Repost on roulette (longish)
You might be interested in this repost of the rudiments of a system I posted a while back. I made a little money with it, but I have no idea if I have an edge or not. The system is worth its price, which is nothing:)
To pursue this subject seriously you need to study everything on Laurance Scott's site at Advantageplayer.com, buy lots of somewhat expensive literature and spend a ridiculous amount of time practicing. Then you will probably find you have little or no edge. Sorry I can't be more positive.
I have some roulette experience. A remark in one of Thorp's academic papers on roulette prediction "Approximate calculations yield a 15% edge" caught my interest.
I derived a method from Thorp's analysis of dealer signature, the fallacious notion that dealer's spin the ball consistently enough to be somewhat predictable. The procedure is basically this:-
1) Calculate the average number of pockets a dealer dealer spins the ball away from the last number he rolled.
2) Calculate the RMS (Root Mean Square) error, presumably expressed in some number of pocket-widths, from the release point to where the ball falls off the track.
3) Calculate the RMS error, again expressed in pocket-widths, in the velocity of the rotor.
4) Calculate the RMS error in the velocity of the ball.
5) Add the root mean square errors together. Any figure less than 18 indicates a positive expectation can be attained. Now you can bet the last number rolled + the number of pockets around the track determined by your dealer's signature.
The problem is, as Thorp records, the RMS at each of the three stages is often larger than 18, let alone collectively. Thorp states that he would be willing to examine proof to the contrary. I tried to find it figuring some of the highly professional croupiers in London might be more predictable-guess what, I couldn't.
However, while the DS phenomena is non-existent or not exploitable as a practical matter, it can apply on individual spins. We can use our powers of observation to determine what those spins are.
For example, it should be possible to reduce the error caused by the ball falling off the track to pretty close to 0 by simple observation.
Similarly, with ball velocity, the error can be substantially reduced by observing the number of revolutions, or revolution-fractions in the period before the coupier calls out "no more bets please".
Often, it is simply obvious that the dealer has changed gear, and these spins can be eliminated as random noise.
Sophisticated roulette predictors use a "cadence", a song-fragment or sound, as a constant unit of time, to measure the number of revolution-fractions a ball has passed through in a given period.
Finally, because the variables of a given spin rarely behave themselves, you will not be betting on most spins. This a little like backcounting.
To fully exploit your advantage you need to understand how to diversify your bets to fully exploit your advantage. Without an understanding of optimal betting theory you are not going to make any money at this. You need to read up on this in Thorp's "Mathematics Of Gambling" available at bjmath.com.