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.

Al wrote: Same thing as the scoreboard.  Follow it this shoe and win and then the next 15 shoes total unpredictability or un-followability!


True, because you are taking one side of the problem. That is the actual single shoe presentation and more often than not it's just a permutations issue that makes us losers or winners.

Example. We're betting toward singles and doubles, triples are our "enemy".

Before considering an actual production let's see how many triples on average will show up per every single shoe.

Answer: say the triples number per shoe (obviously in relationship of how many cards are really allowed to be dealt) is 9.5 .

How many columns are showing up on average per shoe? Say 38.

38:9.5 = 4, and that's the expected value applied to a binomial independent production, that is on average per every triple we'll get three singles and/or doubles patterns. 1/3.

Now we have to set up a strategy based upon this feature, well knowing that such ratio could be easily disregarded at either way. But for a moment assume it remains constant per each shoe dealt.

Say A= singles and doubles and B=triples. 

An ideal world would be to get a presentation pace corresponding to the expected ratio:

AAABAAABAAABAAAB....or BAAABAAABAAAB...

Apparently this is a perfect symmetrical production but in reality is not as A includes two different patterns considered as equal (singles and doubles) and B a vast category of various streaks surpassing the 2 point (triples).

Other productions getting the same "perfect" 1:3 A/B ratio might be:

BBABABBBAAAABBBAAAAAAAAAAAAAAAAAAAAAAA

or

AAAAAAAABBBBBBAAABBAAAAABBAAAAAAAAAAAAAA

We notice that playing toward A in the former example would lead to a total disaster, the latter example lead to the same consequences at the six B streak.

Finally we see that in such extremized (but possible) examples the number of shifts is very low: 7 and 6.
And we're still talking about a perfect 1:3 ratio....
Let's imagine what could happen when B instead of being 9 or 10 is 13 or 14. So proportionally lowering the number of profitable A events.

Obviously we could think about the opposite favourable situation, that is a number of A way greater than 28.5, say 32 or 33.

We can't know if the actual shoe will be rich or poor of A or B, but we do know that the vast majority of shoes dealt will approach more and more to the 1:3 B/A ratio with all the permutations varieties. 

Summary

If the B/A ratio nearly stays in the 1:3 field, it's virtually impossible to lose. Providing to take care of the permutations issue.
When the B/A ratio is quite higher than expected it's more probable that even the permutations issue doesn't make a substantial role in spotting favourable situations. And vice versa when B/A is quite lower than expected.

Put the probabilities into numbers and check your shoes and you'll see that shoes producing more than 11 or 12 triples are quite rare.
More interestingly is the fact that triples are less likely to come out consecutively for long time than what singles and doubles could by considering a proportional unit loss or win. A mere permutation issue.

In a word, it's way more likely to encounter a 21, 23 or higher single/double consecutive streak than crossing a proportional 7 or 8 consecutive triples streak.
That means that the average card distribution makes more probable to get a greater number of consecutive side shifts than the proportional consecutive triples counterpart.

So when the B/A ratio tends to go too far from the 1:3 ratio and assuming those scenarios have the same symmetrical probability to appear, B clusters will be proportionally slight shorter than A clusters.
Therefore 'bad' could happen even for long, but 'good' could happen for longer.

Anyway, algorithms need more precision to forecast outcomes; they can't accept to consider a 3 or a 11 streak belonging to the same "triples" category.
And at a lesser level, even singles and doubles cannot constitute a "same" class.
So they ought to register every single hand.

But the basic process remains the same.

as.

Hi KFB!
Even if an existent app would work (and I highly doubt it) there are no many Stadiums around the globe to use it.
Then casinos via eyes in the sky are able to scrutinize winning players even at Stadiums, especially if they use a cellphone. So it's quite likely that as long as "app" players are not consistent winners no heat is going up.

Another problem with Stadiums is that we have very few time to place our bets and it's more probable to make mistakes as dealer doesn't wait for our action.

Finally it seems we all agree that card tracking isn't the way to make 'better than average' choices at BP hands.

as.

Algos work because they try to estimate the relationship between the actual card distribution and the "average card distribution".

Players "guessing" doesn't work because it tends to get rid of the "average" privileging the "actual".

On the contrary, "triggers" players do not win either as they put too much emphasis about the "expected" instead of assigning a proper value on the "actual".

Of course algos are instructed to know very well how to deal with those rare unfortunate or lucky situations touching or surpassing a 4 or greater sigma probability.

Another important issue is that algos cannot care less about the actual bankroll status or about previous results: that's a pivotal aspect completely diverging to any method, system or approach invented to beat this game.

As already sayed, algos can only fear a "short term" unfavourable permutations issue that can easily overcome by the use of two simultaneous algos applied to the same sequence.
Or by calibrating the main one under stricter parameters when positive clusters tend to be silent for long.

We should remember that the aim of playing baccarat is to make money and not to have fun or collecting comps.
And algos will make their best efforts to exploit symmetry/asymmetry of the deck not visually determined, otherwise it would be a too simple task to accomplish.

More later.

as.

As always a good post from Al.

I concur with KFB about your #1 point: yes, we're obsessed to win no matter what, it's just a matter of time we'll get the full power of our advantage.

#2: Yes, definitely you're right. In the vast majority of the times we players are going uphill but sometimes the road descends so steeply that we couldn't care less about the normal road unfavourable conditions.

#3 and #4: Perfectly spot on!

as.

I'm very glad you enjoyed this book.

When talking about gambling I like to hear from the best of the best and in this field Billy is the GOAT!!!

as.

'Runs' are classified by the number of shifts of one side to the another one.
For example a BPBBPPBPPPPPPBB sequence forms 6 runs and the same concept applies to every A/B binomial model where we assign a given status either to A or B.

The average runs number of a given production is one of the simplest way to assess randomness without calculating it by other statistical parameters as by using the chi square test, for example.

The state of art of baccarat has taught us that BP hands are moving around unbeatable values as B and P results are corresponding to a kind of a constant undetectable EV- coin flip succession.
That is a random undetectable world.

That's a complete rattlesnakesh.i.t: if this statement should be true, any bac succession should confirm the perfect place selection independence, in the sense that no matter which spot we select to consider (or bet) outcomes will invariably follow a general (unb) probability line.

To get a better idea of what I'm talking about say we're deadly sure that the card distribution of every shoe dealt will be asymmetrically shaped.

Now the sure asymmetrical card distribution will entice either asym patterns or sym patterns.
And at both categories different densities of those opposite scenarios will reach more probable values.
When some cutoff values are touched, we're not interested to know whether things will change or not, we simply let results to flow without betting a dime. 

Algorithms just exploit the asym/sym various AVERAGE densities happening along any shoe dealt and as already sayed the main algorithm will get the lead over the backup algorithm because A algo will catch a slight more profitable situations than the B algo, as the former consider a slight more complex rhythm than the other one.
For example giving a lesser impact over the moderate homogeneous streaks but relatively suffering of other strong symmetrical but unlikely sequences favoring the B algo wins.

as.