EC bet selection

[Carsch]

I had been looking for a way to help me with my bet selections when playing EC events so to experience the least amount of losses and smaller gaps between Wins/Losses. I think the best way might be to simply make up your selections in your head and follow through with a short session, let's say, of 3 spins.

For example, just think of 3 consecutive outcomes for R/B (e.g., RRB). Play them in that order for a session of 3 spins.

Win the first bet and you end the session. Make up another selection in your head from a different EC event (e.g., HLH)

Lose the first bet, and continue with the other 2 bets till you end the session.

Try not to go all same (e.g., RRR, LLL, EEE)

Now, at times it's obvious when there is a trend taking place. In such case, to maximize your winnings, go with the trend till it's over then switch back to your random mind-bet selection.

The idea with this type of bet selection (random bet selection, that is) is that within three spins, most of the times, you will be either +1, -1, or you break even.

Try this and see what happens. You might surprise yourself.

With this, i'd recommend a money management system made up of small sessions (3 spins is a good way to go). For each session, raise 1 unit every time when in minus (-)

Any thoughts?

[GLC]

Sounds like an interesting idea.

I suppose it's like all of these ideas, you have to try them to see if you can develop a feel that can give you a slight edge over random.

My problem is that I get side-tracked by another thought or posted idea before I have time to adequately test something that if I did test it more, may be useful.

Speaking of being side-tracked.  I was thinking of an idea similar in ways to yours.

Pick a pattern to play for.  Use any method to pick a pattern that you want.

Let's use the idea Bayes presented of trying to follow recent patterns.

If the last time B showed it was a single B.  We would bet on are for 3 chances to hit.  Bet progression could be something like 1-2-3.

If are hits within the 3 bets you either win 1 unit or get your money back.

If are doesn't hit within the 3 bets, you pick a new pattern to bet for and go to betting 2-4-6 until you recover the lost 7 units.

Just another idea for a little different angle to the game.

Cheers,

George

After further thought, this is just betting that a certain color won't hit in the next 3 spins.  Sometimes we approach a simple thought in a complex way.   Cheers.

[Carsch]

Sounds like an interesting idea.

I suppose it's like all of these ideas, you have to try them to see if you can develop a feel that can give you a slight edge over random.

Yep, that's the idea. For a while now, i've been working around an idea given from hints from someone who says he's got a system that never loses betting ECs. He averages around 30-35 units per 100 spins. He bets no more than 4 units and uses a Stack & Split money management system. He's down in average no more than 6 units, the worst would be 9 units. When losing he's usually betting 1 unit, or he may not bet at all until a win. He may switch from RB to EO, to HL. He says his system is designed to catch streaks; thus, for his bet selection he is also using his human factor instead of relying on mechanical system alone.

My problem is that I get side-tracked by another thought or posted idea before I have time to adequately test something that if I did test it more, may be useful.

That's my thing too. I have bunch of systems i'm looking at the same time, then i spend more time on one and forget about the others, then switch to another that might have an appealing idea to it. 

Speaking of being side-tracked.  I was thinking of an idea similar in ways to yours.

Pick a pattern to play for.  Use any method to pick a pattern that you want.

Let's use the idea Bayes presented of trying to follow recent patterns.

I'll look into that. I'm still learning my way around this board. 

After further thought, this is just betting that a certain color won't hit in the next 3 spins.  Sometimes we approach a simple thought in a complex way.   Cheers.

True!

[mr.ore]

That's interesting, I returned back to researching roulette this weekend as I neglected my hobby for a few weeks, and had similar idea. I know, that math tells us that we cannot predict random, but I have decided to take a more practical approach. I generated 1000000 RNG spins with high quality RNG, and decided to find a bet selection on EC, that gives edge over these spins, or at least have a lot of long winning runs with occasional long losing one to make a balance. I ignore zero, to make things easier, zero must be ignored and dealt with later.

The logic behind bet selection is to catch a long enough trend, so you can capitalize on it. If you draw a graphs of all ECs on no zero roulette, you can see, that (pseudo)trends formed. They are not really trends, because it would be against math, they are just random fluctuations, therefore I call them pseudotrends. Good think is, that trends can last for tens thousands of spins, so if you find a way how to jump on a winning one or form one by jumping from one to another, than you have (pseudo)edge. It is not real edge, just a random fluctuation, therefore I call it pseudoedge, it's just a delusion - but money are real thing, you know What I am trying to do is to find a set of rules to catch a good trend.

And now, we can move from science to magic, from math to alchemy. What is the basic theorem of alchemy?

AS TOP, AS BOTTOM. RULES, THAT GOVERNS SMALL DETAILS, ALSO GOVERNS THE WHOLE

On no zero roulette, probability, that you win 10 units, is same as the probability, that you lose 10 units. If you name event of winning 10 units W, and the event of losing 10 units L, and make a patter of them, you will get similar pattern to that of ECs. You can also fabricate bet with any payout. If you name event of winning 10 units W, and an event of losing 20 units L, and count Ws and Ls, than there would be 2/3 Ws and 1/3 Ls on average. I have counted it, it is true. If you have a game with zero expectation, ie. zero house edge, it does not matter, what is the payout, as long as it is (1/probability-1):1. You can create many other bets with even chance, it just takes longer to get up or down. The same rules governs all.

Now - how does bet selections works? We all know, that the event that today will win same numbers in lottery as those drawn yesterday is virtually zero. So it would be [tex]\begin{array}{cc}stu&pid\end{array}[/tex] to bet them, it would be also "regular"(grr - I can edit my post again - there should have been [tex]\begin{array}{cccccc}s&t&u&p&i&d\end{array}[/tex], now it is finally what I wanted it to be) to bet those drawn before yesterday. The lower the probability of event, the lower the lower the probability of repeat in near future. On the contrary, events tends to repeat, because p+(1-p)*p+(1-p)^2*p+... + (1-p)^k*p is getting bigger slower and slower as the k increases, so most cases are on the beginning.

Now it is where we can gamble, we want to catch a streak where the rare event does not repeat immediately, so we bet against it.

Example:
Imagine you have two spins on RED/BLACK. You can have RR,RB,BR,BB. What is the probability that after RR another RR won't come? 3/4. So we can bet against it, and as long as we choose to bet any other pattern than that which just came, we have pseudoedge. We can do that on ECs, but same idea works for trends. If you make winning 5 units X and losing 5 units Y, then after YY usually you can see XY,YX or XX. Because graph is much longer than just ECs, you play very few plays, and so have a big chance that the selection won't bust. This way the risk can be moved from bottom to top. If you have a good trend over hundreds or even thousands of spins, then even if it cannot overcome house edge without progression, as long as you can count on it not to go down too quickly, you can have a long term winner (it is my belief, I cannot do that yet).

So with "alchemy" we must fight the game of roulette, because rigid logic won't work. But don't throw logic and math from window, they are great supporters. But practically - while it is not possible to beat roulette, it might be possible to beat given 1000000 spins even with zero included.

[mr.ore]

The second word "regular" in my sentence regarding lottery should also be [tex]\begin{array}{cccccc}s&t&u&p&i&d\end{array}[/tex]. Regular is for example grammar in Chomsky's classification, but not betting on bad numbers (hmm...). I am -pedantic -O3, that's right.

[mr.ore]

I have just realized that I use certain word regularly without noticing it is maybe actually bad. Is that certain word really so bad in English speaking countries? In my country no person can be that, it is not used for persons at all, but in English it can be used to describe someone as one with certain intelligence deficit, thus the word being used as an insult. Heh, sorry

[warrior]

That's interesting, I returned back to researching roulette this weekend as I neglected my hobby for a few weeks, and had similar idea. I know, that math tells us that we cannot predict random, but I have decided to take a more practical approach. I generated 1000000 RNG spins with high quality RNG, and decided to find a bet selection on EC, that gives edge over these spins, or at least have a lot of long winning runs with occasional long losing one to make a balance. I ignore zero, to make things easier, zero must be ignored and dealt with later.

The logic behind bet selection is to catch a long enough trend, so you can capitalize on it. If you draw a graphs of all ECs on no zero roulette, you can see, that (pseudo)trends formed. They are not really trends, because it would be against math, they are just random fluctuations, therefore I call them pseudotrends. Good think is, that trends can last for tens thousands of spins, so if you find a way how to jump on a winning one or form one by jumping from one to another, than you have (pseudo)edge. It is not real edge, just a random fluctuation, therefore I call it pseudoedge, it's just a delusion - but money are real thing, you know What I am trying to do is to find a set of rules to catch a good trend.

And now, we can move from science to magic, from math to alchemy. What is the basic theorem of alchemy?

AS TOP, AS BOTTOM. RULES, THAT GOVERNS SMALL DETAILS, ALSO GOVERNS THE WHOLE

On no zero roulette, probability, that you win 10 units, is same as the probability, that you lose 10 units. If you name event of winning 10 units W, and the event of losing 10 units L, and make a patter of them, you will get similar pattern to that of ECs. You can also fabricate bet with any payout. If you name event of winning 10 units W, and an event of losing 20 units L, and count Ws and Ls, than there would be 2/3 Ws and 1/3 Ls on average. I have counted it, it is true. If you have a game with zero expectation, ie. zero house edge, it does not matter, what is the payout, as long as it is (1/probability-1):1. You can create many other bets with even chance, it just takes longer to get up or down. The same rules governs all.

Now - how does bet selections works? We all know, that the event that today will win same numbers in lottery as those drawn yesterday is virtually zero. So it would be [tex]\begin{array}{cc}stu&pid\end{array}[/tex] to bet them, it would be also "regular"(grr - I can edit my post again - there should have been [tex]\begin{array}{cccccc}s&t&u&p&i&d\end{array}[/tex], now it is finally what I wanted it to be) to bet those drawn before yesterday. The lower the probability of event, the lower the lower the probability of repeat in near future. On the contrary, events tends to repeat, because p+(1-p)*p+(1-p)^2*p+... + (1-p)^k*p is getting bigger slower and slower as the k increases, so most cases are on the beginning.

Now it is where we can gamble, we want to catch a streak where the rare event does not repeat immediately, so we bet against it.

Example:
Imagine you have two spins on RED/BLACK. You can have RR,RB,BR,BB. What is the probability that after RR another RR won't come? 3/4. So we can bet against it, and as long as we choose to bet any other pattern than that which just came, we have pseudoedge. We can do that on ECs, but same idea works for trends. If you make winning 5 units X and losing 5 units Y, then after YY usually you can see XY,YX or XX. Because graph is much longer than just ECs, you play very few plays, and so have a big chance that the selection won't bust. This way the risk can be moved from bottom to top. If you have a good trend over hundreds or even thousands of spins, then even if it cannot overcome house edge without progression, as long as you can count on it not to go down too quickly, you can have a long term winner (it is my belief, I cannot do that yet).

So with "alchemy" we must fight the game of roulette, because rigid logic won't work. But don't throw logic and math from window, they are great supporters. But practically - while it is not possible to beat roulette, it might be possible to beat given 1000000 spins even with zero included.

question what do you do when you are betting RR and the streak of reds com. thanks

[Bayes]

I've been experimenting with a plan (still in the early stages) which tries to pick the "best" pattern from a sort of hierarchy of patterns. Yes, I know all patterns are equally likely, and this may well turn out to be nonsense. 

Something along these lines has been done before, I sure, but the idea is to track the sets of patterns of length 2,3 and 4, with bets being selected on the basis of how the "higher" patterns are performing. For example, there are 4 patterns for a sequence of 2 bets:

RR
BB
RB
BR

And 8 patterns for a 3 spin sequence:

RRR
BBB
BBR
RRB
RBR
RBB
BRR
BRB

And 16 for 4 spins patterns, which I won't write out. Since each shorter pattern is contained in a longer pattern, the plan is to select bets to "complete" longer patterns which meet some criteria. For example, suppose a particular pattern from a 3-spin sequence is "doing well", if there is a pattern from the 2-spin sequence which is also running above expectation, which has just been played (successfully), then to select the next bet I would look at all 3 spin and 4 spin patterns and if there are any which have an EC in common (R or B) AND they are both running above expectation, then I would choose that bet.

That's the gist of it, but I have a lot of work to do yet. It may turn out that the best patterns to select are those running below expectation, or AT expectation, or maybe nothing will make any difference. But, since every bet has a 50% chance of winning, this approach can't make things any worse. 

[mr.ore]

Well, all what I want is to force the line to be in certain area, I'm experimenting with filtering out positive trend, negative trend, and trend which is as close to expectation as possible.

Playing positive trend is good for obvious reason, negative trend has an advantage that you can bet against it going down any further, and the one close to expectation can be used with stoplosses.

Playing after BB any other pattern does not work any better than anything else. But what about playing opposite after EC is 20 units ahead? If each 10 units won is W, then there is 3/4 probability that next it will be WL,LW or LL. So bet LL to catch any of those three patterns. Just negate the last result and be it. It does not work long term, and it should not be better than any other bet selection. But short term it seems to work. The aim of this is to force that graph to be near zero, and if played directly on RB, to have high hit ratio. I would prefer to have a selection that is actually losing if there is a house edge, but have higher hit ratio most of the time.

Another idea - to have several bet selection, and if one selection fails, or criterion for leaving is met, we switch to another bet selection, which is opposite or partially opposite to the actual one. If the previous selection gets trendy, we might switch to hunt repeating. Then make a set of possibly dynamic rules for switching them, tracking the rules, and always select one which just failed, betting that it does not fail two times in a row. If there were 8 equals rules of switching between two selections, then probability that one rule fails twice in a row is (1/8)^2. So the same idea for the selection that is not much good actually can be used on selecting rules.

Another idea - to do the simple selection in different levels. So we use it on RB, after RR we bet ~R~R=BB betting it won't repeat. Now we have another stream of wins and loses - what about applying same idea on this stream? And then again and again? Or what about tracing 2 units down L, 2 units up W, and after 4 spins, where we have for example WLs, we negate another 4 spins accordingly.

Example:
RR      0     bet BB
BB   +2 W bet RR (another W or L will be either at +4 or 0)
BR    +2  bet RB (we bet RR so -1 + 1 = 0 so we stay at +2)
BB    +2  bet RR
RR    +4  W  selections switched to LL, bet RR*
*now we have two Ws in meta level, so we suppose, that next there would not be WW, but anything other, so we bet LL to catch it, so both following two sessions we will play the other bet selections, where XY -> bet XY actually follows. This way we hope not to catch two another Ws in the previous selection.

-----------
RB  +4 bet RB
BB  +4 bet BB
BB  +6 W@ bet BB (first L selection won)
RB   +6 bet RB
BR   +4 L@  (second lost) bet RB*
we had WL, so the next session we bet L, and another W
we won first W@ with streaky bet selection, we expect it to lose, so we shall play
non repeating of previous patter

RR  +4 bet BB
RR  +2 L  bet RR*
*L@ was made with streaky selection, so we bet it will win now, thus for second session we shall use pattern continuation
RB +2 bet RB
RR +2 bet RR
RR + 4 W
So another two meta level games is patter LW
and so on...

The idea can be used on higher levels several times. Not coded yet, just an idea. To apply same rule on different levels of play. If it works for short time in lower levels, why should it not work in higher levels for short time? Because higher levels last longer, it might fool us to work long term, but that's what I want.

[Blood Angel]

All this kind of fits into what i am currently looking into...fascinating stuff...thank you.
If I get anywhere Illl be sure to post it here..

[mr.ore]

So I have programmed that "alchemy inspired" selection, and there are results. I used seven layers of pairs of bet selections XY->~X~Y and XY->XY and always the selection in a higher layer selects from the two bottom ones as if they were even chances. On top of that is XY->~X~Y. On the bottom are two dummy objects which are set to R,B according to game and are themselves simple selection (I did it so that I can use polymorphism an pointers in C++, technical blah blah blah). And the results are - inconclusive. It seems good, but I have not tested yet enough millions spins (I do not deallocate spins history so I cannot run it right now much more spins because that eats a few GBs of RAM, need to do optimizations). Well, will experiment with adding more and more layers. It seems it can partially control the beast in macroscopic view, yet in short term there are still long losing series.

The results are on my RNG that I'm "curve fitting", and then 64000 spins from Wiesbaden and 20000 playtech spins. It does not mean anything though. Green line is play according to bet selection, red is red, black is black 

[mr.ore]

The problem with any bet selections is, that if it can go up, then it can also go down, so a really good bet selection should be able to hold bankroll as near as possible to expected value. If we know that it won't deviate much, we can bet when it deviates enough, and should have quite a consistent winner. We cannot fully control the graph, but maybe we can do so with certain probability, that's where we must bet.

[mr.ore]

Another Wiesbaden spins, I will probably switch my research to these spins, and use RNG just for more testing.

[mr.ore]

Does not work in short term, 1000 spins sessions, useless idea after all. It just does not deal with the little fluctuations that happens. I have always believed that bet selections are just gambler's fallacy.

Well, I will give it a try yet, because progressions I have theoretically mastered - it's just a parachute which in one hit brings bankroll to target with minimal units possible(to avoid erosion by house edge), if such bet cannot be made, just use lowest possible alternative target greater than target, no progression can do better.

But back to selections - where is the logic behind them, I am yet to see bet selection, where is some logic in it. Wrangler seemed good for some time, but it fails anyway.

Well, I will try to do the same with wrangler, add up to 7-8 meta games, and hope to force graph to be choppy even if trending, and capture long trends. I have to find anti-Wrangler first, something that loses when Wrangler is winning, to have antagonists.

Another idea - to use evolution:

To have a bunch of "individuals" who are just say 8 characters long strings like RBRRRBBR, BRBBRRBB,... and have them play simultaneously in virtual mode, in real mode always play majority decision. Tracking how are they performing, and evaluate them - that's their fitness.

There would be an operation after some spins where a new generation would be generated in this way. Copy best to a new generation - elitism. Then create new offspring by selecting two individuals, randomly selecting point for "crossing over" and swap the remaning string.

Examle: RRBRB|RBR                                RRBRB|BBB
                    x                  -------->
             BRRBR|BBB                               BRRBR|BBB

Evaluate them in hindsight(how would have they performed last session),and add them to the breeding pool with their parents.

Now randomly select individuals from the breeding pool with probability according to their fitness (string that performed better have greater probability of survival to next generation). Pool have more individuals than generation, we play always with fixed size generation.

When the next generation is created, we randomly mutate each individual. On each [tex]R[/tex] or B we make a random decision with low probability, that it would be reversed. That is a mutation, to bring in change to the process.

Why this? Evolution is nature's way to deal with "trends" in random brutality of cosmos. Earth itself is just a trend in space, and life is nature's bet selection and adaptation to it's environment. As long as there is something stable, something "trending", than life can find it and use it, adapt to it. Why not use the same idea? Well, it won't work, but I will have fun programming it

[chrisbis]

Just read this thread from top to bottom, most interesting.

I just wondered if one of the IT coders could come up with a
EC chance choosing Machine/guesser??

Something that suggests series of EC bets in a string formate
for the staking of EC combinations.

possibilities are of course vast, tho could be as simple as-

Bet 1. >>Red and High
Bet 2. >>Red and Low and Odd
Bet 3. >>Black and High and Even
Bet 4. >>Black and High and Even

This could be one session, and of course, the player could decide whether to play
the bet like that, or do a Marty progression on each bet till it wins or breaks even.