Human mind, symmetry and edge

Human mind is somewhat 'biased' by constantly looking for simmetry. Several studies have shown that when subjects are instructed to write down 'random' successions applied to a binomial probability, a sure undeniable 'overalternating' feature affects the results.
More interestingly is that real random objective successions are perceived by subjects 'less random' whereas unrandom successions are mostly considered as 'randomly' formed.

At baccarat the vast majority of people bet along those 'simmetrical' lines (widely intended), at the same time privileging just one kind of asymmetry, that is the 'long' streaks possibility.

Then there are 'foolproof' systems that give the subjective 'guessing' a 0 impact as every outcome  must fall into well restricted ranges.
Those worthless systems are mostly based around several kind of gamblers fallacies that many times are fallaciously(!) overtaken by the best short term move anyone could think of: progressive betting.
Nonetheless objective flat bet findings alone cannot lead to any EV+ with one billion of accuracy.
 
Therefore and simplyfing a lot, a 100% subjective way of considering things is EV-, as well as it's EV- a strict objective system stubbornly looking for precise triggers.
So the 'truth' must be in the middle. At least according to the money we've collected over the years at bac tables.

At baccarat the possible player's edge is a dynamical issue, surely defined after having measured long term flat betting results.

No matter the fkng strategy we are going to utilize, either we'll catch more W spots than L spots (after vig impact) or we are destined to fail.

Mathematically there's no way to 'guess' right by inserting a kind of 'subjective' sole element in our strategy whether bac successions are really random.
The same if bac successions are kind of unrandom.

On the other end, 'obiective' findings that tend to mix many different baccarat productions (card distributions) are sensitive to huge volatility that only in the long term will approach the expected values.

Subjective and objective strategies are two different categories of random walks following the same math laws, sometimes converging into the same betting line and other times negating it.
And obviously there are more likely positive or negative steps converging at a same betting line  than events forming long series of 'outliers' that could be 'heaven' or 'hell'.

For casinos the only tools that matter are the math edge and the several gamblers fallacies affecting almost every player. 
Educated players can overcome such factors.

as.

Al: yep Golden Steer at West Sahara (close to the defunct Lucky Dragon casino) is a classic for steakhouse lovers.

I'm curious about 888 Korean BBQ....

What about your preferred list of seafood restaurants?

@KFB.
Nice simple way of classifiying what we should expect and what we actually see at the shoe we're playing at.
I'll be back on this later.

as.
 

You are the best Al!

Let's prepare in advance that summertime session hoping KFB will join us!

First dinner it's up on me, my favorites are:

Picasso or Joel Robuchon (French)

Il Fornaio or Canaletto (Italian)

Wing Lei (Asian)

Xavier (Mexican)

Sushi Yiroyoshi (Sushi)

What about your preferites?

as.

For some reasons I'm more inclined to trust people betting serious money at bac tables and undoubtedly Alrelax and KFB belong to this category.

BTW, let me know if any long term member of this site would be interested to get a total free RFB accomodation in a couple of high end Vegas premises or in Montecarlo SBM properties and I will arrange it in a millisecond.

as.

 

I'm astonished to see that people keep thinking that baccarat is beatable by progressive bettings of any kind.

Without a verified edge progressive bettings don't work and can't work, actually they constitute a sure detriment of any strategy. (Variance is greater, vig impact is greater, maybe comps are the only reason to adopt this silly line). 

About the edge.

A possible edge can only come out after having verified it at large datasets and by adopting the same betting amount.

If we'd think that after betting 1, future 2 or 3 or 1.1 or whatever bets involve a better EV we are completely falling into the worst gambler's fallacy territory.

The EV of any method, strategy, system or approach remains the same whether we bet $1 or $200.000 per hand. And per every intermediate category of such a range.

Humans can't read randomness and without the help of math and statistics can't read unrandomness either.
Why?
As the human brain is somewhat biased about 'overalternating' and 'overclustering' patterns where some event or classes of events are mistakenly taken as 'more likely' or 'less likely'.

So we're sure as hell that 'subjective' methods don't make winners but just deluded people.

Most of the times anytime we put a 'subjective' element in our strategy we are just gambling. And gambling is a EV- move.

Notice that gambling forums and internet videos abound of wonderful winning shoes without mentioning or presenting the specular harsh losing counterpart, just in case labeled as a rare 'unlucky' situation that may happen.

Actual card distributions might be relatively insensitive to math and statistical long term findings but they do are to subjective methods in the same way.

To measure a possible winning strategy

Again our old betting random walk friend will help us to find out whether we're doing good or just for a transitory luck's (short term positive variance) impact.

Per every shoe dealt we assume to start at a 0 point, left side is the negative territory and right side is the positive one. Each bet won makes a step toward the right and vice versa for a losing bet.
Alrelax is completely right about this: every shoe is a world apart in the sense that previous outcomes cannot noticeably affect in any way the next shoe.

Obviously such random walk must take into account the ROI, otherwise a simple steady Banker wagering will approach more and more the far end of the right side. 

What is important is that positive steps must be considered under the 'coin flip' multilevel probability classes, what we name as a 'limited random walk'.
For simplicity and according to my unb plans, we consider just two back to back betting spots (that is a way different thing than betting all of the time two consecutive hands). 

Assuming for simplicity a perfect 0.50% winning/losing probability, odds to get a two unidirectional step at either side are 0.25%, so most of the remaining times per every two-bet wagered (for real or fictionally) we'll get a kind of 'balancement' movement where W=L or L=W.
In other words our betting random walk doesn't sensibly move toward any side.

This movement do apply to every two opposite events fighting and the least battle we should be interested at is the B/P distribution as affected by too much volatility.

So 'complex' opposite events taken under the two wagers line mostly move around a 0.75% probability to show up, that is a kind of 'neutral' situation producing the least number of steps toward any side.

As long as the 0.75% or so probability shows up, we can't lose serious money, maybe just the vig when applicable (actually a portion of winning hands will benefit from the B propensity).
The problem remains about what to do when the 0.25% unidirectional probability will come out.

The answer is about the more likely 'clustered-clustering' values happening at such less likely event.

Say A event will fight vs B event.
Most of the times (75%) AB and BA two-step situations will prevail over the AA and BB patterns (25% each).

After a AA or BB pattern had shown up, next event will form either another univocal AA or BB pattern or a more likely AB or BA event.
Let's classify the first AA or BB pattern as 1, AAAA and BBBB patterns as 2 and so on. 

Notice that differently than a simple B/P patterns distribution, many 'complex' A/B patterns involve a way lesser variance than B/P outcomes as more hands are needed to form a A or B pattern.

Therefore BB and PP patterns tend to distribute themselves by a stronger variance than AA/BB patterns.

Moreover BBBB or PPPP patterns (two steps deviating from a more likely albeit proportional course) are slight more likely to show up than the same AAAA and BBBB counterpart and this last feature is more and more predominant whenever we take into account several steps of such kind.
Of course and as already sayed, single shoe productions are finite so anytime a more likely value or class of values are surpassed, we better consider the room left to get the more likely events coming out.

as.

The important thing is leaving the casino with more money in our pockets. 

as.

Baccarat is a game of multiple asymmetrical steps

It's almost certain that baccarat may only be beaten by exploiting its innumerable asymmetrical steps acting per every shoe dealt, steps that by definition cannot be contained in a univocal cut and dried strategy.

There are infinite asym steps to take care of, and each of them must be studied by testing a large shoe sample then collecting more informations than we can.

Here's a brief list of such asym steps. (Obviously I get rid of the Banker math propensity)

1) Number and distribution of naturals

Naturals come out very often so that if an hypothetical naturals side bet would give the house a 15% or so edge we could easily destroy any casino in the universe.
And I'm not talking about card counting 8s and 9s, just considering an average distribution.
For that matter casinos could even cut off from the play two or 2,5 decks and nothing will change for us.
Unfortunately such side bet doesn't exist.

Anyway we know that naturals probability is symmetrically placed in theory but actually distributed by 'more likely' asymmetrical ranges we should consider before choosing the side to bet.
We can't catch when and where a natural spot will fall at, we do know that naturals move around more likely ranges depending upon the segment of the shoe considered.
One sided naturals probability is more than double of the asymmetrical math strenght favoring Banker, so just to simplify a lot the issue whenever we'd think a natural will come out soon we're way better to wager Player for some (few) consecutive hands than betting Banker by taking advantage of the math propensity to win.
A natural coming out at P side is a huge win whereas a natural coming out at B side is a sure long term loss.     

2) The highest two-card initial point is strongly favorite to win the final hand.

Almost every bac player do not give a cottontail rabbitsh.it about how this parameter went along the shoe dealt. (Let alone in their testings if they did any). They are interested about the win/lose destiny of the hand, bet or observed.

In reality such parameter is essential toward the final hand destiny and it moves around more likely ranges.
Sayed in another way, as long as our bets have caught more two-initial higher points than expected, the final hand destiny shouldn't bother us. And obviously I'm not referring about naturals and standing points, just about all other hands that must draw one or two cards.

3) Third card(s) nature distribution

This is a more intricate issue as many times it intervenes at both strong unfavorite or favorite situations.
It's undoubtedly sure that Player predominant shoes benefit a lot from the third card impact, way more than what third cards will help the Banker side.
In fact Banker side more often than not doesn't want to draw a third card unless it has a 0, 1 or 2 initial point.

I mean that it's more difficult to track the third card actual impact range than the above mentioned factors without registering the real outcomes, so we might simplify the problem by assigning a general positive value only at cards different than 0, 1, 2 or 3.
8s and 9s are generally bad cards for the Player but they could transform a very bad situation into a wonderful one.

Anyway the third cards impact follow some more likely ranges, surpassed whom we're not interested to 'guess' anything unless we want to take the 'sky's the limit' approach thus considering 'potential' probabilities as 'virtual' probabilities but with no guarantee to succeed.

as.

Notice to not make confusion between final points probability and actual final winning probability.

as. 

Hi KFB and thanks for your reply.

Original authors of such a system investigated deeply the game by a strict mathematical approach, they didn't mention the previous BP gap other than by the rules I've written here.
So I do not know if your idea could be implemented in the system.
Only a program could solve the issue.

What I could provide is this:

The largest majority of winning points at either side are included within the 9-4 final points.
At Banker side this probability is 73.1% vs a Player's 70.8%.
Considering this final points range, Player odd and even final points are perfectly symmetrical in their probability to happen whereas at Banker side odd points amount to a 37.2% percentage and even points to a 35.9% probability.

Obviously this odd/even final points range must someway invert at the other less likely winning points range, that is points 3, 2 and 1.
Now Banker side will get a 19.4% share and Player a 21% share.

Naturally a 0 final point can't win, anyway this possibility comes out 7.5% of the times at Banker side and 8.2% at Player side.

Not suprisingly and taking into account both sides the highest final points gap shows up at 5 point (3% propensity for Banker side) and at 4 point (2.4% propensity for Player side), the actual situations where baccarat rules advantaging Banker will work most.

In addition, Banker final odd points account for a 37.2% probability vs a final even points of 35.9%.

After all it's not a bad idea to play toward math percentages under conditonal probabilities happened. 

as.

A card counting system

Card counting doesn't work at baccarat, right?
Not necessarily.

Some time ago I've picked up this interesting method involving a card counting technique along with simple math features that gave us some valuable hints. After innumerable tests, we've modified it in such a way:

9 = +12
8 = +10
7 = +7
6 = +5
5 = +8
4 = +8
3 = +2.5
2 = +2.5
A = +1
10s and Paints = 0

Differently to any other card counting method where the sum is 0, here card values only add up giving some totals.
Our range of interest considers a cutoff point after 2/3 of hands are dealt (ties included of course).

Thus we assume that 50 hands dealt are a fair way to take such 2/3 percentage trigger.

Rules

1) We'll consider to bet only whether after 50 hands dealt the total sum is restricted within the 952-1180 range.

2) In the previous 50 hands shoe sample, Player results had to come out more by even winning points than odd winning points.
In the relatively unlikely situation that P even winning points = P odd winning points, we may still consider to bet if the total sum approaches the low end of the 952-1180 range (average point is 1066)

3) The initial burnt card must be counted.

4) For obvious reasons, lower the initial burnt card is more reliable will be our counting.

Naturally it's a wise move not to play at those casinos which like to cut down from the play many cards, anyway as long as less than one deck is unplayable we're doing good.

What to bet if all of the above conditions are fulfilled.

In the remaining 25 or so hands, there's a substantial math propensity to get Banker side not forming heavy negative outliers, meaning that at this 25 hands 'conditioned' sample Banker side will get a fair number of winning hands over the possible total outcomes.

Notice that this doesn't mean the Banker will get more winning hands than losing hands, just that sd values will be way more restricted than expected.

This is the only exception to step away from a strict flat betting scheme as now we have reasons to adopt a kind of progressive plan (multilayered multiple step schemes).

Disadvantages of such a system

1) Shoes fulfilling all the above parameters are rare to happen.

Even if the maximum limit of total sum range stays around an average value, best opportunities come out at the lower end of such range and they are quite unlikely to happen. 

Think that at baccarat more odd points than even points are made and we need a sort of inversion distribution to show up at Player side.

2) Possible numerous ties happening at this 25 hands segment could dilute a favourable conditional probability up to a point where we may find ourselves to be stuck without the room to get a fair number of winning hands.
We've found out that this is the very threat of such a system.

3) The indispensable tool to observe the shoe right at the start of it.


An opposite line of thought could orient us to think that whenever such conditions aren't fulfilled (it happens on the majority of the times) Player side will be more likely to get the same opposite propensity but our tests have clearly shown that this kind of reasonment is untrue.
Mostly for the odd/even points math nature of bac outcomes.

as.

Best baccarat player in the world (again)

The best bac player in the world is one capable to be ahead after 7-10 shoes played, meaning that in the vast majority of the times the 7-10 shoes probability range will happen by ascertained values not likely reaching outliers that of course could be astoundingly good or terribly bad.
Therefore under normal situations and providing a fair number of hands bet per shoe, recreational players and 'I'm the new genius in town' players will get a very slim probability to be ahead after such 10 shoes sample.

As sayed many times here, progressive plans do not solve such a problem, they just dilute it.
So without any doubt, a flat betting scheme or a very slow multilayered wagering (as brilliantly proposed by KFB in his section) is the answer to know if we're really doing good or just getting a fluke.

No room to false illusions.

After a 10 shoe sample, in the vast majority of the times we must be ahead of something by flat betting (or any strategy very close to it).
If not we're fooling ourselves.

This statement is so true that even by betting just 5-6 hands per shoe and after a 10 shoe sample, probability to be eventually ahead will be strongly shifted toward the negative territory. And by percentages well surpassing the math negative edge.

If this weren't true, baccarat wouldn't exist.
Transforming the issue, there are times when we should follow average clustered situations and other times where the 'clustering effect' on average is less likely to happen.

Lesser the number of bets we'll make per shoe, higher will be our probability of success.

as.

See you next week

as.

   

As long as an asymmetrical game provides not perfect random but dependent and finite successions we can't lose

Providing we can bet whenever we wish.

This statement may be set in stone by a 100% accuracy.

Besides the key assumptions made by R. Von Mises on randomness and M. v. Smoluchoswki studies we took decisive hints from, a shuffled deck makes more probable than not and at various degrees some events than others.
Diaconis and Bloom (and others) made extensive works on the subject, obviously not specifically considering baccarat.

Shuffle thoroughly a single deck of cards and itlr you'll get precise values of some events happening (for example red/black card apparition in streaks or 'chops'). By increasing the number of decks employed, such features will approximate but tending to proportionally deviate in relationship of the number of decks utilized (for the impact of variance).

It's obvious that at baccarat red or black cards distribution doesn't matter, so the process is more complicated to be grasped and naturally it should be assessed toward the 'average' key card concentration/dilution.

Since we do know that an average number of rows and columns will be formed, we may infer that at some points such key card distribution will make more probable some events than others.

Fortunately no need to track cards, average patterns and other tools will make the job for us.

See you in a couple of days.

as.

What many players may think is a simple event is in reality a "Complex" event. Meaning many things have to line up for it to present.

+1 KFB!!

We humans are shaped to look for patterns and ancestors looked for patterns coming out by 'clumps'.

More interestingly, a lot of studies proved our inability to detect random and unrandom patterns, in the sense that most subjects wrongly assigned the proper 'random' and 'unrandom' feature at the tasks examined.
(See Muller, 2001) for example.

So it's not surprising that most baccarat players like to detect patterns knowing or thinking the distribution to be random (so unbeatable) whereas on the contrary manufactured unrandom patterns are considered as 'undetectable' for possibly showing 'too low' concentrated clumps to take advantage from.
That's why the 'complexity' factor will be one of the keys helping us to solve the puzzle as acting both at random and unrandom shuffled shoes.

Actually perfect randomly shuffled shoes do not exist or at least the 'random' feature cannot be proportionally distributed along a 416 cards shoe.
This is a different statement than accepting a sure dependence of the results that solely taken cannot lead to nowhere.

Thus in the real world some (or many) unrandom segments will happen per every shoe dealt, so the complexity parameter might take a different role whether applied to a random model or to a unrandom model, in the sense that the previous patterns and the number of hands dealt so far constitute a decisive parameter to look for.


More on that later.

as.

Thanks KFB!!

Average card distribution

Don't make the mistake to consider baccarat as a game where everything could happen per every shoe dealt; but do not make the more fatal error of thinking that a method could get the best of it at every shoe dealt.

For example, we know that a given amount of rows and columns will be filled per every shoe no matter how weird is the actual card distribution.

Since we are not interested to guess every hand or most hands dealt per shoe, we instead should be inclined to understand what's the actual patterns flow in relationship of the general average findings.
And this is a strict card distribution issue obviously depending upon the possible unrandomness as outliers (strong deviations) are more likely to show up when the unrandomness level is quite high.

Naturally such unrandomness is symmetrical, in the sense that we can hugely benefit from it or giving us harsh times when it doesn't go in our favor.
Most of the times unrandom levels are low or very low, so they do not constitute a sort of 'stop' or 'sky's the limit' situations affecting our strategy, yet itlr it's more likely to get strong deviations when the distribution is unrandom rather than attributing it to normal fluctuations a random model exhibits. In a word unrandom distributions get a slight than expected probability to form clusters. 

Example.

Our targeted complex pattern will come out, on average, 4 times per shoe.
Since we've tested that generally we could get a huge control of its dispersion values, we number the general probabilities of such event coming out (Y) or not (N) at the actual shoe we're playing at:

YYYY
YYYN
YYNY
YYNN
YNYY
YNNY
YNYN
YNNN
NNNN
NNNY
NNYN
NNYY
NYNN
NYYN
NYNY
NYYY

Now we're not interested to hope that Y>N as Y=N, let alone to 'follow' seemingly univocal patterns (e.g. betting Y after YY or N after NN), just to get at least one symbol of each category to show up clustered at some point, obviously knowing that the 'average' 4 apparitions could fail, (it may be 3 or 5 or other).
Thus the position of such complex apparition along the shoe does make a huge impact over our strategy.

The expected probability to get any YY or NN pattern in a 4-betting trial applied to an independent random model is 14/16 (87.5%); naturally the remaining 12.5% YNYN and NYNY symmetrical occurences will proportionally cancel any advantage.
But baccarat is not an independent random model, especially while taking complex events as 'triggers'.
It's like that the 87.5% probability increases to 89% or more. Now we get an advantage.

But another pivotal effect of baccarat distributions is that complex events show itlr a minor than expected propensity to produce YYYY and NNNN patterns so increasing the apparition of the other more likely 12 patterns not forming an 'alternating' process (YNYN and NYNY).

Therefore less likely events showing up are NNNN, YYYY, YNYN and NYNY.

Remember again that we're not talking about B and P hands but about 'complex' events.
So in any instance B or P do correspond to Y or N.

The 4 average event apparition per shoe is just an example.
There are infinite 'complex' patterns to look for. I've chosen a decent average event's rate, a kind of 5.33% trigger considering 75 resolved hands per shoe.

The important thing to know, imo, is to never chase a situation to be clustered unless deriving from some kind of a 'complexity'.
Complexity tends to get rid of short term variance (platykurtic distribution), the actual distribution casinos make a lot of money from.

as.