What jurisdiction were you in, Mangetout? This certainly was not true in Nevada, where I was a mechanic in the early nineties. Really ancient slots – like fifty years old at the time – did have tubes in them. When a minor jackpot was won, the bottom of the tube would open and the coins dumped out all at once. For bigger jackpots, two or three tubes would empty. Subsequent coins would then refill the tubes until they were topped off, then start to go into the box in the bottom of the machine again. Since machines so old were not installed where I worked, I don’t know what happened if a jackpot was hit and the tube(s) were not full.
Modern machines have a motor-driven hopper rather than tubes. Much higher in capacity, they work the same way – coins go into the hopper until a sensor detects the hopper is full, then the coins are diverted into the drop. The difference is that on a payout each and every coin is counted. If the machine has been on a streak and the hopper is depleted before the payout finishes, everything stops and a light on the machine flashes to attract the attention of a key or mechanic. Forms are filled out and a sack (or rarely, two sacks) of coins are brought out and dumped into the hopper to let the payout finish. This is no big deal; when the small club I worked in was busy, this would happen several times an hour.
However, there is no mechanism in the machine that keeps track of the payouts. The processor in the machine doesn’t decide, “Hey, I’ve had four big payouts in the last fifty games, I’d better get tight for a while.” The Nevada Gaming Commission wouldn’t allow it. Records are kept for the coins removed from the drop (since they represent the profit for the club) and the number of fills. The number of coins in are also known from meters in the machine. A little leeway from the expected percentage is allowed, but if it gets too large – in either direction – the machine will be examined as to why. That percentage is averaged out over hundreds of thousands of games per week, but the chance of each individual game is just that: chance. Whatever happened in the past has no influence on where the reels stop now.
If you think about it, the last thing a casino wants is predictability, either to play a machine heavily because it is about to win, or avoid one because it won’t.
I don’t think that necessarily follows. It depends on what “99 percent payout” means.
It could mean “We’ve designed this machine so that it is guaranteed to pay out exactly 99 percent every 24 hours,” (or every week, or every year, or whatever.)
or it could mean “On average, this machine pays out 99 percent.”
If it means the former, then yes, you could look for low-paying machines and play on them in order to gain money as they catch up to their payout.
If it means the latter, then there is no strategy applicable.
I’m talking about fruit machines installed in UK pubs (used to have a maximum payout of £3 when I was working on them, I think, but that has subsequently been increased) - they had one tube for each coin denomination that could be paid out and at the bottom of each tube was a solenoid that would eject one coin at a time (so a big win would be accompanied by a stream of coins and a loud chunk-a-chunk-a-chunk noise). the tubes had (IIRC) two sensors - one that would be triggered if a tube was one third full, the other at two thirds. If the tubes were completely full, the coins would be diverted down a chute into the mixed overflow coin box in the base of the machine (and usually, it was only this coin box that would be emptied by the cash collectors - the tubes would be left as they were)
Good point. However as has been pointer out by ultra filter the length of time that it takes for good random number generators to repeat can be extremely long. I say the went to the random well three times for every pull (once for each wheel) then it would take trillions of pulls to get through the whole sequence. So while they might be pseudo random the time scale for a due machine to achieve the 97% payout is much longer than the life of the machine.
If you apply a small amount of cleverness to designing your gambling machine you can add a lot of uncertainty for the world around you. For instance in the case of the slot machine every millisecond choose a new random number even if the play has not chosen bet at that time. Now the fruits that come up depend on exactly when the button was pressed.
Here’s how you do it. Find a nickel machine. Buy $50 in nickels, that is 1,000 coins. Cycle them through the machine in question once. Then ask the lady at the cage for cash.
I regret to report that casino machines (in Vegas) no longer accept coins. The slots areas as strangely quiet nowadays, not nearly as much fun.
Why do you think 1000 coins is enough to find this out? You really only get 200 tries because the big bonuses only payout if you play 5 coins at a time. To really figure if the odds are correct you need to play enough times that you would expect the the high payouts to come say 5 to 10 times. Much less than that you cannot really say with any confidence if it is randomness or bias that your payout is 94% instead of 97%.
I’m not saying it’s possible in practice to find a machine that hasn’t paid out its programmed percentage, only that it’s possible in theory.
The numbers generated by the machines aren’t random. They don’t even try to simulate randomness, because the machine actually is programmed to return a certain percentage of what’s played. You can’t both 1)program a certain payout percentage, and 2)have random numbers decide what gets paid out.
I mention the procedure recommended by Scarne. While more trials would be useful, at some point I wander off to the strip clubs, so 1,000 trials is all I have time for.
Sure. The machines I described above though weren’t even pseudorandom - they had a pseudo random element, but they were programmed to maintain an average payout as closely as possible - I believe it was 76.5% at the company where I worked (the law stated that they must pay out at least 75%) - the machines contain a microprocessor that runs the whole game (the reels are driven by stepper motors, so they are absolutely controllable) and this did far more than just dish up a random selection of plays that just happened to come out near the desired percentage payouts
Here is an example of a typical UK fruit machine - incidentally, along with a discussion of one of the methods the machine uses to manipulate the wins.
Sure you can; at least, you can achieve low probability of deviation of more than “epsilon” from the desired payout percentage, this low probability of failure being on the order of the probability that your knowledge of the laws of physics governing circuit design and all that is wrong anyway.
OK…simple example. RNG spits out numbers between 0 and 9. You enter $1 and hit the button and a number is generated. 0-5 pays off nothing. 6-7 pays $1; 8 pays $2; 9 (jackpot!) pays $5.
Random numbers have decided the payoff. Average payoff will = 90%.
Specifically, for any positive epsilon and delta, no matter how small, the probability of a payoff percentage which differs from 90% by more than epsilon drops below delta as the number of plays gets sufficiently large. Which is perfectly well captured in ordinary language by saying the machine is programmed for a 90% return percentage. Any other approach to achieving a 90% return percentage, like ones which use machine memory to “play catch-up”, so that different plays are not probabilistically independent, is just grossly silly.
Now, for all I know, machine manufacturers and/or the laws governing this sort of thing actually are grossly silly…
They achieve that exact payout percentage precisely by generating the numbers as randomly as is humanly possible. To do it any other way would be both illegal and massively unprofitable for the casino.
Well, if they could actually pull it off some other way, it would be equally profitable; 90% payout is 90% payout, the only potential difference being the number of players. Of course, the danger is, attempting to do it in a way which can be “gamed” might simply fail to attract a steady stream of players: if almost everyone knew how to game the system, then almost no one would play except when the machines were “due”/“hot”/whatever. They’d just sit stagnant collecting dust. But ignoring that (perhaps a big ignore), if the casino can somehow get a steady stream of players for their machines either way, the stupid way and the proper way would be equally profitable.
I know nothing about the law, so I defer to you there. At the very least
Machines WILL pay out their programmed percentage. They don’t have a “chance” to pay out their programmed percentage based on whether random numbers turn out close to what is expected.