System beated 700000 Wiesbaden spins, but just were lucky...

[mr.ore]

Well, recently have I moved my testing into over greater number of spins, because microoptimalizations are to be made with math. Today I returned to my old system where computer develops a system with policy improvement algorithm (Markov decision problem). There is yet some work to be done to fully create "sure losing, but maximally slowly" system. With math it is possible to create a system which is "almost" fair when compared to no zero roulette.

Well, even 700000 spins CAN be beaten with luck, and the key is a high variance system. No bet is bigger than 100 units (I don't know the highest bet yet, because it did not fit into my screen, the system is designed to win 4096 with one unit, if loses restart, until a tremendous winning streak happens. If a betting opportunity with such a HIGH VARIANCE is artificialy created, anything can happen...). The method is of course absolutely unhuman and artificial...

Anyway, old systems should not be just thrown away, there might be something good forgotten...

[mr.ore]

Max. bet changed to 50, and still in profit, but so so...

[mr.ore]

Max. bet changed to 25, target for positive progression changed to 8192, of course - it tanked. But before that it at least produced nice profit. And all that with only 25 units max. bet.

[iggiv]

this is just fluctuation. so huge downdraw

[mr.ore]

Same system, but this time used as a negative progression (bankroll=8191, target=8192) played repeatedly. Tanked after 100000 spins, max. bet 50 units.

[mr.ore]

Yes, fluctuation. That's what it is all about. To ALLOW for fluctuation in either side, and hope, that a HUGE profit will come, and if it happens so, take it, NEVER play again...

All what can we control is this fluctuation, how far it can go...

Another example bankroll=4096, target=6144 for each session, max.bet 50 units. 4096/6411 = 2/3, so now the system "emulates" two dozens, but on a higher scale. Makes a quick buck, then busts... The higher the ratio bankroll/target, the lower the variance and fluctuations.

Without fluctuations, house edge is a fast killer. Only weapon we have against house edge is a fluctuation...

[iggiv]

i think playing this way doesn't make sense. It is just luck. if u believe in luck u need no system. just take any lucky number or numbers and play.

[mr.ore]

The luck can be little helped, but not much. For example, say you bet 5 units, 3 on low and 2 on 3rd dozen. That is (18+12)/37 = 0.8108108... probability that you win 1 unit with those 5 units you have.

What if you bet this way: first unit on EC, if lost then second on a dozen, and if even that lost, 3 units again on EC. Probability you win: 18/37 + 19/37*12/37 + 19/37*25/37*18/37 = 0.8218269... , that is higher than betting all five units at once.

0.8218269-0.8108108=0.011016

That is 0.011016*100 ~= 1.1% greater chance that you win. The price is same - five units. You can take the better chance, for free...

Similar optimizations can be computed on greater scale. Anyway, bet selections failed for me, that is just fallacy. Only at least partially good results I had were with attempts to "control" dispersion. If losing, start betting low variance, if winning, try to make it little more l o o s e and see, if it wins...

[iggiv]

U can bet that way if you change your bets all the time. but you can't bet this way consistently.

for example: if you bet 2 dozens, at this very moment  you have 2/3 chances to win. But if you bet them second time you already have less chances to win 2 times in a row, and if you bet third time, you have even less chances to win 3 times in a row. So if all the time you bet 2 dozens you have just around zero minus house edge, cause you gonna met zeroes of course. So to win this way you need to change your bets somehow, but more you play like this -- more your result will be closer to zero minus house edge. I mean in a long run. but if you go into dangerous progressions then probably  you will lose because you will be going lower and lower, you will be catching "low waves" like this. Finally you will reach table limits or your wallet limits. Here casino has superiority over you because it dictates table limits and also it has always bigger wallet than u.

that's why casinos love high progression players.

[mr.ore]

Yes, that's true.

The sessions above was just a test for a system, which based on the theoretical idea that player will gamble only once in a life on a roulette. I wanted to see how it works without having to run simulation many times.

If a gambler were to gamble once in a life with only two units, the best thing he do is to bet on an EC, and if only it loses, then bet remaining unit to a dozen. 18/37 + 19/37*12/37 > 24/37.

I'm computing a similar thing for "lifetime" bankroll and "lifetime" win. As long as it does not hit table limits, it is pretty near fair gamle, so that if target=100 and starting bankroll=50, probability it wins is 48.9501%. I can compute that, without simulation. The aim is of course to avoid situation, where it plays twice same bet with high probability, so nothing above 18 numbers is not played.

Interesting is that what helps probability are bets like 1 unit on dozen, and 1 unit on sixline in that dozen, if it hits target in one win. Good are also numbers, they can create new options with non-standard payouts if there is enough units to bet.

All bets where you can break even are useless, like low and 1st dozen, because if it can break even, there can be a very long cycle, where only house edge is a winner, and it "eats" from bankroll.

[iggiv]

once in a life maybe

[mr.ore]

That is just theoretical, of course 

But if someone wanted to play two dozens three times in a life in order to win 3 units, he should rather look at this table:


bankroll |  bet                     |  W        |  L  |  Pwin        |  avg.return

0         |  no bet                 |  -         |  -    0.0000%    |  0.0000
1         |  1 unit on 8:1         |  9        |  0 |  10.8108%  |  0.972973
2         |  1 unit on 5:1 (1-6),|  9 or 3  |  0 |  21.4755%  |  1.932798
              1 unit on 2:1(1-12)
3         |  3 unit(s) on 2:1          |  9  |  0  |  32.4324%  |  2.918919
4         |  1 unit on 5:1             |  9  |  3  |  43.3893%  |  3.905040
5         |  2 unit(s) on 2:1        |  9  |  3  |  54.3462%  |  4.891161
6         |  3 unit(s) on 1:1         |  9  |  3  |  65.3031%  |  5.877283
7         |  2 unit(s) on 1:1       |  9  |  5  |  76.5562%  |  6.890056
8         |  1 unit on 1:1            |  9  |  7  |  87.9613%  |  7.916515
9         |  no bet                     |  -  |  -  |  100.0000%  |  9.0000


Since he wants to risk max. six units to get three, he is better off to just put 3 on EC and if he loses then 3 on dozen.

[mr.ore]

Now let's compare that with just even chances:


bankroll 
    |  bet              |  W  |  L  |  Pwin        |  avg.return
0  |  no bet           |  -  |  -  |  0.0000%   |  0.0000
1  |  1 unit on 1:1  |  2  |  0  |  8.8636%   |  0.797723
2  |  1 unit on 1:1  |  3  |  1  |  18.2196%  |  1.639764
3  |  1 unit on 1:1  |  4  |  2  |  28.0954%  |  2.528585
4  |  1 unit on 1:1  |  5  |  3  |  38.5198%  |  3.466785
5  |  1 unit on 1:1  |  6  |  4  |  49.5234%  |  4.457108
6  |  1 unit on 1:1  |  7  |  5  |  61.1383%  |  5.502448
7  |  1 unit on 1:1  |  8  |  6  |  73.3985%  |  6.605863
8  |  1 unit on 1:1  |  9  |  7  |  86.3398%  |  7.770578
9  |  no bet  |  -  |  -  |  100.0000%  |  9.0000


They are just worse in any case. There is no point in playing them. Even chances have only one purpose - to win that 1-3 units before system wins it's target. Then they are absolutely useless.

So it is not just luck, but also knowing that some way of playing will not win anyway, because the posibility is not there. When there is nothing to begin with, nothing can be done.