Baccarat 5 column statistics

I just want to illustrate the STDV values.

Here is one example where we never exceed 1.5 STDV and that the bias overcomes the opposite swings.
The sample is 1500 trails with around 30/40 triggers for each 300 trail sample.

LL = 0.50
LL = 1.50
LL = 2.00
LL = 2.50
LL = 2.75
LL = 3.00

Code:

331 lw 
212 w
332 ll 
211 w   
221 lw
311 w
321 lw
313 w 
221 lw
233 ll
312 w   
311 w
333 ll
211 w   
311 w
211 w
231 lw
213 w
321 lw
331 lw
221 lw
321 lw
321 lw
321 lw
231 lw
211 w
221 lw
213 w
331 lw
331 lw
311 w
223 ll
331 lw 
231 lw
321 lw
231 lw
- - -
211 w
312 w
331 lw
221 lw
222 ll
322 ll 
321 lw 
311 w
312 w
221 lw
331 w
223 ll
312 w   
331 lw
332 ll
331 lw 
311 w
213 w
232 ll
233 ll 
312 w   
213 w
212 w
311 w
313 w
211 w
313 w
222 ll
333 ll 
212 w   
321 lw
312 w
311 w
321 lw
222 ll
231 lw 
- - -
311 w
223 ll
222 ll 
331 lw 
231 lw
211 w
323 ll
221 lw 
211 w
331 lw
213 w
211 w
331 lw
231 lw
311 w
322 ll
211 w   
221 lw
331 lw
311 w
321 lw
211 w
313 w
313 w
311 w
223 ll
313 w   
313 w
231 lw
211 w
311 w
211 w
231 lw
- - -
213 ll
211 w
332 w
221 w
311 w
213 ll
311 w
322 lw
232 lw
321 ll
331 w
331 w
212 lw
232 lw
311 w
311 w
213 ll
311 w
311 w
322 lw
211 w
232 lw
322 lw
212 lw
311 w
222 w
331 w
212 lw
211 w
211 w
313 lw
233 lw
223 w
213 ll
322 lw
221 w
311 w
212 lw
213 ll
312 ll
232 lw
- - -
311 w
213 ll
311 w
223 w
232 lw
311 w
221 w
312 ll
311 w
231 ll
322 lw
312 ll
221 w
312 ll
323 lw
313 lw
321 ll
321 ll
211 w
311 w
312 ll
211 w
332 w
222 w
223 w
223 w
211 w
323 lw
321 ll
222 w
323 lw
211 w
213 ll
222 w
222 w
323 lw
221 w
231 ll
223 w
- - -

 

It's around 185 triggers and none exceed four in a row.
And I run 185 RNG trails and get several six in a row, a couple of seven and eights and one eleven in a row.
Just to put things into perspective.

Cheers

 

Thanks Sputnik.

Just trying to make sure I understand your table, the first column 331 etc is referring to streaks and your second column is two bets after the sequence?

As for PRNG I avoid them as they have impossible requirements. They cannot provide an equiprobable outcome of every possible result on the very next output as they have insufficient internal state. It is impossible to output a single deck shuffle from a PRNG due to insufficient degrees of freedom.

But they do try to simulate unbiased outcomes in first order outputs but then when you pair outputs they don't get equiprobable outputs of pairs. They also fail collision tests, so if you generate block of numbers and repeat this over and over you should expect to get a match between blocks as per the birthday paradox, but they can't serve all masters that random demands at the same time when they are simulating random. They also fail basic binary rank tests (basically even vs oddness) and many other statistical tests. It is just not worth using them.

If you want to know more..
https://en.m.wikipedia.org/wiki/Diehard_tests

 

It is impossible to output a single deck shuffle from a PRNG due to insufficient degrees of freedom.

I have tested this on real shoes and get the same results, problem is that you only get four to five triggers.
The output is using the concept of the principal of 1/3 where two out of three events will dominate and strike several times in a row.
No matter real shoes or a pseudo RNG.
That is the concept and Sputnik's March.

The certain thing I don't show is that some trigger can strike up to twenty-five times in a row.
WLWLWWWWWLWLWLWWWLWLWWLWWLL

I just decide to show the entering point that I decide to work by themself.
So the bias comes from Sputnik's March concept.
That is two events that will always dominate more often than you get triplets.
The Bias.

And how do you explain I only get four in a row after testing 185 triggers?
When I run 50/50 ordinary stream and get several six in a row, a couple of seven and eights and one eleven in a row.

For me, it should have been the same with the 185 triggers, but is not.
That proves my point about the insufficient degree of freedom, the bias.

Cheers

 

TwoUp you can have the last word on this, I will not change my mind.
And I only play for entertainment and have a good time at the casino.
For that reason, there is no need for me to continue with my arguments.
I am happy and proud to develop Sputnik's March Concept.

Cheers Patrik

 

I didn't think we were disagreeing, but thanks for your posts

I'm also not familiar with your march concept.

My only point is that RNG will behave differently and fail certain statistical tests compared to real random and this may manifest is odd results vs real random RNG due to many inherent weaknesses in their design that cannot all be overcome.

Quality of output can also vary wildly based on the internal state of the RNG as it evolves, so perhaps this is why you sometimes see differences??

Simulating random is inherently difficult and technically impossible, so the illusion will always have limits.

 

I remember messing with the birthday paradox back on Gamblers Glen. I can't quite remember how to use it, but I remember it working quite well. I think it was betting for a trigger on a series of results against another series of results. I know you pretty much explained most of it here but is there a way to put it to practical use? I also like betting OLD for 2 to 3 decisions after a streak of 4 or more; I feel like that trigger should be tested. Thanks for your insights.

 

Yes OLD is a pretty good bet selection most of the time, and especially after the break of a streak of 4 or more.

Using a two step negative progression 1,2 or 1,1.5 does capture a lot of wins.

With a few rules like waiting 1 outcome after a 2 step fail before entering again (providing it's still chop) or standing aside until a streak ends (or switching to FLD until streak ends).

You lock a win on singles and 2iar and many 3iar but the 3iar will sometimes lose if they follow a streak of 3iar+ perfectly in sync with your bet placement. But 80-90% of the time we don't get another streak after 4iar or more.

 

"
Sounds good to me from a statistical standpoint. Looks like a good way to snare wins after 4iar+.

 

With a few rules like waiting for 1 outcome after a 2 step fails before entering again (providing it's still chopped) or standing aside until a streak ends (or switching to FLD until the streak ends).

- - -

The following is my opinion ...

TwoUp - you touch on something i would like to call a misconception betting against series.
You write "Switching to FLD until the streak ends" which is the correct way of betting against series.

One old selection is betting against a series of two to become a series of five in a row using three-step progression.
But the misconception is that they are betting against a series of six, seven, eight, nine, ten, eleven, twelve and so on ...
The misconception betting against a series of five.

Using GAP methodology you can isolate the series you betting against.
For example betting against a series of three, four, and when you reach five you bet FLD for six in a row.
That way we TRULY lose against a series of five.

The same applies to betting against a series of four in a row.

Cheers Patrik

 

Yes the Gap is important to isolate losses to just one pattern.

If making just 3 bets, the following are the nemises patterns, but still we win on some 3iar using the gap as shown here we lose the first 3iar but lock a win on the second 3iar:
—LLL—LLW
XOOOXXXO
OXXXOOOX

Now if you do decide to follow streaks using FLD, using the gap and then bet FLD for 4iar to become 5iar there are more nemesis patterns.

But first here is an example of the 3iar becoming a 5iar win:

—LLL—W bet FLD, series breaks so back to FLD
XOOOOO
OXXXXX

But, the 4iar is now a nemesis when switching to FLD. The ultimate nemises is a run of 4iar followed by another 3iar or 4iar with no chop:
—LLL—LLL bet FLD, series breaks so back to OLD
XOOOOXXX
OXXXXOOO

Again based on the reversion to mean after an outlier this only happens 10-20% of the time. After a run of 4iar a mixed result is more likely than another 4iar.

I tend to prefer not using FLD and just step aside and avoid that second nemises pattern.

Getting runs on chop, 2iar and 2 out of three 3iar is good enough.

 

Ok I've had some time to do some reading, and your march is quite brilliant. Thanks for sharing.

This following variation I was particularly interested in, because there are 16 ways to win and only 2 ways to lose with just 2 bets.

Anyone reading this, let that sink in.

Even using a two step martingale 1→2, means you lose at most 2×3 = -6 units but win +14 units on the other combinations with a net of +8 units taking all combinations into consideration.

There are other MM progressions such as delayed Fibonacci, Holloway or even labbys or divisor approaches I won't get into.

Now, not all combinations are equiprobable, more on that later.

The table below 1,2,3 means 1iar, 2iar, 3iar+ (3 or more).

Winning bets are bold, underline are losses.

Sputniks truth table:
222 W bet for 2iar+
223 W bet for 2iar+
221 W bet for 2iar+
232 W bet for 2iar+
231 W bet for 2iar+
233 W bet for 2iar+
211x LL bet for 2iar+ and lose, then bet for 2iar+
212 LW bet for 2iar+ and lose, then bet for 2iar+
213 LW bet for 2iar+ and lose, then bet for 2iar+

333 W bet for 2iar+
332 W bet for 2iar+
331 W bet for 2iar+
323 W bet for 2iar+
321 W bet for 2iar+
322 W bet for 2iar+
311x LL bet for 2iar+ and lose, then bet for 2iar+
312 LW bet for 2iar+ and lose, then bet for 2iar+
313 LW bet for 2iar+ and lose, then bet for 2iar+

In simple terms the above variation boils down to betting 2iar×2 after a 2iar+.

A quick look we see the shortest sequences are the ones that lose, so based on conditional probability calculations they are expected to be more common than the other winning sequences that are longer.

 

Ok I did some calculations, and as I suggested in my previous post, the probability of the shorter sequences that lose, dominate the winning sequences.

This doesn't mean the march is not useful. To the contrary if a method provides a stable stream of W/L in the registry without too many L streaks then it can be very playable using a MM method.

I have done the math with flat betting and a 1→2 progression for each of the two bets and also just focusing on the sequences that start with 3iar+.

I also indicate the total percentage of shoe sequences you will be betting (around 16%) which is used to normalise the probabilities for the actual sequences bet.

Hopefully the tables are useful for those wanting to analyse Sputniks March from a probability standpoint.



Note: Combinatorial math is easy to get wrong by missing something, so apologies if I have made a mistake.

 

Thank you TwoUp ...

Cheers

 

Be aware there are some errors in my calculations, but it's late and I will post an update soon. Calculating probability is fraught with danger, especially when it comes to calculating "X occcurances or more" potentially followed by another event (or not).

 

Hey Fathead,
In order to save me some time, I was wondering if you could test your shoes against what Twoup and Sputnik are talking about?
Bet OLD 2 times until a win after a series of 4 or more. I tested 200 shoes by hand and this way of betting seems promising.

 

'Series' meaning a streak (PPPP or more)? Betting opposite the streak for 2 times only? Or bet against the streak until you win?

 

Betting against a series of BBBB+ or PPPP+ for 2 hits.
Example: BBBBP.. bet Banker twice until a win or PPPPB .. bet Player twice until a win.
Betting against a Banker or Player streak that has 4 or more.

 

So the BBBB is the trigger and you would bet P up to two attempts seeking a win. Then wait for the next BBBB or PPPP?

 

If the above is correct I have a whole thread about this.

https://www.gamblingforums.com/threads/pppp-bet-b-update.22290/

I ran it again just limiting it to two bets. After commission, BPPPP bet B and BPPPPP combined win money flat betting. PBBBB bet P and PBBBBB bet P combined loses money flat betting after commission.