Well, how can I disagree with what you have written...

as.

If we'd think to play with an edge, two big enemies must be overcome in order to properly exploit it.

1) Betting few hands, that is selecting at most our betting options.

2) Being totally insensitive of the actual session's bankroll status, knowing that any edge needs time to get its full power.   

No surprises that by disregarding the point #2 even the point #1 will be invariably affected, meaning that when we're losing we'll try to find reasons to unnecessarily bet too many hands with the effort of breaking even.

That's why we advocate a basic flat betting procedure where W/L streaks and W/L spots aren't equally balanced within short frames.

So when you'd think to raise your bet consider the worst scenario (losing a lot more) than the hopeful situation capable to erase some losses by a lesser number of wins.
It's true that no matter what are the betting amounts we will place, once we play with an edge everything will come at our favor.
Yes, but we cannot be 100% sure about the current conditions making profitable our plan devised at different environments.

Then the house and besides the HE has at its disposal way more money than us, so serenely willing to accept a passive mood vs our wagers oriented to get an advantage by mere statistical reasons.

Probably the most wanted action any casino in the world would like to face is getting players betting every hand and raising their wagers with a kind of "sky's the limit" attitude.
So, in some way and generally speaking, our best action would be to play very few hands and always betting the same amount.

Next topic is so complicated I prefer to present it the next week.
Sorry.

as. 

I'm glad a real expert on this subject as KFB agreed on this point!

Of course I'm not advertising random.org site, just taking it as the best probable source to get real random card successions worthwhile to be compared with what you're facing in the real world.

When deep money is involved we've learnt not to trust anyone or anything even though it's very likely that most people on the "other side" are unaware of the randomness topic (let alone the players...), so relying on formulas and/or math laws, etc.

In our opinion collecting large samples of real random successions is the best way to get an idea about how things develop in the terms depicted above.
It's, again, quite probable that unless a software instructs the cards to be distributed by a perfect random source, what we played so far in a way or another are just unrandom shoes where Alrelax posted many valuable thoughts for them.

In any instance, we keep thinking that random production or not, different patterns move around more likely ranges where a large part of situations are completely worthless to be wagered, remember that at baccarat we're constantly going uphill for the HE.

Setting up a strategy about 'ranges' means to approximate at best of our human possibilities the various patterns apparition and lenght with the old rule of thumb that what happened so far could be easily considered as not existent (especially at unrandom shuffled shoes), yet many valuable and relatively more frequent spots will incorporate a higher edge as 416 cards tend to be distributed by quite detectable levels of clumping.

More later

as.

Consider baccarat as a world formed by three different pattern models:

a) Homogeneous patterns (e.g. long 1-2 sequences, long chopping lines, long consecutive streaks, etc)

b) Heterogeneous patterns (patterns that continuously change their shape by a very high frequency)

c) Patterns not belonging to a) or b) categories.

At the vast majority of shoes dealt c>a+b as patterns are more likely to get levels of probability roaming around "intermediate" levels different than 0 or 1 (case b) or getting huge values (4, 5, 6 or more, case a).

Most players like to bet towards the a) class, other players prefer to chase the b) scenario (patterns standing for short sequences); almost nobody but the very rare serious bettors have learnt to set up their plan upon the c) category.

Notice that we're not criticizing to ride the occasional a) successions where we could only lose one bet but collecting several wins, yet for sure the b) situation (unless the player is particularly super skilled) is the worst to set up a strategy upon.


Let's take a look a bit further by a "consecutiveness" point of view.

- a+a situation could happen, a+a+a situation is very very unlikely to happen.

- (b) situation is a kind of consecutive situation by definition, so it's way more probable to get it mixed by (a) situation or, more likely, interspersed by a (c) situation.

- (c) situation, the strongest on average, tends to constitute the "norm", so itlr c-c>c-a, c-c>c-b.
Once c-c happened, the c-c-c sequence starts to get more volatile values belonging to the classic gambling world (no edge).

We see that in a way or another we had to empirically set up a limit to those a, b and c scenarios transformed into numbers. All unsound math considerations that any math or stats expert could dispute and, frankly, that we couldn't care less about.
After all we are just approximating probabilities at our advantage with the aim of winning money and not disturbing the casinos' knowledge about the baccarat invulnerability.

Anyway there's a decisive factor to be ascertained, that is the real randomness of the card distributions we'll have to face in the real world.
On that important subject we think that it's an intricated task for us to assess whether a shoe is really randomly shuffled or not, fundamentally as real random sources cannot be implemented so easily in any card distribution.

Try to run a 416 finite card distribution at random.org site (considered the best), register the outcomes by baccarat rules and take care of the a, b and c situations happening by a consecutiveness and average levels of presentation point of view.

Probably random.org site is the best source to get an idea about how things will distribute by a perfect (or at least, nearly to perfect) random factor.

By comparing random.org successions with real card distributions happening at your table (same table, same conditions) you'll get a better idea about how real randomness is supposed to act or not.
So when in doubt do not bet anything or bet what's happening whenever b and c situations have surpassed their cutoff values.

as.

Basically at a game where is so easy to win in short terms and practically impossible in the long run, the most important skills come from experience, that is learning from the mistakes we continuosly make in our sessions and in our tests.

HE impact interferes very little with our plan once we properly run infinite times similar betting situations, where of course everything will tend to be equal or almost equal.

For example, one of the worst mistakes to make at baccarat is quitting the session after having collected more wins than losses, a thing particularly frequent and made by players who like to hugely modify the betting amounts, so not taking with the proper consideration the inevitable WL permutation issue.

More later (hoping no site maintenance work will go on)

as. 

Yep, your "Sections" tool is a nice idea to try to get a better picture of what is going on.
We adopt the same concept by assessing consecutiveness of certain outcomes.

Problem, obviously, arises whenever any kind of pattern assessment seems to fail (no long 1-2s sequences, no moderate/long streaks, no one side predominance, etc)
And altogether obvious is the fact that casinos will prosper a lot about such natural patterns being by far the most likely occurrence itlr.

That's why we came to the conclusion that betting at an EV- game needs to choose the same route taken by casinos: long term quality will overcome long term quantity; math edge is just an additional factor to win money from customers and not the principal cause.

Experience

Differently than NL hold'em poker, for example, very young players have a 0 probability to win at baccarat itlr.
Aggressiveness here is the sure recipe to go broke or to win a lot but just within short terms.

The best bac players are people who have experienced thousands and thousands of shoes and I mean "real shoes" and not using strategies obtained by simulated shoes, no matter how's sophisticated the software's production.
We've seen that it's very probable that bac productions will feature different levels of randomness and probability, so any mechanical approach disregarding such factor is destined to lose.

Of course a possible unrandomness or a kind of "undetectable randomness" needs a careful study in relationship of the actual conditions that do not necessarily link with the current patterns.
That's why something it's advisable to stay put for many hands (or even shoes) or at the very least to play some hands by an opposite line of what we've found to be profitable in our findings.

More later

as. 

Quality remains a totally ininfluent factor anytime the production is indipendent and perfectly 50/50 shaped for each hand, a thing that can't happen at baccarat no matter how cards are arranged.

Once 312 or 416 cards are shuffled into a shoe, a strong bias constantly acting towards more likely ranges is set up at the start, meaning that each succession is affected by a kind of unrandom element.

Such statement is supported (albeit taken at different issues than baccarat) by studies made by eminent math experts as RVM and MvS.

oOoOo

A bias could show up by "natural" features (natural variance) or defects of shuffling, what is important to understand is that a bias doesn't always act towards LONG sequences of something but towards PREDOMINANT things of something.

In a word, that most of the times such a bias is quite balanced along any shoe dealt, meaning that what was won in the past will tend to get more losses than wins in the next successions.

Many times this assumption means that what many call as a "chaotic world" is just a bias that naturally go in the other direction.

The common trait helping us a lot is that the famous overalternating mood is the least likely to happen.

So even it might appear as an unsound approach, sometimes there are reasons to apply the reversal strategy, of course only whenever a preferred strategy stopped to work.

Once the "overalternating" opposite patterns betting had reached the 2 or 3 value, we'd better stay still or wagering towards the last side being clustered at least one time.

That's one of the best strategies I can suggest here, perfectly fitting the Gambler's Fallacy law where it's impossible to guess this or that by assessing previous outcomes.
So we can safely assume we're considered as clowns by casinos.

as.

There's no uncertainty when you look for patterns from a quality point of view

No matter how's random or unrandom the production we're facing, the quality performs way lower levels of variance than quantity. 

Quantity is beloved by short term players, quality remains a factor obviously needing more time to be properly ascertained in its inevitable fluctuations.

It's very likely that the few people making a living at baccarat prosper about "quality" parameters, making the  qualities "lenght" just a second-step possibility.

So basically the first thought shouldn't be focused about how long a profitable pattern will happen but WHEN a given pattern would be supposed to show up in relationship of the previous patterns and in constant relationship of the "expected" patterns.

Anytime we'll place a bet we must confide on the probability that patterns stop or show up by values capable to erase or better invert the HE; then the "quantity" factor remains just an additional element.

More later

as.

I enjoyed reading your thread.

Very often players forget that dealers are there to work so deserving the utmost respect.
In our opinion the vast majority of bac dealers are excellent employees even though there are some rare exceptions.

Now I'm looking forward for a post about floormen, Al!

as.

Hi Bally!

Below are just personal thoughts.

The waiting time is one of the most important concept to implement in every bac strategy, KFB knows very well this EC feature.

About doubles we know that 2s are the most probable pattern happening at baccarat, but long term quantities are strongly affected by a "quality" factor.
In a word, doubles are predominant as an overalternating mood is slight less likely to happen and, of course, long consecutive streaks at baccarat are relatively uncommon unless several "incidents" happen along the course of every shoe dealt.

Simplyfing, it's the clustering effect that matters.

as.

A couple of good links, Al!

We've extensively wrote a lot about it...

When a slice of pizza costs $12, a cocktail $25, a T-shirt with a simple "F1" logo $80, no free parking available, etc...well maybe people will look for better places to spend their money.

Where are the famous "Vegas prices" now?

as.

Welcome back!!

as.

RFS is another way to consider streaks lenght, more specifically how rows #2, #3 and so forth are more probable to jump in a new column, that is stopping the "slowing down" CFS.

Think that most of the times, long streaks (say longer than 3 or 4) are the by product of "coincidental factors", so denying a kind of relative general propensity to stop.

We've learnt at our expenses that per any shoe dealt the room to get a general propensity is a limited factor acting in the same way as the probability to get asymmetrical hands favoring Banker.

Short streaks are the rule, long streaks are exceptions.

But whenever long streaks happen at the shoe we're playing at, we should apply a kind of "quality" factor that basically could be translated into the propensity to get a specific streak lenght to be isolated or clustered.
Obviously we'll take care of the most probable streak situations, that is doubles, triples or 4s.

Once a streak had surpassed the 4 streak point, more often than not we're not interested to make predictions, unless we want to gamble a previous robust profit.

In our opinion, wagering towards long streaks needs a very long experience and acute thought, so we don't recommend it as a viable plan to make money at baccarat itlr.

Distribution of the most likely streaks

Each specific lenght streak (say doubles, triples and 4s) fight against superior streaks (so singles are ignored): the basic quality every specific streak will feature is the "isolated or clustering" parameter.

That means that we need one specific streak to show up before thinking to make any bet.

2-3 attack (that is wagering to get at least a 2 or 3 streak after a 2 or 3 streak happened) is the basic approach generally denying a long "overalternating" distribution in terms of I (isolated) or C (clustered) patterns.

Therefore the "overalternating" fashion (I-C-I-C.. or C-I-C-I...) is the slight least to happen, in the sense that sooner or later a I-I or C-C will show up.


3-4 attack, despite of needing more hands to be dealt, is even more "precise" as 3-4 clusters are way more probable than 3 or 4 isolated streak occurences.

An important rule of thumb to follow is that what didn't appear so far should be considered as "no existent".
On the other end, once a streak superior than 3 or 4 happened, we have to be more cautious about our wagering, a thing that might entice us to wait for TWO 2-3 or 3-4 patterns before placing a bet.

 
How long to ride a 2-3 or 3-4 I or, way more likely 2-3 or 3-4 C pattern.

2-3 attack is more probable to come out but suffers of more volatility than the 3-4 attack; a general guideline at both cases is to assess the common "3" streaks parameter.
Whenever a 3 streak didn't happen so far, a two-step betting isn't indicated, so leaving more room to the 2 or 4 streaks being clustered (or, more unlikely isolated).

Anyway never ever bet whenever a 4 streak (attack 2-3) or 5 streak (attack 3-4) happened.
Let a 2/3 or 3/4 pattern to show up and act accordingly to what happened in the previous part of the shoe.

As long as a "enemy" won't show up, it's not wrong to keep betting especially when the enemy hadn't come out once.
At any rate, the most likely and profitable situations come out after one step of cluster or, less likely, after one step of isolated pattern.

as. 

The number of shifts is another way of considering the CFS, but there's another factor to take care of, that is the RFS (rows filling speed) that starts to act after two same sides happen as the row #1 is 100% filled by definition.

More later

as.

Start with a general consideration that any WL sequence will get an average number of shifts (W changes to L and L changes to W).
In our example (where the BP real betting options were "concentrated" into the W/L form) we got 12 shifts within a 24 WL outcomes succession.
Moreover out of 12 shifts, 8 were included within the 1 or 2 step ranges (that is 2/3 of the times a univocal W or L sequence will stop after one or two W or L).

This simple consideration (of course completely worthless at independent models) will help us to define the expected average number of consecutive wins or losses: once such a homogeneous W or L pattern had surpassed the level 2 (two consecutive wins or two consecutive losses) we are relatively not interested to chase the W or L side anymore, so waiting for a new shift to happen. 

It's particularly important to understand that the number of shifts roam around average values but consecutive W or L successions could take strong deviated lines, so denying the aforementioned average shifts number.
That means that sometimes the "expected" will make more room to the "actual", so we've chosen to either let the shoe without betting or, maybe, to ride it by a kind of "reversal" fashion.

Since we have theorized that symmetry (at least under our empirical guidelines) tends to succumb to asymmetry as more often than not is a by product of "coincidental" events, once a symmetrical pattern had come out clustered more than twice (that is producing a LLLL or longer sequence), we'll have reasons to consider such shoe as a kind of "abnormal" shoe. That is not fitting the average requisites, so considering it as unplayable or worth to be chased by an opposite line of thought as long as no wins towards the asymmetry show up.

So in some sense, most part of asymmetrical patterns will show up either heavily clustered or not happening at all (or quite dispersed in limited ranges of apparition), so conceding a greater than average probability that asymmetry will be denied right at the first level of consideration.

See you in a couple of days.

as.