Number of observations to guestimate slot return

I have been reading the forums for a while and mike deconstructs several machine to figure out the return of the game. My question is what sample size do you need to get a close enough guestimate of the return of a slot machine.

Thanks

James

With what level of confidence?

90-95%

I would be fascinated to see the Wizard, or someone else, deconstruct a modern video slot that has bonus rounds and/or different free spin scenarios. I would imagine it would take way too much effort to approximate any of the newer popular slots (ie. Willy Wonka, Family Guy, Flintstones, Big Bang Theory, Walking Dead, etc).

I'm thinking more on the lines of a three coin blazing 7 machine. Just a simple machine. How much of a sample would you need to make a reasonable guesstimate of the actual odds.

Zero.

The average hold based on denomination and geographic location will give you a reasonable enough guesstimate.

Quote: Jim123

I'm thinking more on the lines of a three coin blazing 7 machine. Just a simple machine. How much of a sample would you need to make a reasonable guesstimate of the actual odds.

This one is tough. The problem with a single-line three reel game is, each position on a reel does not necessarily have an equal chance of appearing on the line. The positions on either side of the main jackpot symbol on each reel probably appears much more often than the jackpot symbol itself, so there could be more "oooh, so close - the next one ought to hit it" moments.

This is just a gut feeling, mind you, but for a Blazing 7s, you probably need at least 5000 spins to have a reasonable idea of how often each symbol lands - and this assumes that each reel acts independently of the others (i.e. the probabilities on the third reel do not depend on where the first and/or second reels land). I know there used to be something in Nevada's gaming regulations that required this, but I can't seem to find it at the moment.

Quote: TomG

Zero.

The average hold based on denomination and geographic location will give you a reasonable enough guesstimate.

No, it won't.

So enlighten us what sample size would it take to approximate the payback percentage of a three coin one line blazing 7 machine or a double diamond.

Quote: Mission146

No, it won't.

It probably will within +/- 3% return in many cases. Unfortunately, I agree with others, most modern games, you're better off getting a par sheet copy.

Quote: tringlomane

It probably will within +/- 3% return in many cases. Unfortunately, I agree with others, most modern games, you're better off getting a par sheet copy.

I mean, that's fine unless he is wanting to AP Progressives or something like that, then that 3% is obviously going to be a tremendous difference. For my part, I'd much rather do the reel tracking (especially for a single-line three-reel game) for a few thousand spins and see if the associated probabilities/pays end up being in line with what the machines in that state pay.

Further, depending on denomination, it's also important to note that most published State Pays include VP returns, so if you are talking about a denomination where there are more VP games than others, (quarters v. pennies, for example) then you'll probably see slightly greater percentage disparities on slot machines v. state posted returns.

Quote: Jim123

So enlighten us what sample size would it take to approximate the payback percentage of a three coin one line blazing 7 machine or a double diamond.

I'd want to see 3,000-5,000 spins, or so. Probably the low end of that range would be okay. I'd just determine the probability based on per reel symbol frequency and multiply same by the pays and add it all up.

Quote: Mission146

I mean, that's fine unless he is wanting to AP Progressives or something like that, then that 3% is obviously going to be a tremendous difference.

Then he shouldn't be asking about "Guesstimates"

I think it would require a large number of "trials".

If you looked at a wheel for 3800/3700 spins, then it would be very unlikely that each number would come up exactly 100 times (normally 2sds you'd see figures between 80 and 120).

You're trying to work out how often the symbols happen and using the same idea except perhaps 5000 spins as your trial. So a rare symbol might have a 1 in 64 chance: on average it should show 78 times but (using 2sds) you might see 60 - 96). Using 50000 spins, the average is 781 and you might see 725-837; much closer.

If you knew the machine was using a 64-slot concept then for rare numbers you'd be able to guess 1 or 2's in 64 since their likely ranges don't overlap (1) 60-78-96 (2) 131-156-181. However larger numbers do overlap (10) 730-781-832 (11) 806-859-912, so sometimes you couldn't be sure.

Quote: Jim123

90-95%

How close to you need the return to be? I'm looking for a question like, "How many spins would it take to get the return to within x% with y% probability."