Breaking Vegas:
"The History Channel's series, "Breaking Vegas," is created out of some of the most fascinating stories of people who went on incredible runs raking in thousands -- sometimes millions -- of dollars from the casinos."
"Beat the Wheel." "Childhood friends and physics geniuses Doyne Farmer and Norman Packard came up with a system to beat roulette."
What about reverse engineering the computer that calculates what zone of 8 where a ball is likely to land? By adding layers of speed, distance, 10% for hitting bumpers, and 33% for getting channel margins hits for slot boundaries it's possible to create the perfect American & European wheel simulation for testing actual spins during tournaments.
I need the advice of VB players. Does anyone know the approximate circumference of the ball track of a typical roulette wheel?
Is there a known range where the ball can be known to drop out of the track? Like for instance when the ball drops to a certain speed above a certain sector to sector speed on the wheel can the location for a sector estimation be made.
I want to mimic that for a really cool RNG.
I went over to the physics section and got some interesting information.
So a ball will be released at different speeds allowing to travel different distances from the release point on the track. Add to that the wheel will travel an equally sized disk of distance if you mentally visualize that the turning disk, the wheel, has an orbit that is symbolic to the same diameter or radius. At that orbital location, radius, is a speed coefficient, going the opposite direction of the ball. I can put together a release point that is relative to a location on the wheel's celestial position directly above that relative position on a wheel. It's like finding where you are on planet earth by knowing the time and the right ascension of a star relative to the north star. You can find stars that way too. At a certain time they will be straight up.
So, as far as an RNG goes, that is based on real averages of physics, I need two distances that cross each other at differing random points and has random collision components, and apply that to exact locations along a length that is symbolic of the slot track. It should be in millimeters so that collisions with slot boundaries causes ten different types of bounces. At that point a collision with a slot boundary can happen again. I will also employ three ball sizes that will effect collisions respectively.
Gizmo, mate, please do keep us posted on the progress on this. Be it this from your brainstorming to the announcement of the creation of the perfect RNG generator in software :thumbsup:
It would be nice to have this one ran through "post-processing and statistical checks"* AND after that be used in challenges!
*: yes I read wiki! nolinks://en.wikipedia.org/wiki/Random_number_generator (nolinks://en.wikipedia.org/wiki/Random_number_generator)
Kudos to you mate.
Victor
Quote from: VLSroulette on January 26, 2009, 11:53:58 PM
Gizmo, mate, please do keep us posted on the progress on this. Be it this from your brainstorming to the announcement of the creation of the perfect RNG generator in software :thumbsup:
It would be nice to have this one ran through "post-processing and statistical checks"* AND after that be used in challenges!
That's why I'm building it.
What kind of tables would confirm this: "post-processing and statistical checks?"
I think several trams of no less than 1.000.000 spins.
Kon-Fu-Sed can surely elaborate more on sample size to start trusting any RNG as "fair".
Quote from: VLSroulette on January 27, 2009, 12:46:53 AM
I think several trams of no less than 1.000.000 spins.
Kon-Fu-Sed can surely elaborate more on sample size to start trusting any RNG as "fair".
Well I'm sure it's going to be random to 100 million or more before a repeat and it may never repeat.
My version of a random function is the newest, most powerful version going for small computers like PCs. It gets its random seed from the clock. Whenever the application fires up it grabs a new unique random seed. These new algorithms don't go into a repeat phase until after a half a million spins. I can create many new random seeds during the process of calculating a string of random spins.
Just on that I could produce millions of truly random outcomes. I'm planing on combining that changing the random seed with millions of combinations based on real world physics. All that just to get 37 or 38 numbers randomly generated for tournament playing. Technically I could set it up to generate spins for anyone that wants it. Each time a request for a certain number of spins are requested it will set a new seed from a random selection from trillions. What the heck, I'll throw that into it too so that anyone can use it to get all the unique spins that they want. I'm going to guarantee that it's as random as Random.org results, for free.
Hi guys,
Quote
Kon-Fu-Sed can surely elaborate more on sample size to start trusting any RNG as "fair".
Thanks for your confidence, Victor.
Unfortunately I know really nothing about RNGs.
:(
But regarding tests, I found this:
nolinks://nolinks.stat.fsu.edu/pub/diehard/ (nolinks://nolinks.stat.fsu.edu/pub/diehard/)
There seems to be a zip-file with Windows software...
Also this:
nolinks://nolinks.phy.duke.edu/~rgb/General/dieharder.php (nolinks://nolinks.phy.duke.edu/~rgb/General/dieharder.php)
"DieHard" test or "DieHarder" test... :)
Best regards,
KFS
Quote from: Kon-Fu-Sed on January 27, 2009, 09:10:05 AM
"DieHard" test or "DieHarder" test... :)
Wow, that's from geekville for reading, for sure, to the MAX.
Thanks for the links.
My tournament application is using "Mersenne twister"
From the DieHarder tests:
"randu fails this test. To be more precise, the probability that randu is a "good" generator but produced the observed distribution is less than 0.000001, according to this test. mt19937, on the other hand, still returns perfectly reasonable values of p from the final KS test, even when run repeatedly with still larger -t"
mt19937 is "Mersenne twister" so it has already been power tested by DieHarder
Check out where it is in use:
nolinks://en.wikipedia.org/wiki/Mersenne_twister (nolinks://en.wikipedia.org/wiki/Mersenne_twister)
When I add the simulated physics to this it should be bomb proof. What's really interesting is that it is meant for runs like 100 to 400 spins at a time. In that regard it should be fine. It should also be fine for testing 10 million spins for a long term tests of the Martingale too.