Baccarat 5 column statistics

It may be true that these patterns resolve into a 50/50 proposition given enough data, although in my testing and Fatheads testing these patterns hold strong for now. I'm more curious why these patterns seem to appear more than others. I mean some of these triggers have a better than 54% win percentage with the data Fathead has tested and that I have tested. It seems that I have found winning patterns, but I don't really know why they seem to win more than other patterns. Anybody have any ideas why this may be? Could it be a fluke? After all, no card counting is involved, so I can't prove an inherent edge except for the data streams we have tested against. So, why do you think some patterns appear more than others? I'm curious of what some of you think about the subject.

 

At 54% your in an upswing. Hell it could continue for you for the rest of your life. Question is when the 46% is happening. Is it all at once or spread out. Will your bankroll disappear in that 46% of wrong outcomes or will you make it through the tide. Will you use a slow progression and what is the number of wrong outcomes in a row on average.

This is all more important questions that if it shows up more. Money management #1.

 

MM is the most important key to success as using the wrong progression for any given bet selection, no matter how good, will cause your bankroll to tank. I was thinking a very slight negative progression coupled with a positive progression is the ticket or just a positive progression if your expectancy is at 54%. MM must survive the variance of the B/S.

 

Bets selection is king, baby, and strictly betting flat is the best way to go, hey hey!

 

flat ? Have some balls :3 2 1 can not bite you .



But timid 1 2 3 and up does bite .

 

If one understands the actual difference between gamblers fallacy and regression towards the mean, then we can expect future outcomes to be closer to the mean (closer to an average expectation or mixed result) immediately after we witness an outlier event.

After a streak of 4iar we can expect around 80-90% certainty that the next 4 results will not be another 4iar. A possible betting opportunity right there.

Note I am NOT saying that results are "evened up" as that is a fallacy, an example of fallacy being if we see 4 bankers then we expect the next outcome to be player. What I am saying is that the next 4 results are expected to be closer to an average or mixed result and not another 4iar streak whether on the same side or the opposite. Still 10-20% it can happen as it's only around 80-90% certain. Outliers do happen and can occur back to back.

I won't post a link, but you can google an excellent explanation of the above and test the simulations yourself. Google "Regression to the Mean & the Gambler's Fallacy - Simulated" on a blog called "Tantamanlands".

 

Well, you need to know, why, where and when regression happen (get a visual) understanding.
Let's say you bet against a window of events not reaching 2.5 STDV.
That is 12 versus 2.

The perfect window with a 2.5 STDV would be BBBB P BBBB P BBBB
Among that, your expectation is the third outcome of the player to show.

The trigger can come in any combination.

BB P BB
B P BBB
BBB P B
...

Here your expectation is two more players for the future of eight outcomes, two-point regression.
Is a little tricky to use a progression that capitalizes on two outcomes among eight attempts?

You can do the same with six attempts to capture a two-point regression.
Take a window of 14 versus 2.
3.0 STDV

PPP B PPP
BB P BBBB
...

Now, wait for six versus 1 in any combination.
Then play that you will get two-point regression for the next six outcomes.

Here is the math and probability ...

That 68.3% of the time the divergence would be one SD or less. Either side of the MEAN.
That 95% of the time the divergence would be 2 SD's or less. Either side of the MEAN.
That 99.7% of the time the divergence would be 3 SD's or less. Either side of the MEAN.
That only 0.3% of the time would the divergence exceed 3 SD's

Cheers

 

Nice post @Sputnik.

I am a little hesitant at the idea of suggesting the next result is more likely to being P as it's always 50/50. That is not exactly what regression to the mean implies, but over a set of results we can expect a P as a feature of regression to mean.

This next paragraph clarifies a valid application of regression to mean over a set of outcomes.

Regression to mean doesn't imply making up for an imbalance with an excess of the opposite outcome. It just means that the results will be closer to normal, like 50/50, "average", or "mixed" as we expect with results within 1 sigma (standard deviation).

And I agree with your table as that is a Normal/Gaussian statistical curve.

 

Excellent post. It really does explain why some patterns can occur more than others.

 

"After a streak of 4iar we can expect around 80-90% certainty that the next 4 results will not be another 4iar. A possible betting opportunity right there."

So you're saying that the 4iar could also be a mixed bag of results and not just a Player or Banker streak? Could these maths also be applied to a W/L registry do you think?

 

That's an interesting comment about the win-loss registry. I was trying to make a point of that in another thread. If all bet selections are the same then the wind lost registry should reflect the same randomness.

 

I think it could, given the underlying variable is random it will be random as well. However the ultimate W/L events will not have a 50/50 occurance if you are using a progression and recording a final W/L of a series of bets as it will be biased to certain patterns. So your distribution curve and mean will be different.

 

Agreed. The math should hold up in this case, and W/L should be the same as the math of the expected outcomes unless I'm reading this wrong

 

So using a progression with the W/L registries will distort the maths to different percentages are what you are saying?

 

Yes, the probability of W vs L will not be 50/50 so the mean will be offset.

Imagine a martingale it will skew probability to be lots of W and very few L.

For that reason best to focus on the actual outcomes and look for the mean reversion there.

There is nothing magic or fallacious in what I am suggesting here, just the obvious that outliers are by definition rarer than the average results we expect within 1 standard deviation which accounts for 68% of all observations.

Here is 1000 coin flips with 4iar streaks highlighted.
Clearly over 93.65% of the next 4 outcomes are NOT another streak of 4iar.



And another 1000 coin flips again showing 89.83% of the next 4 results not being another 4iar streak, but occasionally the streak does go longer than 4iar and sometimes one streak of 4iar+ runs into another opposite streak of 4iar+ all representing the 10-20% exception.



You can test this over and over and it is very stable to expect 80-90% of the next 4 results to be closer to the mean expectations which is a fancy way of saying "a mixed result".

I've just given 2000 coin flips, doing a million is left as an excercise for the reader

Bet selection wise, betting OLD after the break of the streak for a series win in the next 2 or 3 outcomes isn't terrible, perhaps with a 1-2 or 1-2-3, 1-1.5-3, 1-1-2 or similar progression to taste.

 

You can map results versus LW-Registry.
Let's say you want to explore Z-score

Then you can say that the two first groups of three increase by a whole point.
Then the other two groups of three increase by half point.

LLL = 1.0
LLL = 2.0
LLL = 2.5
LLL = 3.0

That can work as a guideline for someone who likes to get with groups of three and have an easy map of the values with one simple rule.
Two first two by a whole point and the second by half.
Easy to memorize.

They can be done with any combination.
Here is a little more complex by still easy to memorize using groups of two.

One whole
Three half
Two twenty-five

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

This way you can trach and chart the likelihood where you have 0.3% exceed 3.0 Z-score on either side of the mean.
This way you can see how common or uncommon you reach stronger and weaker Z-score waves using LW-Registry the help mapping the math and probability values.
No need for a calculator or mathematical knowlede.

That 68.3% of the time the divergence would be one SD or less. Either side of the MEAN.
That 95% of the time the divergence would be 2 SD's or less. Either side of the MEAN.
That 99.7% of the time the divergence would be 3 SD's or less. Either side of the MEAN.
That only 0.3% of the time would the divergence exceed 3 SD's

Cheers

 

Here is a chart of the z-score that can be used for any existing playing model.
Banker versus Player, singles versus series of three, or any other combination using a 50/50 probability.
You can build and memorize your own map based upon your way to trackk and charting LW-Registry and get one strong idea where the likelihood for the win.

1. Z-Score 0,02 - 1 singles contra 1 serie of three
2. Z-Score 0,73 - 2 singles contra 1 serie of three
3. Z-Score 1,18 - 3 singles contra 1 serie of three
4. Z-Score 1,53 - 4 singles contra 1 serie of three
5. Z-Score 1,82 - 5 singles contra 1 serie of three
6. Z-Score 2,07 - 6 singles contra 1 serie of three
7. Z-Score 2,30 - 7 singles contra 1 serie of three
8. Z-Score 2,51 - 8 singles contra 1 serie of three
9. Z-Score 2.58 9 singles contra 1 serie of three
10. Z-Score 2,70 - 10 singles contra 1 serie of three
11. Z-Score 2,89 - 11 singles contra 1 serie of three
12. Z-Score 3,06 - 12 singles contra 1 serie of three
13. Z-Score 3,22 - 13 singles contra 1 serie of three
14. Z-Score 3,37 - 14 singles contra 1 serie of three
15. Z-Score 3,52 - 15 singles contra 1 serie of three
16. Z-Score 3,66 - 16 singles contra 1 serie of three
17. Z-Score 3,80 - 17 singles contra 1 serie of three
18. Z-Score 3,93 - 18 singles contra 1 serie of three
19. Z-Score 4,06 - 19 singles contra 1 serie of three
20. Z-Score 4,18 - 20 singles contra 1 serie of three
21. Z-Score 4,30 - 21 singles contra 1 serie of three
22. Z-Score 4,42 - 22 singles contra 1 serie of three
23. Z-Score 4,53 - 23 singles contra 1 serie of three
24. Z-Score 4,64 - 24 singles contra 1 serie of three
25. Z-Score 4,75 - 25 singles contra 1 serie of three
26. Z-Score 4,86 - 26 singles contra 1 serie of three
27. Z-Score 4,96 - 27 singles contra 1 serie of three
28. Z-Score 5,07 - 28 singles contra 1 serie of three
29. Z-Score 5,17 - 29 singles contra 1 serie of three
30. Z-Score 5,26 - 30 singles contra 1 serie of three
31. Z-Score 5,36 - 31 singles contra 1 serie of three
32. Z-Score 5,45 - 32 singles contra 1 serie of three
33. Z-Score 5,55 - 33 singles contra 1 serie of three
34. Z-Score 5,64 - 34 singles contra 1 serie of three
35. Z-Score 5,73 - 35 singles contra 1 serie of three
36. Z-Score 5,82 - 36 singles contra 1 serie of three
37. Z-Score 5,90 - 37 singles contra 1 serie of three
38. Z-Score 5,99 - 38 singles contra 1 serie of three
39. Z-Score 6,07 - 39 singles contra 1 serie of three

 

I know this is facts and logic - but still, mind-blowing.
You can track and chart any trigger at one table and then play at another table.
Your individual personal permanence also follows the same math and probability.

This is why Hit & Run has no advantage with that aspect.
But I like the idea of using Hit & Run because of that aspect.
Because I rather rely on my individual personal permanence jump from table to table and sting like a bee

Cheers Sputnik The Gambler

 

Marigny de Grilleau ?

 

Hello TwoUp - the Z-score example is just a straightforward hit or no hit.
Yes, Marigny De Grilleau uses overrepresented events versus underrepresented events.

And I have the simulation software that pinpoints out all 3.0 STDV.
Where I can see how the regression towards the mean behaves for real.

Here is a chart where you can see the overrepresented versus underrepresented events.
And here we can clearly see how much the STDV drop when one or several underrepresented events show.
Overrepresented events to left and underrepresented on top.