Hi, I’m a software engineer for a slot machine manufacturer
There is no easy answer to that question, as it depends on which gaming jurisdiction the casino is in. In the average jurisdiction, the average slot machine will have a programmable range of average payback, as an example 5 levels around 89%, 91%, 93%, 95%, and 97%. These paybacks are adjusted either by fiddling with switches on the mainboard (older, less common) or by applying a configuration program to the slot machine to set one of the other payback levels. Typically, payback rate is adjusted by changing symbols on the reels in a predetermined, mathematically-proven way, which the average jurisdictional authority has carefully inspected.
No machine I’ve ever seen - at least in the US or Australia - is capable of being changed remotely or instantly on the whim of a casino manager - in some jurisdictions governmental authority must be physically present to inspect seals, verify programs, and even themselves conduct the configuration changes.
Sounds plausible, depending on the beliefs of a particular casino management. Much like the practice of supermarkets cluttering up the aisles to make their customers wander more and foster impulse buys.
They’re published to the jurisdictional authorities which approve them for installation - anyone who cares enough to want the algorithms probably wouldn’t be the type to play the games anyway
The regulatory agencies have all the source code for review, and have verified and approved each game title. Mistakes can happen, as in anything - but intentional hinkiness gets (and has gotten people) jail time.
Both my company and the governmental authorities which have oversight on machines respond comprehensively to complaints made both by players (“this machine should’ve paid me!”) and by casinos (“this machine is paying out (too much / not enough)!”).
Some machines use EEPROMS, some use hard drives, or CD-ROMs, or maybe even CF cards for their programs. In both cases, gaming commissions verify the contents of the media, depending on their regulations. Some tamper-tape up the chips, the door locks, and the ports on the boards.
Late last night on CSPAN, there was a rerun of a program that ran earlier in the day, in which a Congressional committee was probing the efficay of electronic voting machines.
“Test, test and retest” seemed to be one of the mantras for keeping these machines unsullied.
However, the computer gurus who were testifying made it clear that it’s all too easy to hide bad stuff in codes. One told a story that went something liike this:
At Vegas there was a casino computer honcho - last name, Harris, I think - who progammed a slot machine (or several machines, perhaps) to payoff in a big way. The rigged jackpot was activated by the way the coins were fed into the machine (no elaboration whatsoever, here).
Okay, so his shill is playing the slot, feeds the coins in properly, and hits it big time. Ah, but the guy had no ID on him, so casino security escorted the shill to his room to get it. And who just happened to be there? Casino employee, Mr. Harris.
The guru’s point was how easy you can plant some very subtle bad shit in the code of a slot - or a voting - machine and that it might go unnoticed for years and years. Unless someone gets extraordinarily lucky.
Obviously playing a slot machine is a losing battle, but do you know if you are more, less, or equally likely to come out ahead or should I say less behind if you play more than one coin, line, ect. at a time? Seems to me the odds are going to remain the exact same, you will just speed up the process. If you get 95% of what you put in back, I would rather put in $100 and hour than $1000…
As an aside, refurbishing a slot machine to be an e-voting machine would be an interesting exercise - they’ve already got tamper-proof hardware, battery-backed memory, printers for receipts or auditing, touch screens, and secure networking. Insert coin to vote. (click) You lose!
Yes, I remember. Here’s a link. For what it’s worth, there are procedures to look for what may be termed “layer violations” - essentially, global variable references or function calls to sections of the program having nothing to do with, say, coin acceptance, wouldn’t belong in coin acceptance code.
This Harris character, the article notes, got his inside information and leveraged his employment at the gaming control board. He wasn’t acting to cheat the player on behalf of the casino or the industry, he was acting to cheat the casinos on behalf of his own pocketbook.
Usually the odds are the same. It depends on the particular game, however - some games require you to play the maximum bet for a shot at the biggest wins, and the overall return-to-player for non-max-bet play is a few percentage points lower. For one to four credits bet at video poker you may get a 250-1 return for a Royal Flush, but an 800-1 return (4000 coin) for five credits bet, or a 1000-1 return for ten credits bet. So the return to you might be:
1-4 creds: 96.5%
5-9 creds: 98.2%
10 creds: 99.3%
(these are not real figures, just representative examples)
Usually this is pretty clear from examining the particular game’s pay table. Most likely, if it’s sort of a weird game and it’s not obvious whether your payback chance is affected by your style of play, then the return percentage will not vary by more than a half-percent (owing to a New Jersey regulation, if I remember correctly - manufacturers tend to build only one version of any given program, compliant with all jurisdictional rules).
AmbushBug, I’ve also heard that the electronic slot machine algorithm includes a routine to encourage long-term play like this:
When a player sits down and plays a few spins, the machine will start paying him off somewhat more frequently than average, building up his bankroll. The player says “Hey, I found a loose machine!”
Then, at some point, it starts taking back the winnings but, since the player knows that it’s a loose machine, he doesn’t stop and keeps on trying until he ends up putting more money into the machine than he originally planned.
Is this truly part of the algorithm? Or just an individual’s perception of the normal runs of statistical probability?
Just the perception. The random-number generators have no memory - they’re always running in the background (100+ times per second, per NV regs) and when you hit the button to start a game fixes the spin of the reels or the shuffle of the cards.
Games are designed differently - some games can have dry spells between big wins, some can dribble little wins at you all day, some games trigger interesting feature side games or free spins more or less frequently. A “highly volatile” game tends to run in streaks - lots and lots of losses and tiny wins, and a surprisingly large one once in a while.
Class 2 games, aka “video lottery terminals”, found mainly in some Indian casinos and racetracks in the US, work slightly differently. They aren’t my area of expertise, but to my knowledge they have a pot of wins and losses (which may be shared amongst all the machines in the casino, simultaneously) and the randomization entails your play drawing one from this pot, and that’s what you get. The pot might contain very few maximum wins, and once one is hit, the likelihood of hitting another is reduced.
AmbushBug, I must say you have a cool job. Is there any way of knowing which “design” a machine has – is it a “big win only”, or “little dribbler”? Do the casino personnel even know? Do they ever switch an individual machine from one payout pattern to another?
I assume that video poker machines aren’t as programmable–that they mimic the same probablities that you would see with a physical deck of cards. You can’t program one to deal fewer pairs but more straight flushes. True?
The only way I can tell, is by first examining and understanding the pay tables and the nature of the bonuses you may get during the game, and then by actually playing the game. It’s very difficult to tell just how well a given game will play to casino customers; slot players themselves are different the world over and what is popular in New Jersey may not be in Vegas, Mississippi, Australia, or Canada.
Right. The return-to-player on for video poker is based on how much is paid back for the wins, and the frequency of the wins is only affected inasmuch as the “best play” for a deal changes on the payback. If you’re dealt AA886, and 2 pair pays twice what jacks or better pays, you’ll keep AA88 - but if 2 pair pays even money same as jacks, and 4 of a kind has a bonus pay, you’ll just keep the aces and go for maximizing your chance at improving the aces alone.
I think I’ve heard – probably on the Travel Channel, which is All Vegas All The Time – that slots are programmed to tease. That when a losing arrangement is selected by the random number generator, it’ll put the symbol that you need to make it a winner one click above the payline. Any truth to this?
I used to work as a “mechanic” (slot machine technician) in northern Nevada and what AmbushBug says jibes with my experience. Out in the field, at least in Nevada, it is really, really difficult to get a look at any of the firmware used in the machines. Sometimes the proms crap out so casinos like to have spares on hand. The Gaming Commission will allow prom duplicators that go direct from the master to the ones being burned, but will emphatically not allow one that can read a prom and store the bits in a file for burning new ones later – too much chance of mischief.
Every quarter someone from the Commission shows up at a casino and some of the gaming machines are inspected at random. Part of the inspection is to check that the payout prom’s program number matches the record kept in the slot office, that the program number has been approved by the Commission, and that the checksum of the prom matches the records the Commission has. We bought a bank of ten used keno machines and two of them came under this inspection. One of them had a program that was allowed once, but was no longer. All ten of the machines were then examined, and three had that disallowed program. Those three had their chips pulled and were dark a couple weeks until we could get new PROMs from the manufacturer. Whatever the reason that program was disallowed was subtle; we hadn’t noticed any difference between the three “bad” machines and their brothers.
From a magazine that the hotel leaves in the rooms, along with a TV guide and room service menus…the name of the mag escapes me…Atlantic City News or something like that…It stated in an article, that, ONE slot machine garners an average profit of $80,000.00 per year. ONE MACHINE!. Do the math.
Nope - intentional “near-miss” skewing is generally frowned upon. It’s possible, however, that symbols on physical reels or secondary game spinners (coughwheel of fortunecough) are weighted somewhat so that the highest value symbol on a (for example) 25-position reel might only come up once in a hundred spins. In such a case, symbols on either side of the high value symbol will be weighted differently and could come up 3-7 times as often as their adjacent high-value symbol. No game I’ve ever made (one) which had weighting like this worked this way; if either symbol next to the “super 7” came up on the payline it was because that symbol had the same weighting as the “super 7” and it wasn’t the other instances of those symbols on the reel which have weightings six times as large.
There’s a lot of scrutiny on this kind of thing by the authorities, because people do have the perception that the machines are programmed to tease.