does it?? You may think so.

Did you include all possible (or close to it) shuffled 8 deck shoes?
I have seen that number before

416! / ((32!)^9 * (128!)) = 377 digits

164944883791287598411943116655903865118824068362
983914007087347084013829196364680846614865667459
022964650355867714652074291682596726831437397546
695056560355540746705637562240501155776567564785
022522219704232111595011221458008894065762070053
057540438153207938737327767112605303189053871399
047436327387252956448398946650681893111055183549
42178373831442020424406884197134050000000

and your sample size was 1000
Good Luck with your small samples

Your right Lexinger.....the one thing that I've learned during my two month stint in the forum is that nobody ever wins except the AP's. Any attempt to beat a negative expectancy is naive. Only a "chump" would try...... Since I'm not smart enough to be an AP, I'll move on, maybe try my hand at gold mining.

People win all the time playing casino games. A few hours here, a few weeks there and even some over a year of play.
Most all times due to plain old good luck.
The longer one keeps playing against a negative house edge game, the less chance they have of being even or ahead.

The OP needs more shoes to play against. A larger sample size to see how really good his method of play is.
He should show also his data (net win) per shoe since most Bac players do not play every hand.

here is a link to 10,000 more shoes
I would right-click to download as it is very large (21.3MB) or your browser window may not open it all.
Chrome took awhile to open it.
10k 8 deck shoes results


Good Luck

Thanks, 7craps for the link. Back testing ideas against 10,000 shoes will take time, maybe years. Years that would have been spent in the casinos losing money.

Quote: treetopbuddy

Thanks, 7craps for the link.
Back testing ideas against 10,000 shoes will take time, maybe years.
Years that would have been spent in the casinos losing money.

The 10k shoes fits into my Excel sheet (2007 version)
I cut it up so the file was not that big.
A few formulas and in 30 seconds I see one method (always betting Banker) is a waste of time.
It had a very large total loss.
So I bet it every other hand. Same total loss, but smaller than betting every hand.
(I went to find more methods at http://baccaratforums.com/ but that site has been down for some time now)

So my 3rd method
is waiting until a streak of 3 Player wins then bet Banker for the next 3- flat betting.
(takes a bit more coding but Excel can handle it no problem)
Still lost.

But for 2,000 shoes I did have a net win. (not over 10k)
Maybe that is they key.

Play one method for X shoes, then before you lose too much, change to another method.
This only took about 5 minutes in Excel.

Next test, wait for 4 Banker wins in row and bet Player for the next 4 hands with a 1,2,4,8 progression.
This one did well for hundreds of shoes. Not thousands

This ain't no fun testing, more fun playing for real.
Only took me 10 minutes to test 4 methods.

http://imspirit.wordpress.com/tag/baccarat/
here is a guy that did many simulations and tested many Bac systems over the years
We should start there and weed out the good from the bad

Compare your results to the chance of getting the same result betting randomly for the same amount every time you make a bet. Work out the standard deviation and expected value. See how far your results are from that s.d. Is 1-2 standard deviations away from the expected value in profit?

Say you have a system that only bets 1 hand in 5 and makes a flat bet. That's 2,000 hands. Expected value ~1.1% = -22 units.

The variance of a single Baccarat bet is 0.93, which over 2000 hands is ~42 (0.93 * sqrt(2000). Thus over 2000 hands you'd expect to be between +20 and -64 units around 66% of the time.

If your system hits say +10 units after playing through 10,000 hands (betting one in 5), it's well within normal variance.

(Hopefully I got the math about right here).

You can use this as a rough measure to see if you have anything useful.

Your math looks good.

Problem is, maybe not a problem...

a player that uses Money Management (stop loss and win goals)
and Bet selection methods

do not for one split nano-second believe they are just being the same as a random player.
Their results will be (with 100% degree of certainty) way better than just the random player.

MM and Bet selection makes bankrolls grow.
No need to fear the house edge.

Quote: 7craps

Your math looks good.

Problem is, maybe not a problem...

a player that uses Money Management (stop loss and win goals)
and Bet selection methods

do not for one split nano-second believe they are just being the same as a random player.
Their results will be (with 100% degree of certainty) way better than just the random player.

MM and Bet selection makes bankrolls grow.
No need to fear the house edge.

7craps, thanks for link, reading the Lee Evans e-book. "MM and Bet selection makes bankrolls grow, no need to fear the house edge" ....do you believe that or are you sandbagging "chumps' like me?

Quote: 7craps

Your math looks good.

Problem is, maybe not a problem...

a player that uses Money Management (stop loss and win goals)
and Bet selection methods

do not for one split nano-second believe they are just being the same as a random player.
Their results will be (with 100% degree of certainty) way better than just the random player.

MM and Bet selection makes bankrolls grow.
No need to fear the house edge.

Indeed... I'm just suggesting to map against the random player and see if you really are doing better than a 'ploppie'. You probably aren't.

" MM and Bet selection makes bankrolls grow.
No need to fear the house edge.'

Whether you fear it or not, the house edge will take your money.

Stop loss and win goals just means you will lose slower than a random player.

But per dollar invested you loss will eventually be decimated by the house edge.

MM & Bet Selection are all bullshit if you choose a negative expectation game.

Quote: Buzzard

" MM and Bet selection makes bankrolls grow.
No need to fear the house edge.'

Whether you fear it or not, the house edge will take your money.

Stop loss and win goals just means you will lose slower than a random player.

But per dollar invested you loss will eventually be decimated by the house edge.

MM & Bet Selection are all bullshit if you choose a negative expectation game.

Oh I might have misread 7craps as being sarcastic... :)

Money Management is a smoke screen, and bet selection is reading the entrails.

For those actually interested in the math. Out of 83,640 hands, I made 11,705 actual bets and had about a 5% net win rate which worked out to about a 4% net gain after commisions. I developed 3 different bet selection methods and only bet when two agreed. And I also varied the chips bet according to a formula.

The 11k bets may be too small a data set, but it was based on analysis of the full 83k results. I have found many systems which work for a time, sometimes hundreds of shoes. But I am still not sure what would be the minimum data set for proof of success.

Thanks for the comments.

Quote: BDMWA

For those actually interested in the math. Out of 83,640 hands, I made 11,705 actual bets and had about a 5% net win rate which worked out to about a 4% net gain after commisions. I developed 3 different bet selection methods and only bet when two agreed. And I also varied the chips bet according to a formula.

The 11k bets may be too small a data set, but it was based on analysis of the full 83k results. I have found many systems which work for a time, sometimes hundreds of shoes. But I am still not sure what would be the minimum data set for proof of success.

Thanks for the comments.

Proof of success? I guess it depends on the audience. It's impossible for a system to beat a game with a house edge over the long term, so you'll never be able to prove success in the eyes of most people. There's even a challenge set up by a member of this board because he's so sure that what you're trying to do isn't possible.

Systems work better as the min and max bet are further apart. The frequency at which they fail is typically related to how far the min and max are allowed to go.

The most impressive betting systems are the ones that don't change the bet amounts at all.

The ones that do work are the ones that are allowed an infinite bankroll.

Quote: BDMWA

For those actually interested in the math. Out of 83,640 hands, I made 11,705 actual bets and had about a 5% net win rate which worked out to about a 4% net gain after commisions. I developed 3 different bet selection methods and only bet when two agreed. And I also varied the chips bet according to a formula.

What's your range of bet? if it's 1-10, with 1/10th of the bets at the 10 unit level, only the higher bets will really adding to the variance, so we started to look at the variance of around 1,100 bets. On baccarat, that's +/- 308 units on an EV of -121. So being ahead is well within one standard deviation.

Hence my question. Hope this makes a sort of sense why you might want to calculate the possible results on a null-hypothesis (your bets win and lose at random).

Quote:

The 11k bets may be too small a data set, but it was based on analysis of the full 83k results. I have found many systems which work for a time, sometimes hundreds of shoes. But I am still not sure what would be the minimum data set for proof of success.

Many 100,000's of hands, best generated randomly, rather than data mining through sets. The set you use to build a method with should not be the set you use to test it. I'd generated the hands randomly, using your a RNG (streaming random numbers from a psuedo random number generator on a home pc should suffice, but if really keen, you could fold in some real random data streamed from a variety of places on the web).

Quote: BDMWA

I have found many systems which work for a time, sometimes hundreds of shoes. But I am still not sure what would be the minimum data set for proof of success.

Define "proof".

A. Logic shows no system is working. There is no need for samples or data sets. Statistics cant beat logic.

B. (For other settings) Statistical "proof" is a matter of confidence. It has nothing to do with proof in the sense of proving a theorem. You can never be 100% sure, so technically the minimum data set should be infinite (i.e. does not exist) or cover the whole population (prohibitive).