+ Register

[AsymBacGuy]

Betting plan

Baccarat world is made by infinite situations that we could restrict by considering streaks.

More hands we'd try to guess lesser will be our profits, or better sayed, greater will be our losses.
There's no fkng exception about that, believe us.

Mathematicians and 'so called' gambling experts will say to us that no matter how's diluted our betting, the EV remains negative yesterday, now and in the future.

This is a complete false statement as it considers any shoe dealt springing from the same shuffling source.
We've seen that it makes quite a difference when a shoe is shuffled by a machine or manually or by "only gambling gods know" preordered shuffles.
The 'cut' made by players (when allowed) isn't a proper tool to break the shoe sequences, it just postpones them.

Hence we must be prepared to deal with sd values (at both positive and negative ways) and machine sd values are way better controllable than other shuffling procedures.
Next are manually shuffled shoes and at the bottom of the list come preordered shuffled shoes where we do not get any hint about how they were shuffled.

1) How many 5/5+ streaks are we expecting from any shoe dealt?

We strongly think that percentages presented above are accurate, we do not care less about millions of shoes dealt especially when a homogeneous source wasn't considered (so you could put in the trash every data you've collected unless coming out from the SAME source). In the same way we shouldn't give a fk about B/P long term raw percentages as sure as hell they will be approaching more and more the 0.5068/0.4932 useless ratio.

We need to assess the average streaks distribution after having devised one or more proper random walks.

After having classified every "long streak" happening as belonging to the 5/5+ category, the positional distribution of them is paramount as way more than half of the times we'll expect zero or one or two 5/5+ streaks happening along the entire shoe; obviously more such long streaks happen at the start or intermediate parts of the shoe greater will be the probability to get a shoe surpassing the inferior than three/superior than three streaks number ratio.

Remember that per every seven shoes dealt on average you'll get ZERO shoes presenting a 5/5+ streak, and at the worst situation you're almost 3:1 favorite NOT to get three or more 5/5+ streaks.

More simply, well more than half of the shoes dealt will include just one or two 5/5+ streaks whatever distributed.

Best scenarios and worst scenarios

Well, best scenarios are when zero 5/5+ streaks will happen along the entire shoe (nearly 13.8% of the times), worst scenarios are when the "four" cutoff streaks number  will be reached or surpassed (say it's  about a 5% value or so).

So we're nearly 2.77:1 favorite to get zero 5/5+ streaks than having four or more 5/5+ streaks at any shoe dealt.

Problems arise when we have to decide the more valuable spots to deny such possible 5/5+ streaks

We'll see this issue in a couple of days.

as.

[AsymBacGuy]

BTW: try to spot two different mechanical random walks getting a slight different pace capable to exploit the most asymmetrical bac features and you'll get one of the solution to get a sure indeniable long term edge.

We'll see how a multilayered progressive plan will get the best of it.

as.

[AsymBacGuy]

IMO and according to our data, baccarat outcomes move around a kind of constant 'asymmetrical' propensity widely intended.
We've found that streaks are a better target to set up a plan about and they always start after a same result happened at least twice.
On the other end, when a same result happened 5 or more times, we are not interested anymore to assess the streaks destiny as longer streaks most of the times are affected by a kind of "math inversion" due to the third(s) cards impact.

In poor words, we think that the 2-5 streaks range is the best to risk our money at.
Moreover we know that the 'overalternating' results succession is slight less likely to happen, privileging back-to-back same outcomes where by far the more probable cluster is one (and at a lesser degree, two).

Back to the 'asymmetrical' feature.
The interesting point is that the baccarat asymmetrical world tends to make a decent portion of symmetrical events in a row, meaning that same streaks lenght or two streaks classes are slight more likely to show up clustered than at a perfect binomial independent symmetrical model.
Obviously the very slight propensity to get the opposite event already happened makes a role in that, yet and generally speaking successful random walks do not take into account B and P hand sequences as they're too much influenced by low levels of "statistical limitation".
And, more importantly, no preordered mental schemes based upon too long term findings (for example knowing that P singles and P doubles are slight more than 3:1 favorite to come out than P triples; or that B streaks are more probable than B singles, etc) will help us.

If we want to play baccarat with a possible edge we should understand that "common" stats won't help us too much, otherwise the game wouldn't exist.

The whole EV- picture presents many EV+ spots

See you later

as.

[AsymBacGuy]

The main factor why bac players keep filling casinos' pockets is because they try to get a utopian task to guess longer winning streaks than losing streaks where sure as hell the long term net sum is zero (before HE).

More acute players have studied both the average "prolonging" and "stopping" pattern expectation of what the actual shoe produces in relationship of a "general" probability.
When such ratio strongly differs from a more likely flow of the outcomes (say when long streaks came out) they simply do not bet as such long streaks may easily belong to "math inversion" events.

In some way long streaks should be considered as negative bj card counts where positive situations more often than not are not coming out so easily in the remaining portions of the deck.

The beauty of baccarat is that long streaks (longer than 4) still implement a "finiteness" attitude verified by checking out long term live samples.

Let's consider the issue more technically, so by using numbers:

If a 5/5+ streak accounts for a 5 number, a 4 streak for a 4 number and so on about inferior streaks (up to 2), any shoe dealt will provide a final total number made by a simple arithmetical operation.

Now we want to add the previous streak number with the next number, so for example a 5/5+ streak followed by a 4 streak accounts for a 9 sum (5+4).
At the next streak occurence the last 4 number could transform into 6 (4+2), 7 (4+3), 8 (4+4) or 9 (4+5).

Each sum of two adjacent streaks will form a number succession where some numbers can't be produced.

For example, after a 5/5+ streak happened, the most inferior possible number is 7, then a 8, then a 9 and finally a 10.

Since we have reasons to play towards low sum numbers and knowing that the best scenario will be to get a 4 sum (2+2) and so on (2+3 and 3+2 = 5) or (2+4 and 4+2 =6) or (3+4, 4+3, 5+2, 2+5 =7) and finally (4+4 = , a 5/5+ streak will deny many possible "profitable" spots as sooner or later 10 or 20 or 30 (or more) back to back sum values must happen.

Approaching the issue in another way, it's like a final sum made after those simple rules is going to provide an average final shoe's result.
But anyway getting some peaks, some steady values and some decreasing values.

To help defining when some spots are worthwhile to be bet, we can confide about the relative unlikelihood that 5/5+ streaks will be hugely produced along any shoe dealt.
I've provided a couple of different source samples so you might get a better idea of what I'm talking about.

But what's important to be emphasized is that streaks of precise lenght that didn't appeared so far must not be considered to show up as what we're really interested about is the clustering effect.

After all we're not fighting to get a potential outcome whatever is entitled to show up unless it came at least one time.
For sure "isolated" specific streaks will come out but not for long.

Clustered streak classes to look at

a) First are doubles coming out clustered at least one time.

b) Then doubles coupled with triples.

c) Then doubles coupled with 4s whenever triples didn't seem to show up.

d) Then triples with 4s.

At very rare occasions even mere 4s (so not getting any help from inferior streaks classes, so battling alone vs 5/5+ streaks)

A multilayered betting scheme will take care of those a/b/c/d possible events having the main task to get at least ONE clustered event.
No need to set up a fictional betting strategy waiting for some losses (albeit being wonderful whether we have the attitude to be extremely patient) : what happens still remains more likely to come out again at a given events' category.

Then the same "propensity" will act even at the singles/streaks battle, yet getting more intricate tools to be grasped.

I'm deadly sure you're on the right site to ascertain whether this wonderful game would be beatable.

as.

[KungFuBac]

Thx Asym.
I read all your posts re: grouping or clustering effects.

Can you respond a little more on your ABCD lines above. Sometimes I find it helpful to observe the clustering and run lengths as it helps me identify the "weak link" which in turn tells me the current strong side. The key in my opinion is early detection as just one or two spots earlier winning spots can often mean the difference in our "Win-In-A-Row" length (At least for me).


Continued Success,

[AsymBacGuy]

Hi KFB!!

Sometimes I find it helpful to observe the clustering and run lengths as it helps me identify the "weak link" which in turn tells me the current strong side. The key in my opinion is early detection as just one or two spots earlier winning spots can often mean the difference in our "Win-In-A-Row" length (At least for me).


IMO, you couldn't condense the issue in a better way!! 

Think that basically casinos will lose only when long streaks of something univocally shaped will happen, all other "low deviations" patterns will favor them in a way or another (HE, players' greediness, wrong adaption of the actual results, etc).
Moreover players being favored by such (rare) long univocal patterns will lose very soon what they've earned and we well know that the "quitting when you're ahead" move is a complete worthless bighorn.stuff.

Therefore an acute player should be prepared to know that what constitutes an "easy way" to win (rarely happening) actually is just the fuel to mantain live this game.
In other words, an acute player should selectively bet towards low levels of deviations, being way more likely than the counterparts.

Now, it's sure as hell that low degree of deviations (whatever considered) will come out clustered or not by different levels of probability (0= no clusters, 1= one clustered event, 2=two clustered events, etc).

Then even each clusters streak already classified by a 0,1,2, etc class will come out isolated or clustered and so on.

Wholly considered, it's virtually impossible not to get at least a back-to-back same class category to show up, even considering the 0 (isolated) value.
In fact 0-0 (or 0-0-0, etc) is a cluster.

More later

as.

[AsymBacGuy]

If you want to play baccarat successfully you must imprint in your mind that you'll get an edge ONLY  by getting more wins than losses: trying to erase or invert the HE by adopting progressions is a sure recipe for disillusionments and eventual disaster.

Technically such task can only be achieved by exploiting the "clustering effect" (CE) factor, always working at any shoe dealt but at different levels.
Most of the times the CE provides easy detectable situations, sometimes it'll be too restricted to be exploited, finally some very rare shoes won't make any room to get EV+ spots.

So baccarat is a kind of coin flip succession game only under the eyes of losers (2+2=6-2 forum 'experts of our a$$', for example).

Moreover, baccarat successions are quite dependent of the actual shuffling source and obviously you can get an idea about that only if you know (and get at your disposal) the several shuffling machines casinos will employ to deal bac shoes.

At the end you'd want to play baccarat only if you are able to manage a kind of 4.5 or 5 sigma probability NEGATIVE deviation that sooner or later may come out. (To simplify the issue it's like to face, without getting sensible damage, a 20 or 25 negative B/P streak).

Notice that by exploiting the CE, in the long term a 4.5 or 5 sigma will be more likely to show up at the positive end than at the negative one.
Yet those are very very unlikely situations, after all we're interested about the more likely ones.

The clustering effect moves around more likely steps

In a way or another and considering the common horizontal distribution of the outcomes (rightly displayed at 95% of the casinos' screens), every pattern fills the slots by a long term asymmetrical fashion.
If we'd consider streaks (so not giving a damn about the first row), empty column slots at 2nd, 3rd and superior rows will limit "geometrical" spaces more and more forming long empty shaped sequences where the "empty value" 1 will be more likely followed by superior empty spaces in relative relationship of the row filled.

Providing to set up a decent random walk, back to back 1-1 empty third row successions will be followed by superior empty row classes by a close to 59%/41% probability, meaning that our bets will get an astounding 18% edge (before vig) to win.

Anyway it's important to understand that such edge won't be linearly distributed as 1-1-1 sequences now are 'even money' to get wins than losses and the reason is because many "univocal" events happened somewhat deny a math propensity to show up (being 'consumed'by the actual card distribution).

As long as we're going more deeply at the rows evaluation, empty slots limited by possible or actual "streaky" patterns will form longer empty sequences than 1, but it's up to us to choose the minimum profitable risk to get a win.
And of course, deeper we're trying to get such empty rectangles NOT to be limited by a 1 gap, greater will be our probability of success.
At the cost of missing many profitable opportunities mostly belonging to a undectectable random world we can't do anything about that.

Say that what our algos are interested about (after being instructed to take two different mechanical random walk shapes) is the empty row ranges considered at various levels:

1) When considering the third row, singles and doubles "empty" rows matter only when the 1-1 back-to-back empty row range trigger happens. And it must be played just one time. 

2) When considering the fourth row, we have to get rid of the first row events, meaning we 're taking care of the second and third row vs the fourth (or superior) rows. That is singles are considered as neutral. Here the same 1-1 trigger range concept applies but now we're adding an isolated  superior than 1 situation vs superior clustered events.     

3) When considering the fifth row, we're getting rid of both singles and doubles, so focusing about triples and 4s vs 5/5 streaks empty ranges.

4) Sometimes and for the slight more propensity doubles are showing up, even doubles coupled with 4s present profitable clustered situations, mostly as triples haven't come out so far.

Without any doubt and as already sayed, shuffling machines will make way easier the task to get predictable empty row ranges than expected.

That's the reason why HS rooms keep offering "preordered" shuffled shoes we do not know a fkng about.

See you in a couple of days.

as.

[AsymBacGuy]

Just tracked this live manually shuffled shoe:

B
PPP
BB
PPP
B
P
BBBB
PPPP
BB
PPP
B
PP
BBB
P
BBB
PPP
BBBB
P
BB
PPPP
BB
P
BB
PPP
BB
P

A very good shoe where our algo 1 have found a couple of 5/5+ streaks even though there were none at the Big Road.
Anyway the remaining shoe's texture came out so good for every other pattern the algo would elicit a bet.
And, curiosly, at this shoe there were many.

See you later

as.

[AsymBacGuy]

Our algos constantly work by assessing numbers and not situations: obviously numbers can only derive from situations, that is from B/P sequences entitled to be more likely to happen.

The basic procedure is extracted by simple arythmetic operations made on various patterns where sums may stay put, going back (decreasing their value) or going forward (increasing their value).

Thanks too the Alrelax and KFB fine posts, we've instructed our algos to consider each shoe as a world apart, thus considering it by an "intricate" conditional probability factor moving our random walks into 'predictable' ways capable to erase and invert the HE.

Such task could only be accomplished by evaluating the most likely deviations happening at thousands and thousands of shoes dealt with a special regard dedicated to the strong positive and negative situations' values.

Strong positive and negative situations

To properly work, algos must rely upon verified long term data and the best tool coming in our mind is the ability to get the best of positive events and at the same time the ability to get the possible lower damage at negative events and this can only be achieved by knowing that strong positive events > strong negative events.
Meaning that progressions can't alter normal flows for long, unless the W/L ratio will be shifted in our favor.

But the words "positive" and "negative" must be related to specific situations: technically the issue is resolved by an evaluation of positive streaks lenght and average distribution.

For example, if we'd consider as "enemy" the 5/5+ streaks, we know that shoes producing such streaks are 2:32 against to happen, yet some random walks will make such probability lower than that, so making more room to inferior streak classes to show up.

Same about inferior streaks where the best "streaks trigger" to look for are doubles.

Problem is that many (nearly half of the times) patterns will deny the streak apparition right at the start (singles), so we have to devise random walks capable to concentrate their strenght upon streaks of specific lenght in order to come out clustered.

Once we've discarded the singles apparition, problem focuses about doubles and so on about superior streaks.

Let's take the shoe I've presented above by the simplest Big Road sequence.

No one 5/5+ streaks happened, yet betting towards singles (first step to deny such 5/5+ streak apparition) would have been a lousy strategy as singles came out by this succession (1=isolated singles and 2 or superior numbers=clustered singles):

1, 2, 1, 1, 1, 1, 1.

On the other end clustered streaks (again 1=isolated streak and 2 or superior numbers= clustered streaks) of any lenght had come out as:

3, 4, 2, 3, 3, (3)

Notice that by merging two streak classes inferior than 5 (that didn't happen in this shoe) we'll get such scenarios (1=isolated streak, any superior number = clustered streaks):

2/3 streaks classes: 3, 6, 1, 4

3/4 streaks classes: 11

2/4 streaks classes: 10

Merging such streaks classes will get a general 0.75% probability to happen.
Notice that a supposedly "lower streaks are more likely to happen" general course (that is the 2/3 class) will get us the lesser profitability, whereas 3/4 and 2/4 classes will get us all wins.

Nonetheless, our algo 1 had found out a wonderful 23 winning single/double streak as opposed as to a maximum BR 3 streak and:

2/3 streak classes: 10

3/4 streak classes: 1

2/4 streak classes: 9

Moreover the BR provided 18 streaks and 7 singles and our algo has found out 14 streaks and 20 singles.

Anyway our algos are more focused not to lose than to win multiple hands in a row, so assigning a different value to any spot considered by a 1-2-3 scale where 1 is the less desirable value and 3 the maximum profitable value (0 is a very frequent value not suggesting any bet).

We'll see the actual spots of the shoe provided above where our main algo suggested a 1 (light), 2 (moderate) or 3 (heavy) supposed profitability.

as.

[AsymBacGuy]

All possible patterns derive from a math expectation where card rank positions will make a role in determining the most likely final outcome.

Different ranks help or not a side:

1/3 rank card positions are 9/13 favored to get the B side to win;

2/4 rank card positions are 8/4 favored to get the P side to win;

5 rank card position is 1/1 (even money) to get any side to win;

6 rank card position is 1/1 (even money) to get any side to win.

You must consider such values in order to build valuable algorithms by approximating the key cards falling here or there by more likely ranges.

So itlr you can't expect to win Player bets whether the first or third card isn't a 6,7,8 or 9 (1.45:1 against) but you are 2:1 favorite if while betting the same Player side the second or fourth card is an A,2,3,4 or 10.

If the hand needs one or two third cards (5 or 6) the hand will go even money.

Some random walks are more capable to grasp the "more likely card distribution" considered by certain patterns, especially by exploiting the natural 'clustering effect'.

And that's a huge edge we could rely upon.

"Incidents", that is hands not following a math propensity, are surely happening but they should be considered just as a kind of systematic error not influencing our long term results.

See you next week

as.

[AsymBacGuy]

We know that baccarat BP sequences can't be solved mathematically (actually it's possible but by getting insignificant long term values) so we're forced to study other ways and we've found particularly valuable to consider BP successions by "ranges".

To be really worthwhile a "range" or "ranges" must move around finite and slight dependent factors affected by a constant asymmetrical strenght (due to a unequal card distribution).

So in some way baccarat BP sequences are reproducing "biased" successions where the B slight math advantage won't make any substantial effect in order to gain a possible edge over the house.

Anyway and since we're talking about ranges, we should know that the edge will be influenced by huge variance, so we must take many countermeasures to avoid it (or to try to get the best of it when positive), a thing apparently easy to do in theory but very difficult to put in practice.

More later

as. 

[AsymBacGuy]

A "range" is formed by one or more patterns delimited by "boundaries" making such pattern(s) to stop at some point of any shoe dealt.

The beauty of baccarat is that sometimes (say not so rarely) those boundaries do not happen at all, so leaving plenty of room to extract value upon those clustered pattern(s) situations not crossing  any losing spot.

It's altogether obvious that boundaries formation is in direct relationship of the general probability of success, meaning that it's impossible NOT to get boundaries (streaks) at singles successions (whatever considered), very very very very very difficult NOT to show up at least one time at double patterns (battling vs 3/3+ streaks boundaries), and so on with triples vs superior streaks or 4s vs 4+ streaks.

Since our algorithms are instructed to put a kind of "limit" to the realm of randomness, 5/5+ streaks are the maximum value where they stop to consider ranges. The same way insurance companies put limits to possible compensations after customers had payed the insurance fee.

In some way acute baccarat players like to get a kind of "insurance policy" that some very rare events won't come out short-gapped or, more importantly, by numbers not giving room to way more likely events considered by ranges.
In addition and since the main aim will be always oriented to get inferior patterns classes to be clustered at some point, back-to-back boundaries (in our example 5/5+ streaks) won't get us any damage as no inferior pattern can come out when two or more adjacent 5/5+ streaks happen.

Knowing that the 5/5+ streaks appearance is well limited per any shoe dealt (especially and foremost by utilizing specific random walks), a multilayered progressive betting scheme joined with a super selected betting will crush every casino in the world.

So instead of thinking about the missed profitable opportunities when things seem to come out confused, we should focus about how many losing spots you'll get rid of by waiting for superior ranges to show up clustered and by passing more undetectable inferior ranges.
After all whenever we won't bet a fkng dime, casinos will someway hate us and when casinos hate us is because we're taking the right side of the operations.
So when in doubt do not bet a fkng cent: more hands we want to guess, greater will be the HE.

Technically is just challenging that bac productions won't make 2/3 streaks or 3/4 streaks or 2/4 streaks NOT coming clustered "for long" at least one time.

Since 5/5+ boundaries are well defined in their sd values (at least by adopting some random walks), each losing step will be more likely followed by a kind of clustering inferior streaks propensity capable to erase and invert the HE.
By a 100% accuracy.

as. 

[alrelax]

"A "range" is formed by one or more patterns delimited by "boundaries" making such pattern(s) to stop at some point of any shoe dealt."

SECTIONS???

"The beauty of baccarat is that sometimes (say not so rarely) those boundaries do not happen at all, so leaving plenty of room to extract value upon those clustered pattern(s) situations not crossing  any losing spot."

BUT-THEY DO WITHIN NEW SECTIONS.  A PERSON JUST HAS TO BE ABLE TO DEFINE SECTIONS, WHICH WAS WHAT I WROTE AND DEFINED——WHAT SECTIONS ARE, HAPPENS AND ACTUALLY HAVE TO OCCUR IN EACH AND EVERY SINGLE SHOE PLAYED.

"Knowing that the 5/5+ streaks appearance is well limited per any shoe dealt (especially and foremost by utilizing specific random walks), a multilayered progressive betting scheme joined with a super selected betting will crush every casino in the world."

ANYTHING AND EVERYTHING HAPPENS AS WELL AS, NOTHING EXPECTED OR DESIRED OR ENVISIONED MIGHT AS WELL HAPPEN.

NOTE:  But the bottom line in actual wagering at a table of bac, is the ability to play with total consciousness, focus and enough buy-in to have a decent drawdown.  As well, if you depleted your buy-in, risk capital, you must leave and chalk it off.  Why?  Because then your emotions and frustrations will 100% set in.  The previous said is not fallacy or my opinion, it is multiple decades of experiences and facts.

ADDITIONAL NOTE:  Online testing is completely different than real table game wagering.  Online testing is either one of two results.  One, you eventually come out winning.  Two, the program is enticing you to signup and deposit real funds, etc. 

[AsymBacGuy]

Hi Al!
Thanks for your fine comments.

Sections

Yes, we could use this word.
Specifically by "range" I mean the various distribution of empty slots at the different rows (horizontal registration).

BUT-THEY DO WITHIN NEW SECTIONS.  A PERSON JUST HAS TO BE ABLE TO DEFINE SECTIONS, WHICH WAS WHAT I WROTE AND DEFINED——WHAT SECTIONS ARE, HAPPENS AND ACTUALLY HAVE TO OCCUR IN EACH AND EVERY SINGLE SHOE PLAYED.

I agree, anyway algos can only move by objective findings (single gaps, streaks lenght, consecutive or isolated streaks, etc)

ANYTHING AND EVERYTHING HAPPENS AS WELL AS, NOTHING EXPECTED OR DESIRED OR ENVISIONED MIGHT AS WELL HAPPEN.

Of course, but always by specific levels of probability. And our efforts are always directed to spot "more likely" ranges with a special regard of NOT falling into those very rare negative (but possible) distributions.

NOTE:  But the bottom line in actual wagering at a table of bac, is the ability to play with total consciousness, focus and enough buy-in to have a decent drawdown.  As well, if you depleted your buy-in, risk capital, you must leave and chalk it off.  Why?  Because then your emotions and frustrations will 100% set in.  The previous said is not fallacy or my opinion, it is multiple decades of experiences and facts.

100% true.
That's why we leave the bet selection to algos that are totally insensitive to emotions and frustrations.

as.

[AsymBacGuy]

Let's trace some technical elements again by considering streaks.

Streaks are patterns made by a multistep probability floating around the "back-to-back" same side (widely intended) apparition.

Nothing wrong by limiting the bac outcomes into streaks (so ignoring singles).
Then all possible streaks will be classified within the 2-5 range, so considering irrelevant all streaks superior than 5.

That's the range our algos are interested upon.

Now algos want to be instructed about the "maximum positive or negative value" every streak class (2, 3, 4 or 5) will appear per every shoe dealt.
Of course not giving a damn about previous shoes as each shoe is a world apart.
Say we'd assign a correspondent number to every streak belonging to the 2,3,4 or 5(5+) class.

If we'd sum up the two last streak numbers we'll get those three situations:

1) The value remains still (for example a 2-3-2 streak succession (sum=5) or a 2-2 (sum=4) or 3-3 (sum=6), etc. 

2) The value will increase (any streak followed by a superior streak)

3) The value will decrease (any streak followed by an inferior streak).

Obviously not every situation will make the next sum belonging to every different category.
For example a 2 streak apparition must only produce an increasing or still sum.
3s and 4s will make any scenario possible and 5(5+) cannot increase their value (either they stay put or decreasing the sum), a banal specular situation happening at 2s.

Therefore we might think about how are the best and worst possible events making such sums to be decreased (2s if no 2 happened) or 5(5+)s (if no 5-5+ streak happened).
In addition, we want to get at our favor all the possible situations making a still sum (so a back-to-back same streak apparition).

The luxury tool we rely upon is that 5/5+ streaks are well determined in their apparition (that is by their density average apparition along any shoe dealt), 3s and 4s streaks are very common to show up and itlr doubles are the most likely streak shape any BP distribution will provide.

In a word, streaks distribution (providing a proper random walks action) will make more probable to get decreasing or still sums (5/5+ streaks aside) of two adjacent events than increasing values.

That's just a general propensity that must be evaluated by how the actual shoe is doing.
In fact most of the times sums are in direct relationship of the previous specific streak classes appearance, in the sense that we do not want to chase doubles when no double had come out so far and at the same time we must always be prepared to face the inevitable 5/5+ streaks erasing any decreasing or still sum (yet at an interesting portion of the shoes they won't come out a single time!).

Putting things into numbers

Since we have learnt here that it's way better to chase the model NOT to provide expected numbers (or situations) at two consecutive betting steps, we should assess how many decreasing/still/increasing sums will happen along any shoe dealt.
Obviously by betting (or fictionally betting) two situations out of three (when applicable), that is wagering towards still or decreasing sums, we'll get a better idea about how bac things work itlr.

Let's take the above presented shoe registered in real time at a online site.

As already sayed, we're just considering streaks as numbers.
First by the common Big Road sequence, then by our main algo and finally by our backup algo.

1) BR sequence

3,2,3,4,4,2,3,2,3,3,3,4,2,4,2,2,3,2.

Sums are 5, 5, 7, 8, 6, 5, 5, 5, 6, 6, 7, 6, 6, 6, 4, 5, 5.

2) Our main algorithm:

3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 5, 5, 2.

Sums are: 5, 4, 4, 4, 4, 4, 4, 4, 4, 7, 10, 7.

3) Backup algorithm:

2, 3, 2, 3, 4, 5, 5+, 4, 4, 2, 3, 3.

Sums are 5, 5, 5, 7, 9, 10, 9, 8, 6, 5, 6.

This shoe (bad manually shuffled) was pretty good as no "boundaries" (5/5+ streaks) happened at BR  sequence; moreover at our main algo registration the couple of 5/5+ streaks were fortunately coming around clustered giving plenty of room to inferior streak classes to show up (here by a consistent clustered doubles appearance).
Backup algo (despite of crossing just one 5/5+ streak, went more badly as most sums did increase their value than lowering it). 

But it's not a coincidence that the main algo will get way better results than the backup one.
Anyway and putting the main and backup algos into the decreasing (D), still (S) or increasing (I) sums (stopping when a 5/5+ streak happened and waiting for an inferior streak class to show up) we got:

main algo: D, S, S, S, S, S, S, S, S, I (stop)

 
backup algo: S, S, S, I, I, (stop), D, D, D, D.

Just for curiosity let's see how the BR succession performed:

S, I, I, D, D, S, S, I, S, I, D, S, S, D, I, S.

Notice how different went the three different successions but focus about how's easy to spot what are the most likely occurences to look for.

Now we are quite sure best ATM in the world are casinos offering baccarat tables (at least as long as the global warming effect won't cancel the human species from this planet, and unfortunately this thing will happen very very soon)

as.