PDA

View Full Version : Question about Random Number Generators

bhizzle
01-23-2006, 06:16 AM
I, like alot of people, am addicted to online poker. I recently came across a program on the internet that claims to be able to give you a huge advantage, the advantage or knowing what other players have and what the upcoming cards are going to be. The company claims that the program synchronizes with each individual poker site's RNG and this is how it predicts the outcomes. The program doesn't even have to be connected to the poker software, you simply put in the data that you see(your hole cards and the community cards). The company says that after about 2200 hands played, you will have 99% accuracy with results, because it takes around 2200 full hands to gather enough data to have synchronized with the RNG. My only question is: Is this even theoretically possible?

thanks

brannan:confused:

JamieR
01-23-2006, 10:06 AM
The rules of this forum state that no illegal activity is to be discussed...cheating in games is one such form.

jkd
01-23-2006, 06:13 PM
The possibility of this software relies solely upon how these services generates random numbers. I highly doubt online poker websites reveal what their generator is.

If it is something simple like a linear congruential generator:
y(n+1) = a*y(n)+b mod m
Where y(n) denotes the nth random number, and a and b are some carefully chosen constants, and y(0) is the seed, then it might be feasible to reverse engineer the constants, but I would assume after many more hands of poker than 2200, given the widespread use of this algorithm.

If the psuedo-random algorithm is any more complex, or if they use natural data, such as atmospheric conditions or fluctuations of a lava lamp or anything, then such pattern-software is certainly impossible.

jkd
01-23-2006, 06:34 PM
Upon further thought (and memories of abstract algebra), 2200 hands is unrealistically low.

Any decent set of values would modulu numbers close to 2^32, or 4,294,967,296. Therefore, the maximum theoretical period is close to that, and if the values are chosen well (which they most definitely are), then the true period of the function is probaby the maximum theoretical one. So you can either iterate over 4 billion consecutive hands, or try performing a 4 variable LCG-regression over your data and attempt finding some statistically close sets of data; however, this algorithm is chaotic, and variables close but not exact will still lead to vastly different outcomes. If such analysis software exists, certainly it must utilize the latter approach, building a collection of 4-variable tuples that work so-far, and refining them as play continues. But the initial set of these tuples must also be prohibitively massive, plus who knows when the poker game actually invokes the random() method, and in which order the cards are internally dealt, and so on.

I pronounce such software a ripoff.

slushy77
01-31-2006, 03:12 PM
to work out a random number you need several things
a supercomputer (1000 node beowulf should do)
a team of the best academics you can find
a lot of caffeine - to feed the academics
lots of cash - to pay for the massive amount of electricity used by the supercomputer
lots of time - a couple of years maybe
and some very precise information on how the random number was generated

slushy77
02-02-2006, 08:23 PM
one poker site i play on, uses random numbers generated by mouse movement on the customers' machines to add entropy to the random numbers generated on the server - impossible to work out, so the site is happy to publish some details of its RNG system.

most poker sites will be monitoring their network ports for unusual activity and will quite happily freeze accounts, cancel winnings, and contact authorities if they suspect someone of cheating

jkd
02-02-2006, 09:39 PM
one poker site i play on, uses random numbers generated by mouse movement on the customers' machines to add entropy to the random numbers generated on the server - impossible to work out, so the site is happy to publish some details of its RNG system.

That's an interesting way to implement some randomness... though one must wonder how "random" mouse movements are, considering that they tend towards points of interest on the screen. Points of interest = strange attractor? Would be an interesting exercise in chaos theory and the like.

gsnedders
02-02-2006, 09:45 PM
That's an interesting way to implement some randomness... though one must wonder how "random" mouse movements are, considering that they tend towards points of interest on the screen. Points of interest = strange attractor? Would be an interesting exercise in chaos theory and the like.
Well, if they use something like x*y, it could quickly get difficult, as you'd have to be correct down to the pixel.

slushy77
02-02-2006, 11:25 PM
The theory is that each person will move a mouse slightly differently, and position the client software in a different location on the screen. Factor in different screen resolutions, different types of mice, and the fact that there could be any number of users(more than 1) connecting to a particular server, hosting 1 or more games, you would have enough variation to generate true randomness.

jkd
02-02-2006, 11:43 PM
The theory is that each person will move a mouse slightly differently, and position the client software in a different location on the screen. Factor in different screen resolutions, different types of mice, and the fact that there could be any number of users(more than 1) connecting to a particular server, hosting 1 or more games, you would have enough variation to generate true randomness.

It's not that the coordinates wouldn't be different from each user. But in general, mouse-movement is goal-oriented, e.g. move mouse to click on button (unless the user is just bored, in which case circular movement might be engaged in, still somewhat organized however). Yes, the coordinates would differ, but overall, coordinates would be grouped with some sort of regularity to certain areas on the screen. A good RNG will produce very different numbers even for seeds close to together, so it's a non-issue for this specific case, but mouse movement in general strikes me as more similar to a chaotic sytem exhibiting strange attractor-behavior than purely random, unrelated sets of coordinates. I'm still convinced this would make an interesting study to perform.