Roulette Chasing - betting on the imperfection of the wheel

[Erazman]

Roulette "Chasing" -- betting on the imperfection of the wheel...

I've come up with a simple idea called "chasing". First off let's look at the fact that a roulette wheel is NOT PERFECTLY random. Small dents, unequal traction, etc.
These imperfections may only give a .01% improvement on certain numbers, but stay with me here..

You would think to take advantage of this .01%, you would have to somehow figure out which numbers are insignificantly favoured by imperfections.
It would be nearly impossible to choose the right numbers. BUT ..Simply betting on the previous "winning number" ensures that you are taking advantage
of these dents/scratches by letting the results determine your future odds!

If this doesn't make sense, let me use a very simplified roulette wheel with only 4 numbers
1,2,3,4. Let's say there's a large dent in this exaggerated example, causing the ball to usually land on 3. Let's say this dent is so severe, it causes
3 to win 60% of the time.
The dealer spins the wheel and drops the ball. It lands on 3.
So bet on 3 next time.
It lands on 2..
So bet on 2 next time.
Usually you will be forced to bet on 3, and only sometimes on 2.
This will be MUCH more profitable than randomly picking 1-4. By "chasing", you are letting the winning numbers tell you what works. Sometimes its wrong,
but in this case it's 60% right!


Anyways, this is the extremely simplified version of what "Chasing" is. Instead of betting on 3 60% of the time, you're going to be betting on
say, 31 2.7% of the time (as opposed to 2.6% or whatever it is), because 31 is coming up SLIGHTLY (0.01%) more often, and so you are betting on it more.

By chasing, you are betting WITH the imperfection of the wheel.

With 36 numbers to land on, and with dents/scratches not even visible to the naked eye, this advantage is EXTREMELY small, and by no means
signifcantly profitable (or is it?).

Now here's my question: on a European Wheel with a 2.6% house advantage, do you think a wheel might be flawed enough to have a 3% "chasing advantage"? Or is my initial estimate of 0.01% much more accurate?

It's also interesting to see how deliberately damaging a wheel could make profits with chasing very possible (and no need to even predict the track of the ball). It would be illegal but interesting to think about..

 

[HallsofIvy]

You are basically arguing that succesive trials are NOT independent. This is, then, similar to Peng's argument about flipping coins but just the opposite.

He was trying to argue (and apparently was surprised that no one was agreeing with him!) that if you flip a coin many times and get heads every time then the coin is much more likely to come up tails on the next flip. because "the final result has to be close to 50-50".

You, on the other hand are arguing- and I agree with you- that if a coin comes up heads many times in a row, it is probably not a "fair" coin and so is MORE likely to come up heads again than tails.

Perhaps a better example is a baseball player with a good 'batting average" (number of hits per appearance at the plate) who has recently been hitting below that average. You will often hear people say he is "due for a hit" when it would be more correct to say that he is probably in a slump and is less likely to get a hit than his batting average would indicate.

 

[matt grime]

It would be better, perhaps, to keep track of the winning numbers and bet on the modal average (that which occurs most often) at any given point in time. Chasing implies at some point you will almost surely place really bad bets. (Not that there is such a thing as a good bet on Roulette, necessarily). Of course that is harder to do, unless you've got a good memory.

 

[robert Ihnot]

Often time the roulette area will have a sign that supposedly keeps track of the last 10 or 15 numbers chosen. (This should aid your memory of past play.) HOWEVER, I have seen circumstances where these numbers are way off because(?) the system is not working and just gives long gone past performance, or what? Also, the board may not keep track of 0 or double 0, so that while the machine says red has just come up three times (making black due?) this was not really true since there was a zero in their somewhere, or even two zeros.

When those cases occur most payers just shrug their shoulders and keep playing, LIKE IT MAKES NO DIFFERENCE ANYWAY!

I have never, never forgotten the time I was at the horse races and my friend, an ardent handicapper, notices that the scratches and jockey weights were not posted where they usually were. So he goes up and asks the woman behind the counter,

"How come this information is not posted?"

Lady looks at him with disgust and answers, "DOES IT REALLY MATTER?"

I have often thought over her comment. I once heard someone who argued that these things do not matter, so even if the race is fixed, it makes absolutely no difference to those who don't know.

 

[ssd]

The idea of keeping track of past winners is quivalent to estimate the probability distribution from a sample of tracked numbers. Keeping the bet fixed on the mode of the estimated distribution is a nice idea (assuming that the mode does not change frequently due to machine defect).

 

[robert Ihnot]

If you want to play red vrs black the house edge is given as:

1-(probability of win)(return) = 1-(18/38)(2) = 1/19=5.26%.

Where as if you want to play just a single number paying 35 to 1, the house edge is same as above figure. What chance is there to overcome such odds?

 

[robert Ihnot]

In a small sample the irregularities are going to be greater than in a large sample. If p=1/38 and we look at 38 samples, by Poisson distribution we have u=np=38*1/38 = 1. P(0) =e^(-1)u^0/0! = e^(-1) = .37; P(1)=.37, thus the chances of a number coming up twice its expected frequency of 1 is 26%.

But if the same test is applied for n=380, or 10 times the above n, then the chances that a number would be twice its expected frequency or more, 20+ times, is reduced to .35%, about 1/3 of 1%, or almost nothing.

Your example is using only 4 numbers, this is a much simpler case than having 38 different numbers. For one thing p=1/38 and q=37/38 are widely seperated, so I used Poisson distribution. As is obvious, the chances of find a real bias increase as the sample size gets larger. Using a very small sample greatly increase the chances of chasing a "false alarm."

As for denting the ball, or damaging the wheel, NOT IN VEGAS! They would never let you even touch the equipment. In fact, they don't want you to even lean over or put your hand above the wheel.

But anyway, to repeat: the game has a built in factor favoring the house of 5.26%, and so gaining a 1% advantage in favor of certain numbers would help mostly in getting you so excited that you would keep on playing as you lose a bundle.

 

[Erazman]

If you want to play red vrs black the house edge is given as:

1-(probability of win)(return) = 1-(18/38)(2) = 1/19=5.26%.

Where as if you want to play just a single number paying 35 to 1, the house edge is same as above figure. What chance is there to overcome such odds?

5.26% on the american wheel (0 and 00).
2.6 on the european wheel.(only 0)

Here in Canada we have both variations.

 

[Hurkyl]

http://en.wikipedia.org/wiki/Multi-armed_bandit