So how exactly does this work?

[Railroader Travis]

Community,


Now that I have really gotten into the slots stuff and have learned so much, I'm still trying to solve just how exactly slots determine payouts and how they maintain a payout percentage profit for the casino.


My main confusion stemmed from the fact that a slot has to maintain a payback percentage. 95% so let's say it takes in $100,000 it should of paid out $95,000 at this point.  Well then I go, wait a minute... Don't slots especially IGT S2000's use RNG (random number generator)? If so, that would make all wins completely random, therefor would not follow a particular payback percentage. I have heard of slots paying multiple top prize jackpots in the same day, maybe to the same person back to back, or by different individuals... OK, so then I'd assume well it probably won't payback for a long time when it's done because it will be trying to take in money to recover and equal out to it's payback percentage....  So is it random with an RNG, or is the slot really an evil little child who plans his attack?

[mikec200]

To quickly explain, the payback percentage is based on statistics collected on a simulation of typically  10 or 100 million simulated plays of the machine given a certain virtual reel configuration and a certain RNG algorithm.  The machine isnt maintaining any specific amount, its the stats foe this game config and rng that over a long period of time will even out wins and losses to yield the payback percentage.  You can have wild swings in short periods of slot play like nothing paid out over 50 plays or back to back handpays.  All of this is possible given the stats in the short term, but long term the game and rng design are tested with 90% confidence to reach the programmed payback percentage with in a very narrow band. This is a high level explaination with getting into a ton of math, but hopefully clears it up a bit.

[CVslots]

Also, that 95% is based and tested on 10,000,000 pulls...I don't know the average pulls the busiest machine in a casino gets, but I'm thinking it would take 2-3 years to hit 10M.

[mikec200]

I dont think the theoretical payback percentage as shown in the par sheet is based on actual pulls, its simulation based on the rng and virtual reels. Now Im sure the manufacturers keep tabs on real world performance but with all the games and multiple programs in each game each with its own par sheet there is no way to actually test 10 or 100 million pulls without doing it in computer simulation.  I wrote my own sim of a machine using actual virtual reel layouts and pay combos and ran 10 million pulls using a standard rng in C#, so nothing fancy.  Only takes a few minutes on an average pc and the results vary, but are fairly close to the par sheet reported values in the 90% confidence interval.

[mikec200]

Also I recently got my hands on some WOF parsheets with 512 virtual stops.  Over 134 million possible combos.  If they test 10 to 100 million pulls on a 72 stop virtual reel with 373,0000 combos, I'm very interested what they test to with 512 virtual stops.  I say that because they dont mention it in the par sheet, but I would venture a guess of no less than 10 billion, possibly 100 billion or more as the standard test.  That has to be simulated... 

[rickhunter]

They are simulations, it wouldn't make sense for them to be actual pulls.  I don't know how they determine the sample size for the PAR sheets, but doesn't it say on the PAR sheet how many pulls?  It did on the S+ sheets.

[mikec200]

Most of the par sheets I've seen do, but most of them have also been smaller virtual reels.  The WOF sheets were 512 virtual stops and had none of the usually info regarding the payback percentage over increasing numbers of plays upto 10 or 100 million in  90% confidence. To me simulation is the only practical way to generate these sheets, and I would think the manufacturers then keeps tabs on real world results for a secondary confirmation.