Testing Gambler's Fallacy

[falkor]

According the Gambler's Fallacy the wheel has no memory, so past spins have no effect on future spins. I thought I would try to test this and now I can't stop winning money! Can anyone else test this and confirm please?

EC bets can only come in a row up to 22 times in one million spins. Now, try waiting for 18 different numbers to come in that are unique and list them; this will probably happen after 20-25 spins due to repeating numbers. Once that happens it means that the other 18 numbers haven't come in yet, but the 18 numbers that have come in means that half the board has already come at least 18 times in a row, so try betting all numbers on the other side still waiting to come in.

If you lose you can double up on the same numbers until you win.

If you win then you need to repeat last bet (but don't hit spin), removing the winning chip and adding a chip to the oldest number in your list that isn't already in play. You'll end up adding to your list at one end; erasing at the other end as you win.

Let me know if you have to double up more than 4 times? If Gambler's Fallacy is true then half the board could hit 44 times if it's not a traditional EC bet; if false then you should never need to double up more than 5 or 6 times max using this strategy.

[GLC]

Falkor,  Does that mean that if you track until you get only 12 numbers hit, then bet on the other 24 numbers you would bet 1 unit on 24 numbers and if you lose, 3 units on 24 numbers and if you lose, 9 units on 24 numbers.  That's 24 units, 72 units, 216 units.  It seems that you could improve your odds if you don't bet until you have at least say 18 spins with only 12 numbers represented.  To lose, you'd have to go 21 spins without hitting 2/3 of the wheel (not counting the zero).  I think it would be very difficult to get 9 units placed on 24 numbers in time.  You might have to limit this strategy to 2 attempts for a 96 unit loss if you miss on the 1 unit and 3 units per number bets.  It can happen, but not very often, just like betting on your 18 numbers.  I suppose 18 numbers would be easier to play.

I know that I've read a system somewhere where you wait until you get like 48 or so spins with 12 numbers not represented and then bet those 12 numbers with a 1-1-2-3-4-5 sequence.  Same concept.  You lose 192 units if your 12 numbers stay missing 54 spins.  We could use a computer and look for bets like this all over the wheel. 

By the way, this concept is what got me interested in roulette in the first place.

GLC

[nowun]

I tracked until only 18 numbers left (25spins) then bet the 18 numbers with the martingale progression suggested by Falkor.  Then followed his betting idea of deleting the number hit and adding the number furtherest back in the tracking if not already included..

Here is a 100 spin session using this idea.  The scary part is going past 6th step (already in hole 1134 units), but in the 100 spins I only got to the 6th once and got to the 5th once, the rest were minor.  I think the progression should be shortened as per GLC's suggested progression. This idea definitely has potential.

[nowun]

Heaps more testing needed but here is another one that was pretty impressive, again 18 numbers left at spin 25 tracking.  This one went to 7th level once as can be seen in the graph at around the 75 spin mark.

This hits regularly enough that a short progression and stop could easily be employed and still be a winner, but MUCH MORE testing is required.  This could just be luck.

[nowun]

Here is the first example I gave, flat betting:

[falkor]

Falkor,  Does that mean that if you track until you get only 12 numbers hit, then bet on the other 24 numbers you would bet 1 unit on 24 numbers and if you lose, 3 units on 24 numbers and if you lose, 9 units on 24 numbers.  That's 24 units, 72 units, 216 units.  It seems that you could improve your odds if you don't bet until you have at least say 18 spins with only 12 numbers represented.  To lose, you'd have to go 21 spins without hitting 2/3 of the wheel (not counting the zero).  I think it would be very difficult to get 9 units placed on 24 numbers in time.  You might have to limit this strategy to 2 attempts for a 96 unit loss if you miss on the 1 unit and 3 units per number bets.  It can happen, but not very often, just like betting on your 18 numbers.  I suppose 18 numbers would be easier to play.

I know that I've read a system somewhere where you wait until you get like 48 or so spins with 12 numbers not represented and then bet those 12 numbers with a 1-1-2-3-4-5 sequence.  Same concept.  You lose 192 units if your 12 numbers stay missing 54 spins.  We could use a computer and look for bets like this all over the wheel. 

By the way, this concept is what got me interested in roulette in the first place.

GLC

Exactly - Same concept! Likewise you could wait until only 6 numbers are left and bet on them.

If it doesn't work then it's quite confusing to me since red/black, odd/even, high/low are fixed to 21 in a row +/-2 every million spins, but if my scope is any 18 numbers on the board can that increase to 42+-1? It should either win without needing to double up 5-7 times depending on how many 0s have come in - or I just cannot get my head around it.

[falkor]

Here is the first example I gave, flat betting:

Cheers for doing all the testing! Are you "rotating" or repeating the 18+ virtual spins after a win? If not then the perceived advantage will only apply to the first bet.

[nowun]

I was doing it as per your instructions, removing hit number and adding oldest number.

If you win then you need to repeat last bet (but don't hit spin), removing the winning chip and adding a chip to the oldest number in your list that isn't already in play. You'll end up adding to your list at one end; erasing at the other end as you win.

[falkor]

OK - that's quite complex to program I think - because you always want to bet against the last 18 numbers that came in most recently unless you are on a losing streak. Kudos to you if you managed to get the software to do that!

[Kav]

Three points:

1) Very interesting approach, but...

2) Falkor, the "fixed" 21 series in millions spins (where did you get that?!) is already invalidated since your first 18 numbers take 25 spins to appear. So, yes you can say the 21 series max is wrong!

3) It is not very uncommon to appear only 20 numbers in 36 spins. So, yes again, the 21 series max is wrong.

Try to calculate how many different 18-number-groups there are in roulette. I'll tell you how many: Too many!
If you want to know what to expect search for the longest EC series not in 1 million but in 1 billion spins.

[falkor]

kav, I meant the 3/6 traditional EC bets are fixed to 21+/-2 every 1 million spins.

[Kav]

Hi,

kav, I meant the 3/6 traditional EC bets are fixed to 21+/-2 every 1 million spins.

Where did you get that? What is your source or reference?
It is already invalidated since your first 18 numbers take 25 spins to appear.
Try to calculate how many different 18-number-groups there are in roulette. I'll tell you how many: Too many!
If you want to know what to expect search for the longest EC series not in 1 million but in 1 billion spins.

[falkor]

Kav, 18 numbers taking 25 spins doesn't invalidate the behaviour of traditional ECs, but it's just something we don't understand.

What is the world record for the longest number of reds/blacks in a row? It can't be much more than 21+/-2 because that's the result I always get when I test 1 million spins for max ECs in a row. I haven' tested 1 billion spins because I don't have that data from any real casinos.

[falkor]

OK, I've managed to code this system. Unfortunately, I haven't quite figured out how to do it on paper yet: it's not as simple as adding the oldest number on the list. Nevertheless, the system checks the previous X amount of spins - renews each cycle - in order to determine the last, say, 18 numbers that appeared (18 times in a row for that portion of the board), etc.

The next stage is to determine the max amount of wins and losses in a row per 10,000, 100,000 and 1 million spins, for each of these:

33 vs. 4
32 vs. 5
31 vs. 6
30 vs. 7
29 vs. 8
28 vs. 9
27 vs. 10
26 vs. 11
25 vs. 12
24 vs. 13
23 vs. 14
22 vs. 15
21 vs. 16
20 vs. 17
19 vs. 18
18 vs. 19
17 vs. 20
16 vs. 21
15 vs. 22
14 vs. 23
13 vs. 24
12 vs. 25
11 vs. 26
10 vs. 27

bold = most likely the optimal range

29 vs. 8 @ 10,000 = 6 wins vs. 30 losses
18 vs. 19 @ 10,000 = 13 wins vs. 10 losses
13 vs. 24 @ 10,000 = 19 wins vs. 9 losses
10 vs. 27 @ 10,000 = 24 wins vs. 6 losses

[falkor]

24 vs. 13 @ 10,000 = 7 wins vs. 23 losses
20 vs. 17 @ 10,000 = 10 wins vs. 11 losses
16 vs. 21 @ 10,000 = 13 wins vs. 10 losses

I've got a 31 step progression for 8 numbers:

1   1   28
2   1   19
3   1   10
4   1   1
5   2   19
6   2   1
7   3   10
8   4   11
9   5   3
10   7   14
11   9   8
12   12   13
13   16   21
14   21   24
15   27   14
16   35   11
17   46   27
18   59   10
19   77   28
20   99   13
21   128   17
22   165   12
23   213   18
24   274   3
25   353   3
26   455   14
27   586   20
28   754   13
29   970   5
30   1248   4
31   1606   18

I've got an 11 step progression for 27 numbers:

1   1   11
2   2   10
3   4   9
4   8   7
5   16   4
6   33   11
7   66   6
8   133   11
9   266   4
10   533   7
11   1067   9