Remember the Fairplay campaign? That's right, the folks who were going to boycott video games in order to protest their high price. Well, now they've targeted gambling machines in the UK and found some disturbing results, namely that there is a gambling machine based on East Enders, the functional equivalent of having a Days of Our Lives nickel slot.
Additionally, they've discovered, using an emulator running the fruit machine software, that the software decides whether you win or lose before the virtual wheels stop spinning. There is a portion of the game where the player is allowed to inrease their winnings by cdoing another gamble. The machine displays a number, and the player chooses whether the next number displayed will be higher or lower than the one currently displayed. However, at this point, the machine has already decided if the player will win or lose.
To determine this the Fairplay folks saved RAM states of the game before choosing a number and repeatedly loaded them to find out if the machine was generating random results. What they found is that no matter what is picked, "higher" or "lower", the machine will generate a number that loses, if it has already determined that the player will lose. They are currently seeing if they can pursue legal action against the manufacturer of these machines.
Although this appears on the surface to be an instance of the machines cheating the player sout of money, the reality may be a little different. It's not like people never win at these machines, otherwise no one would play, of course. Well... less people would play more likely. The thing is, if the machine pays out at a percentage equivalent to what one gets from a mechanical slot (or any slot machine that is considered "fair"), then the user experience remains the same as far as being paid goes. So essentially, you put in your money, and it's random (as far as the player is concerned) whether you win or lose, the spinning wheels and flashing lights are just for show, a way of translating the player's win or loss into something the player can understand and enjoy. If you put in money and the screen just said "Win", or "Lose", it wouldn't be any fun, and nobody would play.
Missing from the results displayed on Fairplay's web site are how often the machine determines that the player is going to win, no matter what they pick. So, for example, say the machine generates number between 0 and 10 and displays a "5", if the machines has determined, at this point, that 50% of the time, the player will win, it's just as good as if it actually generated the number afterwards. If this percentage is the same is where the cruz of the matter lies. Fairplay may have a case if the percentage is less than what it would be if it were truly random.
Also, how they conducted their research seems to betray a basic misunderstanding of how random-number generation works. If they saved a RAM state and loaded it again, they are going to always get the same results. This is because no software can generate random numbers. In fact the term "random" typically means "unpredictable". The "random" numbers that machines generate are created by feeding a seed number, usually the time, or a constantly increasing value (1, 2, 3, etc.) into a logarithmic function, so that the number that comes out is unpredictable to someone who doesn't know the function, and therefore is just as good as if it truly were random. So, if one continually loads a game state by loading the RAM, it's going to generate the same number every time, unless the random function is running off a seed that changes, such as the time, or how fast the player pressed the button, or something like that. However, if the code is running on an emulator, as it was in this study, the machine is not getting the actual time, but what it thinks the time is, which keeps getting reset when the RAM is loaded.
The web page offers ROMs and emulators to download, as well as instructions on what exactly will happen when you do certain actions (thus proving the non-random nature of the game), but an actual machine, sitting in a pub may still be random as far as teh player is concerned, since it seems that the machine generates its numbers from a sequential seed number. The random element, then, would be how many times the machine has been played before, something the player would have no idea of knowing, unless the operator told them how long it had been sitting there, which I doubt they would divulge.
So, the test I would propose is to get the actual machines, run them repeatedly and see how often it pays out. Simple. If the actual machines pay out at a reasonable rate, then there is no problem. The perception of randomness is just as good as the real thing (which doesn't exist anyhow).
However, people do have a certain amount of faith in randomness, which is illustrated by the popularity of gambling machines. So this report may have the same effect as telling someone who'd taking sugar pills to cure their headache that they are swallowing a placebo. The results of this study will be very discouraging, I'm sure, to fruit machine enthusiasts.
BACTA may, in fact, have a case against Fairplay. Barring the copyright issue of Fairplay providing the ROMs to these games on their website, they are also providing a way for the players to cheat the machines. If the program follows the same sequence of events every time, a dedicated player could download the machines, exhaustively study the results and find patterns with the emulator, then take that knowledge to the pub and seek out similar patterns in the actual machine's performance. If they know when the machine is going to pay out, and when it is not, they will be able to make quite a bit of money. The manufacturer of the machines may have essentially shot themselves in the foot by not using a better random-number generating system.
Of course I have no idea how the fruit machines in question really work, and it will be interesting to see BACTA's (The organization of fruit machine vendors (yeah, I don't know what the acronym stands for either.) response to this accusation (though it is interesting to note that nowhere in their volumtary rules of conduct does it say anything about payout rates.)
(link via Metafilter)
