If I had to test an approach the least option coming to my mind would be to use Evolution shoes.

At that site cards are so badly shuffled that it's very difficult to get hints about a "general" line of thought.

I suggest to test shoes coming from "shuffling machines" and/or preordered shuffles as someway we need the "randomness" factor to act for implementing more constant situations.

as.

If baccarat outcomes would be "humanly" detectable, the game would have disappeared a long time ago.
So we could conclude that the human power over the results is inferior than the normal (uncontrollable) fluctuations and of course there's still the HE burden.

So our mind can't get long term valuable hints about those undetectable situations typical of this game as we're facing a machine.
Moreover players' minds are biased by the current bankroll status, so hoping beyond the real probabilities to promptly recover a deficit or to try to get 'sky's the limit' profits around any corner.

And patterns our mind is looking for are quite different than baccarat patterns, so we need to deactivate a kind of "normal" ancestral process.
Only a third part tool can do the job for us, one giving a lesser damn about money, current financial status or other bighornshit as being lucky or unlucky for that session.

Baccarat successions

Bac sequences are affected by a kind of "dynamical" bias related to the actual shuffling but is quite difficult to be grasped as it isn't mathematically shaped.
So we have to approximate what most likely will happen by not taking specific patterns as triggers. And implementing other features.

Since we know well that a single mechanical approach won't do the job (for the actual variance issues), we could utilize two simultaneous approaches slightly differentiated but both exploiting some common triggers until quite strong deviations will happen. Then both will stop their action.
To get a long story short say I'm referring to long consecutive BR streaks (for example).

Obviously there's the problem to know which approach to use and we've seen that generally algo A performs better than B.

The choice of privileging algo A vs B came out from having ascertained that many shuffling machines provide a huge quantity of constant valuable patterns.
So if you are in doubt to pick up a casino to risk your money at, definitely choose a machine producer. Sooner or later you'll be able to spot the proper rhythm to place your bets.

Since such machines were deeply investigated by statistical experts that assured their random production by presenting complicated formulas, we may infer that algo A likes the randomness.
A paradoxical conclusion.

On the same token, preordered shuffled shoes should follow the same randomness aim, we have no reasons to think otherwise.

Since we never know about a given production, we better be ready to use the backup algo that in the vast majority of the times will take care of the job when the main algo seems to fall prey of the variance or just because it can't do any good for supposedly unrandom models.

To get an additional hint about how both algos move say they are particularly sensitive of the doubles positive appeareance...

as.

Thanks KFB!

Your comments seem to confirm our conclusions and we're very glad about that because we think you are one of the 2-3 best scholars ever met in the forums circus. 

If hands are not predictable in any way (and math tells us that), betting a lot of hands makes the casinos' fortune; if some hands are partially predictable (and some studies suspect this), betting many spots just dilute such advantage up to the point to directly fall into the EV- ocean.

Since the almost totality of baccarat players consider any shoe as a source of possible profit, casinos are pretty certain to extract a lot of money from them by levels far greater than the HE.

Algos are smarter than those players as they were built upon the average probability of getting profitable shoes, neutral shoes or unprofitable shoes in relationship of general values compared to the actual source of results, the initial patterns shape, the intermediate patterns shape, the asym/sym hands current ratio, etc.

Moreover they are instructed to put the brakes on even after a single losing spot, that is waiting  longer than what a human normally does.

And more importantly, they are totally insensitive to previous shoes' results and about the actual bankroll as they simply do not care a bit.

OoOoOo

It's true that everything can happen anytime and independently of the production source, yet itlr some values are slightly more predictable than others and they are quite dispersed along the shoe course.
But this propensity is not constant per any shoe dealt, so forming a distribution's curve slightly but more and more oriented toward the positive right side yet suffering some negative and sometimes harsh natural deviations.
In our opinion, such propensity is due either for a substantial defect of randomness or for a very very close perfect randomness reproduction. Two different random walks operating.

Once again we ought to remember that casinos' interest is to offer perfect random shoes where the HE and normal deviations will get their full value, thus giving no hint to the players.
But even unrandom shuffled shoes could be a threat for us as they tend to produce too whimsical result lines that not necessarily belong to more predictable patterns. But those 'unrandom' shoes provide several steps corresponding to a 'random' model. It's just the different rhythm that cares.

Basically and regardless of the two algos utilized, there are three different situations each shoe will provide:

a) Long shoe, that is a shoe filling many columns (so forming many singles and/or doubles with few or any long streak(s);

b) Medium lenght shoe where the single-double/superior streaks ratio is about 1 or close to 1;

c) Short shoe, that is the opposite a) situation, where few columns are filled and there's an abundance of consecutive streaks and/or lack of singles.

As KFB pointed out, it's slight less likely that after a given number of hands dealt a) and c) categories will all of a sudden shifts from a to c and vice versa.
On the same token, b) more likely category to happen MUST stay at a more or less same level of probability for superior times than the a/c classes, so producing many same lenght or almost same lenght streaks and again the clustering effect is of paramount importance.

Since any shoe is a world apart, such three different categories move in direct relationship of what happened in the previous hands of the same shoe, privileging the old clustering effect.

See you in a couple of days.

as.

"Son all you payed is the looking price. Lessons are extra"

From The Cincinnati Kid movie

 

Very good post!

I agree with everything you wrote and speeding up for long is very dangerous at this game!

as.

Well, those shoes were coming out from the same (biased, IMO) shuffling machine.

And the interesting part is that the backup algo got specular losing sequences, confirming the main algo's good predictability.

The problem is that when big money is allowed to be wagered, machines are not utilized or we can't precisely know the card distributions' source.
Fortunately we have found a tool capable to approximate at best the various card distributions by increasing by one step (or two steps) the algos action.

More later

as.
 

Great post!

Math can't be bypassed, yet there are some long term consistent players around the globe getting the best of many considerations Al made in his post.

After all we players are getting a minuscule part of what could happen at baccarat productions, so trying to focus about what is really happening could be an additional force to get a profit or to enlarge profits. Or, more likely, to reduce (a lot) the inevitable losses.

as.

How many consecutive times an algo is expected to lose?

Unfortunately (and obviously) there's no a specific answer and that's why a martingale approach isn't recommended at any game even by playing with an advantage.

But there are other options to be considered, for example what's the more likely LW gap (LW, LLW, LLLW, etc) or WW, WL patterns when utilizing a deep bet selection.

Now there's a better possible answer and it's at the very beginning of any shoe dealt.
And when the selection is ultra deep, the beginning of the shoe corresponds to the total shoe.

In addition, differently than the basic way of utilizing a 0.75 probability now we'll just consider a "same class" clustering trigger, that is waiting for a searched specific trigger to happen and then "hoping" that another one will come out before the losing counterpart class shows up.

Here our last shoes played by taking care of the very first two bets the main algo suggested.

LL
LL
WL
WW
WW
WW
WW
LW
WW
WW
WL
LW
WL
LW
W(-)
L(-)
WW
LW
WW
WL
WL
WW
WW
WL
LW
WW
LW
LL
LW
LW
WW
LW
WW
W(-)
WL
WW
WW
LW
LW
LW
LL
LW
LL
LL
WL
W(-)
LL
WW
W(-)
LL
LW

(-): no second trigger showing up until the end of the shoe.

Summary

Total situations splitted in double W/L results =51

First W=28
First L=23

Second W=31
Second L=15 (5 hands not classified by lack of triggers)

WW=16
WL=8
LL=8
LW=14

Is this a short term positive variance?
No fkng way as the same results were registered after thousands and thousands of REAL LIVE shoes played.

Now let's look at the third trigger eliciting a bet:

W
L
W
W
W
W
W
W
L
(-)
W
W
L
W
(-)
(-)
W
(-)
(-)
W
L
(-)
L
(-)
L
(-)
L
W
W
L
W
W
L
(-)
(-)
L
L
(-)
W
(-)
W
(-)
W
W
W
(-)
W
W
(-)
W
(-)

W=24
L=11

(-)=16

Check each of the 51 rows (shoes) and you'll see that out of 35 resolved situations per shoe, only shoe #2 provided a LLL sequence, that is a 1/8 expected probability.

Since those shoes were quite whimsically patterns shaped, we guess that the main algo while suggesting a general "more probable results flow" seems to make a very good job as no matter which point of intervention it picked up, more Ws than Ls are coming around even if here we haven't considered  the vig impact, anyway applied to a very low number of bets made per shoe.

See you in a couple of days.

as.

That's interesting but I guess that playing by "sections" needs a lot of experience as very often the sections texture is too confusing to be valuably exploited.

as.

Definitely I can't dispute your arguments Al, yet it's quite difficult to spot the favourable/unfavourable sections you've mentioned unless (IMO) we try to put the basis for a strict mechanical way of considering outcomes.

Betting frequency

Trying to be more right than wrong per every hand dealt is impossible and we're pretty certain that it's impossible either when betting half hands of a shoe.
For that matter we have tried to operate our algos by a betting frequency close to 1/3 of total hands dealt and itlr results were a disaster.

To stay put in the vast majority of the times is the key to have success itlr as just a single or a couple of hands that went wrong mathematically could transform a wonderful sequence into a horrible one. At both algos action, I mean.

Yes, quite often positive situations could last for long or very long but the negative counterpart  must be avoided at all costs, even though it will be slight less likely to happen.

After all, from a math point of view it's ridiculous to think that a EV- game could be beatable itlr and our algos know very well that. So they are built toward the maximum possible level of "safety", so putting more emphasis on not losing instead of winning (a lot).

Not every "low betting frequency" player will be a long term winner but 100% of long term winners bet very few hands per shoe. At least when they bet the principal amount they're interested to wager as the "illusion of action" is a common trait among real winners, capable to concede the HE at their lower wagers and able to exploit an edge at pivotal hands.

Without any doubt every soul betting more than 20-25% of total hands dealt is a sure pure loser and among this category maybe just one or two players are really defending their bankroll wisely.

The "experts" argument that people wagering rarely are just diluting their losing expectancy is almost always right but not 100% right. That's where our edge comes from.

Technical features

At a EV- game the best math move to win is not to bet many portions of our session's bankroll but to bet everything in one attempt (Bold Strategy).
Since we have ascertained that some baccarat spots are EV+ interwined in the EV- ocean, we might come at the conclusion that it's way better to bet a lot at rare spots than to dilute our wagers by fearing that unfortunate situations could come out and depleting our session's bankroll.

More on that tomorrow

as.

Al, what you have written is absolutely true: more patterns are displayed in the screen more action is expected from players in the hope that something univocally shaped will prolong by a "sky's the limit" direction.

Nonetheless, some derived lines will make way more probable the appearance of some (rare) univocal patterns than at Big Road.

I've provided some examples about that:

-At BYB road and SR road very long consecutive double patterns are 10-fold more likely than at Big Road.

-At CR road and assuming the same number of shoes dealt, the probability to face long streaks (say superior than 6 or 7) is greater than at Big Road.

But as you've correctly pointed out (and knowing that anyway such situations are relatively rare to happen) most players are not going to wait for them, thinking that every shoe dealt will present such slight abnormalities. (of course assuming they are aware of them)

Casinos are patient to wait for their math edge to show up and for the many mistakes players will make in the effort to win or to win more or to break even in too short terms.

as.
 

Basically algos are instructed FIRST to catch the more likely world (where E/F ratio is close or very close to 1 so the permutations issue takes a major role), THEN to get the best of the F clusters and to get the least damage when E clusters come out. (see later)

It's intuitive to realize that PER EVERY SHOE DEALT patterns as doubles vs 3+ streaks or other simple "fighting" patterns do not belong to a kind of constant E/F ratio as being affected by a fair or huge degree of volatility.
And algos hate at most variance.

This happens because at those patterns the dependency factor tends to be too low to be exploited, so for the purpose of what we're talking about such patterns don't give us an idea of what the actual card distribution should be more likely entitled to do.

Back to the previous topic.

The principal aim of our A algo is to bet towards the more probable permutations applied to a more likely E/F ratio; when rare negative permutations take place we have to run the backup algo or simply quit that shoe.

It's of particular interest to know that in the process of continuing or discarding from the play the actual shoe, the streaks distribution makes a decisive role: that is the clustering or the isolating effect.
By applying a proper rhythm of considering the outcomes, each class of streaks will be more clustered than isolated and of course when E streaks are coming out clustered we have less reasons to attack very soon that shoe. And vice versa for F streaks whether showing up isolated (so preceded and followed by a different streak class).

That's one of the best tool to think about prolonging a possible profitable succession or to stop it.
And in fact when we decide to put in action the backup algo, we'll see that in the vast majority of the times, the clustering streaks effect had taken place.

No wonder that I'm talking about the more probable classes of streaks...

as. 

An algo starts with the assumption that most shoes are concentrating toward a close to equal enemy/friend ratio, of course it must also take into account the actual streaks lenght (besides of complicated card situations I do not want to discuss here). And those streaks are just one of the main indicators of the level of randomness (IMO).

I mean that whenever an "average" shoe is showing up it's virtually impossible to lose as the algo takes care of the enemy/friend gaps and consecutiveness.
Thus whenever the E/F ratio remains constant or close to equal at the end of the shoe, it's just the permutation issue to be "controlled", meaning that we should be prepared to deal with ALL possible permutations.
Obviously there's a "more likely line" to be followed as not each permutation will get the same probability to appear (think of the enemy coming out severely clustered, for example).

On the other end, severely clustered enemies might be a sign of those "deviated" shoes giving us some or a lot of trouble and normally such shoes will more probably belong to the "unrandom category".
For example, when sharp enemy clusters happen at the start of the shoe or in the first 1/3 of the shoe, we might think that the E/F final ratio will be surpassed beyond the normal values.

Then there are the rare enemy situations intertwined by long friend streaks and fortunately those  deviations are slight more likely than the strong enemy clusters counterpart.

Naturally there is no guarantee that the deviated category at either positive or negative side will come out "when we wish" and it's very likely that after one or a couple of negative shoes the player's attitude is flawed.
But not that of the algos.

More later

as. 

While playing baccarat a funny thing to observe is that 90% of people keep betting what "they feel more likely to come", especially if they are used to bet every hand or almost every hand.

Actually it's just what casinos hope for: that is having players betting for this or that in the effort that something univocally shaped will happen for long.
Or, even worse, that something not belonging to a given actual predominant category will magically change as "their tests dictated so".

Algorithms move just around the middle of the operations field: they know that something will prolong and they know that something will stop very soon by specific more likely terms in relationship of the estimated level of randomness of the actual shoe.

They are so acute that they are able to provide close to optimal choices worth of erasing and inverting the HE.
And it's not a coincidence that they need a fair number of hands dealt before eliciting a given real betting action.

Algos are more cautious than us, they know that "easy solutions" are not coming out around any corner.

They can't give a lesser dam.n about our current bankroll status, they simply suggest best options knowing that itlr they'll be more right than wrong by a degree capable to invert the HE.

When algos don't suggest any bet (and such thing happens quite often) it's because they are not able to spot favourable situations either at unrandom shuffled shoes and at perfectly random shuffled productions.

Finally algos are the best permutation forecasters as they constantly evaluate an average enemy/friend ratio related to the actual portion(s) of the shoe.
A thing that we'll see in a couple of days.

as.

Paradoxical arguments and considerations

It's sure that an independent and random binomial production can't be beaten by any means, let alone if results are unfairly payed.

Baccarat isn't a perfect binomial game as winning bets are unfairly payed (B) or getting a lower probability to appear (P).
We can't do anything to alter this.

It remains to evaluate the independence and the randomness factors that, in our opinion, are linked together.

There's no way to beat a game whether we couldn't spot the possible unrandom events but we need to know which direction such unrandom sequences will take as at baccarat a "general rule" cannot work  (differently than black jack where low cards favor the dealer and high cards and aces favor the player).

Moreover and even if the actual production is "random", the independence factor between hands could be investigated in order to assess possible "more likely" patterns.

Gambling experts have tried to estimate mathematically the dependency of the outcomes in relationship of the actual shoe's card composition assuming some cards are good for the B and others for the P.
They were right but the effect is too tiny so useless.

Anyway we have the proof that a given card composition left will make more probable (albeit by a very very very very slight level) a side to happen than the opposite.

Therefore we might infer that a dependency between hands exists so baccarat isn't a coin flip but not because it lacks of the perfect symmetrical BP probability to happen.

To summarize the issue in practical terms, IMO we have just two options of thinking to really beat baccarat:

a) we have reasons to think that the production is quite unrandom and we have tools to valuably ascertain the unrandom segments' direction of any shoe or most shoes dealt; here the dependency factor takes a minor role as it cannot be assigned before or along the shoe dealing.

b) we have reasons to think that we're facing a very close to perfect random production affecting the results, so we must find "mechanical" ways to enlarge profitably the important dependency factor.

Since casinos' aim (at least at HS rooms where a lot of money is wagered at) will be to get more randomly shoes than they can, our main algorithm should take care of a huge degree of randomness and, more importantly, to find solutions about enlarging the dependency world.

In the circumstances where the main algorithm seems to fail, the backup algo will make the job as it is calibrated to get the best of it by bearing the most unidirectional volatile situations typical of a partial unrandom model.(And half of the time such situations are easily humanly determined and the other half are very difficult to be grasped). 

The paradoxical conclusion is that shoes approaching the perfect randomness attribute will be more likely to be "controlled", that is giving more detectable spots to the main algo that takes care of the average card distribution (and able to enlarge dependency features).

We guess that it's not the casino's interest to badly shuffle the cards....isn't it?

as.