Slot Machine Random Number Generator VS. Bet Size

I've got a question regarding slot machines. Does the RNG change at all if you change how much you bet when your playing a slot machine? For example, suppose I start betting the minimum of a $1 on a slot machine. And then I switch to betting $2 or $3 or $4 or $5 etc, is there a different RNG algorithm that gets triggerred?

Payouts differ and so does the house advantage as you increase your bets. Do those get reflected on the same RNG or do they switch RNGs?

Or does this not make a material difference at all? (Rub the Buddha, click your heels three times, throw salt over the shoulder, make a sacrifice to the gambling g-o-d-s, and all in baby!)

Sorry to not speak out of actual, factual knowledge, but I'm pretty sure that you got it right in the last line. There is no need to adjust the RNG, as that is exactly why they change the pay tables for higher bets. The odds are usually improved/made worse by changing the number of lines and the pays, not the RNG. Once again, not speaking from solid data, but that's just how casinos tend to work. However, because of the change in lines/paytables, you definitely usually end up getting better odds if you are betting more money. That is pretty standard for casinos, too (i.e. European roulette is usually only available with $25 minimums, same with single-deck BJ, etc.)

There is only one RNG inside the slot machine and it is not in any way adjusted or subject to being adjusted based on coins or club cards or waitress tipping behavior or time of day or gloom of night.

I deleted this post, based on the answer to my question being..um...three threads down in the forum :)

A RNG is a computer chip or programming subroutine that picks a random number. No machine needs more than one, even if they wanted it to change things as you are implying.


Theoretically, the machine's Virtual Reel Strips could be changed based upon the coin-in, but I have a feeling that would not be allowed under Gaming rules.

On the other hand, the Virtual Reel Strips probably are changed when the player changes the denomination of a multi-denomination machine.

Thanks for all the replies.

Btw, DJTeddyBear, love that last quote.

Quote: DJTeddyBear

A RNG is a computer chip or programming subroutine that picks a random number. No machine needs more than one, even if they wanted it to change things as you are implying.


Theoretically, the machine's Virtual Reel Strips could be changed based upon the coin-in, but I have a feeling that would not be allowed under Gaming rules.

On the other hand, the Virtual Reel Strips probably are changed when the player changes the denomination of a multi-denomination machine.

Correct, the RNG is just that, a random number generator. What happens after that is the important thing. For example, one video roulette game might pay 30-1 for a single number, and another 35-1, but the RNG is the same.

I can affirm that sometimes when you go up in denomination the reel stripping will change, to strips with more high paying symbols. Much like if you move up in coinage on a video poker or video keno machine, you'll see the pay table go up. On some IGT machines you can tell when they do this when the last outcome displayed on the screen changes. Slots do not change the reel stripping based on the number of coins bet. That is why you're generally better off betting a small number of large coins, than a large number of small coins.

Wiz -

Conventional slot wisdom is that the lower the slot denomination, the worse the odds. And you confirmed that the same thing applies to multi-denomination machines.

And I realize that two identical looking machines sitting next to each can have a different return.



But is it safe to assume that a game in a single denomination machine would generally pay better than the identical game at the identical denomination in a multi denomination game, particularly when set at less than the maximum denomination?

To clarify, I figure it costs the casino a little more to purchase/lease a multi-denomination machine than a single-denomination of the same game. Additionally, a player on the multi-denomination machine, when set at less than the maximum, is preventing another player from playing it at the maximum, giving the casino an additional reason to 'punish' that player with a lower return.

I don't play slots enough to care, but I am curious.

Thanks.

It depends on the casino. When I did my slot machine survey I noticed some casinos (let's call them type A) had different returns for different denominations on the same machine. Others (type B) kept the same return on all denominations. In my experience such returns were always matched to the smallest denomination. For example, if a type B casino preferred to pay 90% on a 5-cent game, and 94% and a $1 game, and had a multi-denom (short for denomination) game, ranging from 5-cents to $1, it would set all of them to 90%.

If you're at a type A casino, then it shouldn't matter. The slot manager likely has a strict opinion about what return to give what denomination. If you're at a type B casino then you should be indifferent to playing a single-denom game and the lowest denom on a multi-denom game.

Unless you perform the test I mentioned in my last post, you won't know what kind of casino it is. In that case I would recommend that you assume you are at a type B property. In that case, if you must play a multi-denom game, play at the lowest coinage only. This is especially apropos if you're a 25-cent player, because some games will range from 5 cents to 25 cents, and others 25 cents to $2. In that case, play the 25 cent to $2 machine.

Interesting stuff.

Thanks.

Quote: Wizard

On some IGT machines you can tell when they do this when the last outcome displayed on the screen changes. Slots do not change the reel stripping based on the number of coins bet. That is why you're generally better off betting a small number of large coins, than a large number of small coins.

I performed an expensive "test" on an IGT Mystical Mermaid machine in a bank with "must-hit-by" progressives of $25, $100, $1000, $5000 in Wendover, NV. Line options on this machine were 1, 5, 9, 15, and 20, but I only played 20. Multiplier options were 1, 2, 10, 15, 20. Denominations were $0.01, $0.02, $0.05, and $0.10. Every penny bet on the bank of machines has an equal chance of triggering a progressive.

Although records of hands played were not kept, I am convinced that play at maximum bet (20 lines, 20x, $0.10) was no higher than the minimum payback required by Gaming. Apparently, the casino sees a patron who bets big to win the progessive as someone to extract maximum revenue from.

At low bets, this machine and similar machines were much better than at maximum. From my perception, play at 20 lines, 1x, $0.01 was more loose than at 20 lines, 20x, $0.01. From my perception, play at 20 lines, 20x, $0.01 had roughly the same payback as 20 lines, 2x, $0.10.

I obviously don't know where the changes in payback were, but I believe that the bet amount was the determining factor, not the bet denomination. If I am right, maybe I learned something from my ridiculous loss.

Dunno where to stick this comment so here it is:

If you're in a typical negative expectation, close to 100% casino game (say a 99.5% return online 6 deck blackjack game, where the Kelly criterion is obviously inapplicable) what effect would random bet sizing have on the total expectation (tho, of course, the otherwise fixed mathematical expectation remains unchanged). It seems like random bet sizing would generally enhance variation. I've been unable to determine much about this scouting the net.
Thx.

Quote: APEppink

Dunno where to stick this comment so here it is:

If you're in a typical negative expectation, close to 100% casino game (say a 99.5% return online 6 deck blackjack game, where the Kelly criterion is obviously inapplicable) what effect would random bet sizing have on the total expectation (tho, of course, the otherwise fixed mathematical expectation remains unchanged). It seems like random bet sizing would generally enhance variation. I've been unable to determine much about this scouting the net.
Thx.

Compared to what, betting a fixed minimum? If you want increased Variance, bet as much as you possibly can. If the mean bet of your, "Random bet sizing," results in a greater bet than your, "Regular bet," would otherwise be, then you'll have greater Variance assuming you would play the same number of hands either way. If the Random Bet Sizing results in a lesser mean bet than the amount you would otherwise be flat-betting, then you will have reduced Variance.

Random bet sizing in my parlance simply means varying bet size stochastically, indeterminately - randomly (or at least as randomly as a human can bet). The mean bet size seems irrelevant to me. Thus, if you're lucky, I'd think the random bet sizing could result in protracted winning even in a negative expectation game. Of course, that's true over the relatively short term even with constant bet sizing in a negative expectation game if you're lucky, if variance happens to be falling your way. Seems like there's two variances active, 1) that simply due to repeated game trial results, and, 2) that due to 'random' bet sizing. Intuitively that seems correct to me.

I'm no marthematician (BSME) but I'd like to see a rigorous math analysis of the situation. No luck finding such an analysis on the computer thus far.

It doesn't really require a rigorous math examination, because no set of simulations would be exactly the same.

Furthermore, let's take a look at a really small sample so I can show you what I mean. I'm going to use a PL bet at Craps (no Odds) to highlight my point so we don't have to get into the mess of doubling, splitting or things of that nature. Imagine you have random bet sizing vs. flat betting for a total of six bets, but the parameters are that the random bet size can be anywhere from $5-$100 vs. flat-betting $10.

$10-$10-$10-$10-$10-$10. Okay, so what range do you have there? You can either win $60 or lose $60, (at most) so you're looking at what I would call a $120 range of results, for the purposes of this example. You have a roughly 1.434% chance to win all bets and a roughly 1.7% chance to lose all bets. Over 96% of the time, something else happens that does not lie at the extreme end of those results.

Now, let's look at two different random bet sizings courtesy of our good friends at random.org/integers

66, 40, 32, 39, 7, 66---Mean ~41.67

30, 76, 76, 52, 85, 68---Mean 64.5

If you look at the bottom set of numbers, they sum to 387, which means you have the potential for a $774 Range of results, again, either losing or winning every one of the hands.

Okay, now using the Binomial Distribution, you have about a 31.2311% chance of winning three and losing three, thus, breaking even with the flat betting. I'm not going to go through those numbers in the above sets to determine whether or not there is any combination of wins and losses that would result in a break even, but I do know this without looking: it's not going to happen 31.2311% of the time.

That's where the mean bet comes in and how it is relevant to Variance. "Bold Play," is pure Variance (that's betting everything you have in one go) imagine if you have $500 and you are either going to bet $5 100 times or bet $500 once, you're not going to win $500 by betting $5 100 times and you won't lose that much either. It's technically possible, but eighteen yos in a row absurd, or close to it, anyway. So, which one has more Variance?

You have a negative expectation on all of the bets you make pursuant to the House Edge, as you have already shown yourself to understand. In the extreme long run, everyone will eventually lose the same amount of money based upon x amount of money bet, but in the short-term, your Mean bet is going to have a huge impact on the Variance.

If you want protracted winning in a negative expectation game, then you should use the Martingale, start with Table Minimum, and have a tremendous bankroll. There is a pretty good chance you will enjoy small wins for a very long time, but when you do hit your limit (by losing too many in a row) it will be devastating.

Going back to the small samples above, I don't see how you think random bet sizing is going to give you a better shot at winning. The only thing it does is make it not about how many hands you win or lose, but when you win or lose individual hands. If you look at the first set of random numbers, let's say you lose the first and last bet, but you win the middle four, you have lost overall as the two 66's combined exceed the sum of the middle four bets. With the flat-betting series of $10 bets, if you win four and lose two, then you are always ahead $10 regardless of where the wins and losses come in.

Your Expected Loss will always be the result of your total action multiplied by the House Edge expressed as a decimal. Your Expected Loss on an individual outcome is the amount bet on that outcome multiplied by the House Edge. The Mean amount bet not only affects the Variance, but additionally, it would affect your Expected Loss if you intend to do the same number of hands. If you are going to play six hands and your mean bet is greater than what your flat bet would otherwise be, then you will have a larger range of monetary results...but that works to the negative as well as the positive. If your Mean bet is less than the flat bet would otherwise be, then you'll actually have a smaller range of results, although, you may still find yourself in a position where the results of only one or two bets dictate whether or not you have a winning session, at least, in an extremely small sample size.

Mission is right - if you vary your bet between minimum X and maximum Y, your variance will be greater than if you bet a constant X but less than if you bet a constant Y.

Thx all.

I guess my contention is that in the short term (however defined) luck (variance) does play a non negligible role, tho, of course, the results converge to the house edge over infinite trials.
There's one other thing. My statistic's are rusty but I'm thinking of 'variance' in percentage terms rather than in absolute values (of money). Maybe 'std deviations' is the term I should be using.