[..]
Well generally no, putting aside the specific example which is US centric (I don't know if compensated machines with an equivalent % payout even exist in the UK) but the point in general, someone playing in a finite period with a machine which isn't memoryless has the added uncertainty that at any point in time that machine might have increased or decreased the EV of their bets.
In the case I gave, some guy trying to get diamond status via video poker by wagering a set amount and knowing in advance both his expected loss and the variance around that loss, if someone happened to have hit a royal flush (thus winning several thousands dollars) prior to him playing then an equivalent machine which was compensated (I suspect they don't even exist at that level) could well drastically affect the expected loss of his session.
On the scale you gave, it wouldn't. If he'd wagered a relatively small amount, it would. But not $50K or anywhere near that much. Even then, that scenario is not limited to compensated machines. Wholly random machines will also have temporary variances in payout percentage over periods of play due to random chance. No game rigidly pays precisely the target percentage at all times.
As for whether or not they exist, I don't know. In the UK, they would be category B1 machines or fixed odds betting terminals and I've never dealt with those.
It's worth pointing out that neither you, I nor anyone else here knows whether or not compensated machines are used in the USA. I'd be willing to bet they are because it's a simple and reliable method of achieving the target payout percentage.
It is IMHO a pointless (for the player) additional layer of uncertainty.
IMO it makes no difference to anyone. Players don't even notice or care. Some assume they can detect patterns and predict future outcomes. They're wrong, of course. It's a variation of the gambler's fallacy.
The edge for the operator, presumably over some defined period/number of spins?
It goes by VTP - value of total plays. The total amount staked. For a fixed stake game, that's proportional to the number of spins but for a variable stake machine it isn't, e.g. 1x £1 spin counts the same as 10x 10p spins as the VTP is £1 in both cases.
But for a player that arrives at some random point in time in order to play for a finite period unaware of what state the machine is in there is additional uncertainty. Aside from stating that the machine is compensated do these manufacturers make public the details of how their compensation mechanism functions?
I have no idea. I've never known anyone who cares, so I've never known anyone who's looked.
Now the memory of past payouts/compensation feature could in theory allow for a machine that sometimes gives an advantage to the player (I guess this is somewhat dependent on how much variance is allowed and how long the machine will go for before compensating etc..) - some sharper player observing such a machine over time might well take note of the past payouts and attempt to play at a point where they believe the EV has increased... in theory you could have a machine that becomes +EV for certain periods. In practice, given the rather harsh house edges of plenty of UK machines, I presume this isn't so feasible but rather, at best, the machines would mostly just give slightly more generous/less harsh -EV bets for a bit.
Every machine I've dealt with will be +EV for some periods,
including the wholly random ones. It's not particularly uncommon for a machine to pay out over 100% over a period of a week or so. The edge on our machines is as low as 8%. 6% on a few games, IIRC, but 8% on quite a lot of them. Generally speaking, the higher the stake the smaller the edge. Of course, if you look at small enough periods of time a machine can have a payout percentage way over 100%. For cat C and B3 machines, a payout of 5000% is possible over a short enough period of time. Stake £1 on a B3 machine, win £500 - 5000%. Stake 25p on a C machine, win £100 - 5000%. EDIT: Ooops, maths failure there
re: the motivation for them - I suspect it relates to the original marketing of them to the small venues whereby the payouts to the operator of them (who might well not be the owner of the venue) will have less variance, less faff in having to top them up etc.. if some punter(s) get lucky few times in a row etc.. in a completely random, memoryless machine you could in theory hit a sequence of events that clears it out of cash, in a compensated machine I guess you could design it to minimise or even essentially eliminate the chance of this happening.
That would be wildly illegal and as far as I know it never happens. It wouldn't be worth the risk for any legal operator, that's for sure. Maybe some secret underground gambling place run by criminals, maybe, but they would also have to have custom written software (and hardware in older machines) to make it possible to do at all. A business running in the open would have absolutely no chance of getting away with that. They would be caught, and soon, and be fined and shut down. The operators might be jailed too. The authorities do not mess about with this sort of thing - there are physical inspections on site, accounting inspections both onsite and remotely and the inspectors are one of an extremely short list of people who may not be denied or even delayed access to any place and any records except in the rather unlikely situation of an official with enough authority declaring the premises unsafe to enter, e.g. the rather unlikely situation of a gaming inspector turning up when the premises are on fire, the firefighter in charge can deny them entry.
Compensated machines
do not have less variance than wholly random machines. They
do not have "less faff in having to top them up etc". Machines being cleared out of cash happens regardless of whether the machine is compensated or wholly random because, and I am belabouring the point, which method is used to payout a target percentage makes no difference.
The factors relevant to how often a machine needs refloating are:
The float level set in the machine.
The frequency with which the machine is cashed up, as it will be refloated every time it's cashed up (obviously, you can't cash up a machine without taking all the money out of it and you will then need to refloat it).
How much the machine is played between one cashing up and the next.
The target percentage of the machine.
The stake on the machine, more specifically the stake relative to the float. A 10p machine with a £250 float will run out much less often than a 30p machine with a £250 float.
If the machine pays out only coins, then what will matter is how many coins are put in rather than how much money in total. If players put notes in and collect coins, the machine will run out a lot more often.
If there is a routing problem with the machine. If, for example, you have a machine that routes all coins to the cashbox it will run empty very often. Maybe a dozen times a day if it's being played often.
The biggest factor, of course, is whether the machine pays out cash at all. Most new ones don't - they print a ticket that can be put into a seperate machine to convert it to cash. That was done mainly because of the "faff in having to top them up, etc". With that system, you have a far smaller number of machines that need topping up, usually just 1.
The only machines left at my workplace that need topping up are the older Cat C and Cat D machines
that are compensated machines.
I suspect the main motivation for compensated machines is that it's easier to write the software for them and, very importantly, that software is already tested and approved. A developer can take their already approved gaming software, slap some new game software on top et voila! they have a new game. If that dev wanted to make a wholly random machine, they'd have to go through the whole testing, validation and approval process again and that's time consuming and expensive. Why would they bother, since it makes no difference?
