Al,


I usually stick to the ECs in conjunction with RTM, and there's a reason for that. Betting one or two numbers (or even 3 or 4) would just take too long to find the rare events. For example, worst-case scenario for a single number could be 600 spins or more. Time is money, and not only that, boredom would a factor, and when you get bored there's a tendency to make stupid bets. A lower variance bet is more efficient from my point of view, and ECs hit the sweet spot ("odds on" bets like two dozens are just too dangerous because even a short losing run can cost a lot, and higher variance bets are too slow).


In principle, any bet can be used because the logic is the same. One approach I'm currently experimenting with is to track repeating numbers. The random variable in this case is the number of spins it takes for a repeat in the last X spins.Track the last say 5 spins on a rolling basis, then when the number of spins reaches 3 SD+, bet the last 4 spins (you only need to bet 4 because you're banking on a repeat in the last 5) again on a rolling basis. This is a simple idea, but it can be extended by simultaneously tracking the last 2,3... up to 8 or 9 numbers for repeats (each will have its own standard deviation) and then looking for correlations (i.e., did the number which is a repeat pair with a number which is 2,3...8 or 9 spins back? and what the is mean and standard deviation of this random variable?).

Bayes,
      Say I chose to bet all three EC pairs in euro roulette and trigger being a z-score of -3.0. Do you feel such triggers will suffice to earn flat bet? How many bets should we place after that?

The question is how you do this in real life.

One loss is one event and one win is one event.
One loss has the value of 1 and one win has the value of 1.

So 40 loses and 6 wins as Bayes mention is 5.06 STD

One red is one event and have the value of 1.
One black is one event and have the value of 1.

As i understand it, so does Bayes software using patterns of certain length.
Then the software match or miss-matches does with the real results to find the worst or highest STD.

So how do you do that in real life in real casino?
No software and only pen and paper.

I think i have the solution for that using predetermined window of events (random bits).
Then the benchmark is set before you start playing.
Could be Bayes example with 40 loses and 6 wins or 5.06 STD ...

But it would be interesting to hear what Bayes say about the subject.
How would you charting by hand using RTM methodology.

Quote from: Bayes on December 22, 2013, 09:34:29 AM
I do need to use a progression (albeit not an aggressive one), and the problem with progressions, as we all know, is that they can win very often but when you hit the run from hell they can wipe out all your winnings and then some. So I let the wins take care of themselves; my business is to cut down the length (not so much the frequency, which doesn't affect my profits, only the time it takes to get them) of the losing runs.

Can you please elaborate concerning your progression how and why frequency of hits is less important then length of streak and how do you mean you let wins take care for themselves?

Drazen

Quote from: Sputnik on December 22, 2013, 11:13:03 AM

As i understand it, so does Bayes software using patterns of certain length.
Then the software match or miss-matches does with the real results to find the worst or highest STD.

So how do you do that in real life in real casino?
No software and only pen and paper.

I think i have the solution for that using predetermined window of events (random bits).
Then the benchmark is set before you start playing.
Could be Bayes example with 40 loses and 6 wins or 5.06 STD ...

But it would be interesting to hear what Bayes say about the subject.
How would you charting by hand using RTM methodology.

Not sure what you mean by "predetermined window of events". If it's predetermined, i.e., it doesn't use any history, then you run the risk that your predetermined sequence (or the opposite of it, depending whether you are betting for or against it) may show up, so to my mind this is no better than betting randomly.


Another option which could be used in a B&M casino is the old technique of betting the opposite of whatever sequence has just appeared. Norman Squire has a section on this in his book "How to Win at Roulette". He suggests betting the opposite of the last 37 spins, the rationale being that it would be virtually impossible for the exact same sequence to repeat immediately (You could suppose that you had been betting that same sequence over the last 37 spins, in which case you would have had 37 losses in a row). The flaw in that theory is that the sequence needn't repeat EXACTLY in order for you to still get substantial losses, and the longer the sequence is, the more likely you are to get bunched losses, which is what you're trying to avoid in the first place. I've tried this method in the past, and it isn't very effective.

Quote from: Drazen on December 22, 2013, 11:14:05 AM
Can you please elaborate concerning your progression how and why frequency of hits is less important then length of streak and how do you mean you let wins take care for themselves?

Drazen

The kind of progression I use is such that I will break even with something like LLWLLLWLWLLW which has twice as many losses as wins. So it doesn't matter how long this pattern continues because I won't LOSE (I won't win much either, but as I said, I'm not bothered about that). In that sequence, the losing runs are short between wins. Occasionally I win for quite long periods flat betting, but I never "chase" wins using positive progressions. If I'm on a winning streak, I'll ride it as long as it continues and break off after one or two losses (depending on how much profit I made). Long winning streaks aren't that common though, more often outcomes are choppy which is why positive progressions are dangerous.

Bayes

Would you please show in the above eg  LLWLLLWLWLLW  how you would bet using monetary amounts thanks.

Cheers

R

Quote from: Bayes on December 22, 2013, 01:36:43 PM

Not sure what you mean by "predetermined window of events". If it's predetermined, i.e., it doesn't use any history, then you run the risk that your predetermined sequence (or the opposite of it, depending whether you are betting for or against it) may show up, so to my mind this is no better than betting randomly.


Another option which could be used in a B&M casino is the old technique of betting the opposite of whatever sequence has just appeared. Norman Squire has a section on this in his book "How to Win at Roulette". He suggests betting the opposite of the last 37 spins, the rationale being that it would be virtually impossible for the exact same sequence to repeat immediately (You could suppose that you had been betting that same sequence over the last 37 spins, in which case you would have had 37 losses in a row). The flaw in that theory is that the sequence needn't repeat EXACTLY in order for you to still get substantial losses, and the longer the sequence is, the more likely you are to get bunched losses, which is what you're trying to avoid in the first place. I've tried this method in the past, and it isn't very effective.

Similiar toward what i was thinking, but ...

A match is one event and has the value of 1.
A miss-match is one event and has the value of 1.

Same as ...

Same is one event and has the value of 1.
Opposite is one event and has the value of 1.

This means ...

That if you have 37 past results you can aim for what is underrepresented, no matter if its is same or opposite.
And in the beginning of each sequence you will notice what is overrepresented.
With other words so can i get 5.00 STD with same as overrepresented or opposite as overrepresented.

Bayes,

I do not contest your original points; take the probability for one outcome or ten outcomes, it makes no difference, but even so the probability of events still do not exist unless you bet on it (or them). There is always a probability, but unless there is a wager it's just theoretical. The probability applies directly to your wager, does it not? As oppose to the probability of observing some event.

It's not "there is a 18/37 chance it will be red".
It's "there is an 18/37 chance of doubling your money".

Betting black after five reds can be no different to regression measured over 1000 spins. In fact, betting black after fives reds is itself a regression bet; there has been a slight and fast deviation, now the player expects a slight and fast regression on the very next spin, so he bets with the erroneous belief he has a better chance of winning. And we already know the probability of winning the bet is .486. So, how is regression measured over many more spins different? Yes, you're considering multiple outcomes, but each outcome is still an individual. And you can even argue that where the outcomes are random, the SD is reset to zero on every new spin you record, so the whole idea of regression is an illusion until after the fact.

There is no expected value on a non-wager, the point cannot be contested even if put into any context. There is no probability of you winning or losing a non-wager. Thus, it can have no affect at all on your personal variance.

Let's pose a simple question, to make this explicitly easy to grasp and find out what people do and don't truly believe. Anyone is free to answer it.

You are betting on red, every spin, now you're losing heavily. The SD for red from your first wager is -2.0. At this point you stop betting. Now you continue tracking the SD until it reaches a virtual -3.0. Now you decide it's a good time to start betting for regression. When you place your next wager what is the SD? Is it -2.0 or -3.0?

Quote from: Pockets on December 22, 2013, 08:14:13 AM
It is dependent on only the outcome from the wheel.

This is the point I would like to stress: it is dependent on the outcome of your wager. Where there is no wager, the outcome from the wheel is irrelevant.

QuoteYou are betting on red, every spin, now you're losing heavily. The SD for red from your first wager is -2.0. At this point you stop betting. Now you continue tracking the SD until it reaches a virtual -3.0. Now you decide it's a good time to start betting for regression. When you place your next wager what is the SD? Is it -2.0 or -3.0?

The classical playing model for RTM aim to find one window of overrepresented events - then attack.
If you search for windows on a rolling basis, so is the last present window the current state, so the answer to you question is 3.0.
Each window is independent with overrepresented and underrepresented events.
Its based upon what benchmark you use as reference point when playing and measuring the distribution.

Quote from: Sputnik on December 22, 2013, 02:50:36 PM
The classical playing model for RTM aim to find one window of overrepresented events - then attack.
If you search for windows on a rolling basis, so is the last present window the current state, so the answer to you question is 3.0.
Each window is independent with overrepresented and underrepresented events.
Its based upon what benchmark you use as reference point when playing and measuring the distribution.

For the benchmark, the SD is measured against your wagers, not the events you observe, it can never be measured against anything else. I would love for you to explain why the opposite is not gambler's fallacy.

Similarly distribution is measured against your wins and losses. How can the SD be -3.0 when you haven't even bet to that point?

Quote from: Bayes on December 22, 2013, 01:46:31 PM

So it doesn't matter how long this pattern continues because I won't LOSE (I won't win much either, but as I said, I'm not bothered about that). In that sequence, the losing runs are short between wins. Occasionally I win for quite long periods flat betting, but I never "chase" wins using positive progressions. If I'm on a winning streak, I'll ride it as long as it continues and break off after one or two losses (depending on how much profit I made).

Well I would definitely like to see in which way you are betting amounts and how much, same what Rouletta asked, but  I have a feeling maybe you will keep that for yourself.

You have actually never talked on forums much about your MM and progression, except in quite general terms.

I find very interesting what you have said that you still wouldn't lose in no matter how much that pattern continues... So you can actually "hover" around breaking even with some % of wins all the time, and not getting into deep holes or increasing stakes drastically at the same time... As you said your progression is mild.

That sounds like a true art of MM...

My opinion is that this is much more worth to figure out then argue about RTM...

Drazen

Quote from: Bayes on December 21, 2013, 08:48:28 AM

it CAN help to reduce variance

This leads back to the first question from the beginning of the thread again - how do you measure that? Is there a figure in terms of smaller losing streaks, or narrower parameters between highest high & lowest low? If I'm searching for a lower variance bet, how can I be mathematically sure that I've got one? or is it subjective, you just feel it's lower variance because of experience & through many spins of play & practice?

Quote from: Number Six on December 21, 2013, 04:50:22 PMWhere there is no bet there is no probabilility of winning or losing, and there is no expected value....  All the maths applies only in relation to one's wagers and bankroll.

but there is still the probability of red or black (or whatever you're measuring) and whether you bet or not doesn't invalidate or affect the maths of that probability and its significance in relation to something like RTM.

Quote
The probability applies directly to your wager, does it not? As oppose to the probability of observing some event.

I would say 'as well as', rather than 'as opposed to'.

Quote
The SD for red from your first wager is -2.0. At this point you stop betting. Now you continue tracking the SD until it reaches a virtual -3.0. Now you decide it's a good time to start betting for regression. When you place your next wager what is the SD? Is it -2.0 or -3.0?


...


SD is measured against your wagers, not the events you observe, it can never be measured against anything else.

You can measure your wins v losses with an SD value, & you can measure the outcomes that the wheel is producing with an SD value. They are different values but both exist, both can be calculated, & either or both can be ignored if you so wish, but both are still there and a figure can be put on either of them at any time.

Quote from: Bayes on December 22, 2013, 09:34:29 AMThe only way I know how to do that RELIABLY (I've tried other approaches like following trends and switching from one bet selection to another, without any long-term success), is to "use" RTM. I wait for an extreme deviation (3 SD+) and then start betting when I see some indicators that the tide is turning, using a mild negative progression (it has to be negative, postive progressions don't work well at all).

RTM talks about the next event being more likely average, so why decide to take the side of, or wait for, the under-represented? Upon hitting that 3.0SD, why not bet for the side currently hitting? RTM says the next x spins will be more likely closer to the average, so why not bet the current 'hot' side (as a side must be chosen), and even if rtm began the moment you started betting, you're still likely to be in a period of 'average' which looks fine for your MM?
It seems that you are waiting for another extreme event to a lesser or greater degree (the under-represented catching up) to happen. As long as you got out when you see the indicators you are currently waiting for begin to happen, is that not at least as efficient?

Quote from: Drazen on December 22, 2013, 04:21:50 PMWell I would definitely like to see in which way you are betting amounts and how much, same what Rouletta asked, but  I have a feeling maybe you will keep that for yourself.Drazen

and I don't think we should put Bayes in that position of turning this into a 'Bayes show us exactly how to win' thread. This thread's not really about that, it's in the Maths & Statistics area, about variance, RTM & associated ideas, not really 'show us your exact MM'.

Quote from: monaco on December 22, 2013, 09:41:57 PM
but there is still the probability of red or black (or whatever you're measuring) and whether you bet or not doesn't invalidate or affect the maths of that probability and its significance in relation to something like RTM.

The ball can land in a red pocket, a black pocket or a green pocket. If you don't place a bet, it doesn't affect you then or at any time in the future; you can't win and you can't lose. So it is not significant to you in any way; even the regression itself is insignificant because there is no regression if there is no bet. At best you can define it after you have observed it between two points in time, but that provides no real life advantage.

Regression to the mean can be any length of spins taken at any time. Measuring it across 10 spins is no less valid than measuring it across 10,000. If your first wager is to bet black after 10 reds, is your chance of winning any better that 18/37?  No. It is gambler's fallacy. You will always be playing catch up from the first wager, you'll always be expecting something that isn't going to happen by any means other than coincidence. The same applies regardless of the size of your sample and how complex the bet selection is.

Where there is no expected value there is no variance in relation to your bankroll. Where you don't bet the outcomes are simply irrelevant. Please tell me how there is an expected value in a non-wager, and how a phantom probability can possibly influence your future. The EV explicitly applies to your wager, nothing else.

Here is the formula for the EV on even chances. 1 or -1 represents your return or loss.

(18/37 * 1) + (19/37 * - 1) = 18/37 - 19/37 = -0.027 * 100 = -2.70%

You cannot reach the EV without a bet. It's that simple. I suspect you are bored of me going on about it, I guess you will have to figure it out in your own time.

Quote from: Marshall Bing Bell on December 23, 2013, 01:54:38 AM
you introduce a personal permanence   

Makes sense to me. But I still believe it's a confusing topic.

If random outcomes follow the laws of regression, when you are playing virtually, the spins which are real creates patterns and will follow laws of regression and will have its own SD. This is my understanding of the whole issue and I think you cannot control variance using virtual bets. May be for the same number of spins, you might be better off, but if you consider the same number of placed spins  you will more or less be around the same mark, is my guess.

Virtual bets are more like hit and runs. Its just that instead of running every day after a hit, you will run after a short sessions within a bigger session. And we all know that HAR will not create an edge. A good explanation can be taken back to the day when Bayes challenged JL questioning HAR with a logical example. Same holds good for virtuals as well is MHO.

What Bayes does is a little more than virtual betting. Little more than HAR. There is a subtle difference. Read in his own words, even though it was in different context, it makes absolute sense and can be related here.

"I agree. I've tried to make this point several times but it seems people just don't get it. I play this form of HAR myself - look for "favourable" opportunities, get in then get out. There's a big difference between doing that and the kind of HAR JL recommends in PB and his other systems. Take PB: the "trigger" to bet is determined by random itself, not by any favourable conditions. The number of spins you have to wait when playing PB is a random variable, ie; anywhere between 21 spins and 60 or more.

Now someone might say that waiting for favourable conditions is a fallacy because roulette is a game of independent trials (unlike blackjack) so there is no such thing as a "favourable" condition, but that's irrelevant to the validity of the argument which only says that IF conditions are favourable, then bet and IF conditions become unfavourable, then stop betting. The logic of PB ignores any concept of favourable conditions because you enter the game at a random point in the stream and the random stream itself determines when you actually make the bet, so both "triggers" are randomly determined.

Playing short sessions (HAR) does nothing to change this, it just means you're entering at different points along the stream of data, but in both cases (either playing HAR or continuously) the trigger is selected randomly. Placing 100 bets in one mode or the other (HAR or continuously) thus amounts to the same thing in terms of favourable conditions, ie: they're not taken account of in either case because it's not part of PB's remit to find any favourable conditions; PB is supposed to be an absolute winning bet - the final pattern will materialize at a constant rate which is higher than 7-1 and this will guarantee you a profit. But this simply isn't the case."


Drazen, we may not discuss it here. But you might be absolutely right. The MM is what is creating that edge and not the bet selection here.