Epic slot rate of return - we got 28.5% on Mad Hatters slot pull!

[sjbdtz]

Err, no.

 

It depends on the magnitude of the difference.

 

If you're looking for a 5% difference or so then yes you will need thousands of data.

 

If the difference is larger you may only need 100 or so.

 

And if you do see a significant difference, then there is a difference, regardless of the amount of data.

 

The number of data is only important if you see NO difference, because you can't rule out that more data would lead to a significant difference.

 

You'll usually see something like "although no statistically significant difference was seen, the study was powered to detect a greater than 20% difference, so a smaller difference is not ruled out."

 

Hmmm.....are you just talking or do you have scientific credentials? (I'm not accusing you of just talking! I'm asking for data!)

 

I have a hard time swallowing that 700 concurrent observations are insignificant and you need hundreds of thousands before it is meaningful, especially with such a large difference from any sort of reasonable return, but I've been shouted down and NCL casino management emphasized the "hundreds of thousands" figure. Although I AM trained as (and practicing) scientist I am willing to admit that I don't have a lot of independent knowledge about casino operations, just some knowledge of odds. The NCL casino folks referenced the wizardofodds website and Slot Player magazine. I don't play slots ordinarily because I prefer games where I have a least the illusion of control over wins/losses, such as blackjack.

 

I'd love to hear from someone with some factual knowledge and not supposition! If you have it then that's great!

 

 

If you cross the road 700 times, and only get hit by a car once you can say that you had a 1.4% chance of getting hit.

 

If you cross the road 200,000 times and only get hit once you can say that you have a 0.0005% chance of getting hit.

 

If you cross 14 million times, and only get hit once then the odds are again very different.

 

 

this is exactly what Ellen is asking.... we played 700 rounds.....with 28.5% return....would 200,000 rounds produce same odds? what about 14 million rounds?

 

The fact is that in a game where we know that jackpots are occasionally won (cars occasionally hit people), then the odds increase as the number of opportunities go up. In other words, the more you cross the road the more likely it is that you will get hit (statistically) or conversely the more you play at a slot machine, the likelier you will hit a jackpot.

[sbark1]

Do this same exercise for 1 million more pulls and you will hit your expected rate of return...your sample size is ridiculously small.....:)

[erdoran]

If you cross the road 700 times, and only get hit by a car once you can say that you had a 1.4% chance of getting hit.

 

If you cross the road 200,000 times and only get hit once you can say that you have a 0.0005% chance of getting hit.

 

If you cross 14 million times, and only get hit once then the odds are again very different.

 

 

this is exactly what Ellen is asking.... we played 700 rounds.....with 28.5% return....would 200,000 rounds produce same odds? what about 14 million rounds?

 

The fact is that in a game where we know that jackpots are occasionally won (cars occasionally hit people), then the odds increase as the number of opportunities go up. In other words, the more you cross the road the more likely it is that you will get hit (statistically) or conversely the more you play at a slot machine, the likelier you will hit a jackpot.

 

Ummm.....actually I don't care about jackpots, it's more of an academic exercise for me. What percent of the money that gets put into a slot machine gets returned? I don't care whether it's as a jackpot or as $1 paybacks. It's a very fair question. Asking the percentage of jackpots is (for me) meaningless because there are too few of them to be statistically significant. I'm more curious about the total dollar amount paid out over a large number of pulls vs the total dollars put in. The probability of hitting a jackpot is far too low to be able to determine/verify experimentally!

[gb58]

A guaranteed return of 90% means you are guaranteed to lose 10 cents on every dollar. Anything else is a bonus.:D

[sjbdtz]

Ummm.....actually I don't care about jackpots, it's more of an academic exercise for me. What percent of the money that gets put into a slot machine gets returned? I don't care whether it's as a jackpot or as $1 paybacks. It's a very fair question. Asking the percentage of jackpots is (for me) meaningless because there are too few of them to be statistically significant. I'm more curious about the total dollar amount paid out over a large number of pulls vs the total dollars put in. The probability of hitting a jackpot is far too low to be able to determine/verify experimentally!

 

 

But if you're excluding jackpots from the average winnings, then the percentage would change dramatically....as would the whole notion of slot machine gambling.

 

People gamble on slots for the enjoyment of the $1 wins, AND AND AND the chance to win a jackpot.

 

Over the course of the hundreds of thousands of spins, the jackpots DO get won, and those winnings form part of the 90% return which the machines give.

 

You could conceivably sit and play 10,000 hands and win nothing...and the next person sits down & wins 10,000 in a row.

 

Over the course of 20,000 hands, you have a 50% hit ratio...even though over the first 10,000 it was 0.

[erdoran]

But if you're excluding jackpots from the average winnings, then the percentage would change dramatically....as would the whole notion of slot machine gambling.

 

People gamble on slots for the enjoyment of the $1 wins, AND AND AND the chance to win a jackpot.

 

Over the course of the hundreds of thousands of spins, the jackpots DO get won, and those winnings form part of the 90% return which the machines give.

 

You could conceivably sit and play 10,000 hands and win nothing...and the next person sits down & wins 10,000 in a row.

 

Over the course of 20,000 hands, you have a 50% hit ratio...even though over the first 10,000 it was 0.

 

No, I am not excluding jackpots at all. My method is

 

• How many total pulls were made?
  
• How much $$$$ total came out of the machines - I don't care whether it was all a single jackpot, or $1/pull--I just want the sum of the winnings
  
• How much $$ was put into the machine?
  

That's all! $out/$in = % returned.

 

• 
  

[DrD]

Hmmm.....are you just talking or do you have scientific credentials? (I'm not accusing you of just talking! I'm asking for data!)

 

 

Sure, I'm a physican and published researcher (although along time ago) and have taken statistics. (I try not to talk about things I don't know about, refreshing huh? )

 

I am not an expert on biostats, but for example in an article I'm reviewing now, 31 of 140 patients (22.1%) in study group had a certain outcome, and 73 of 159 (45.9%) in placebo group had the outcome, and that was statistically significant, with a p value of 0.001, meaning those results are only 0.1% likely to be due to chance alone.

 

so I would guess that if 700 pulls resulted in a 28% return, and assuming the normal rate is 70%, that this would be highly significant.

 

hopefully someone with a stats background can calculate a p value for us.

 

but I'm quite sure that what you were told is false from a stats perspective, larger numbers of data are needed only when differences are small, not great.

 

For example, if you have a loaded penny, such that it will come up heads about 90% of the time, you might only have to flip it 20 or 30 times to prove it is loaded, as coming up heads say 25 of 30 is unlikely due to chance alone.

 

on the other hand, if the penny is loaded to come up heads 55% of the time, to prove it is loaded you might have to flip in thousands of times, since coming up heads say 60 times out of 100 could easily be chance alone.

 

so the smaller the difference, the more data you need.

 

In the last example, you might conclude that there is not greater than a 10% difference in the chance of heads vs. tails, but you can't rule out a smaller difference vs no difference.

 

clear as mud???

[garycarla]

Spend a few thousand and making an assumption about rate of return is illogical.

 

Not knowing this specific machine, I will make an assumption. Please allow that this machine has a $10,000 jackpot.

 

So, if the only payout this machine ever makes is the $10,000 payout, and you assume the rate of return is 80%. So, people would place bets totalling $12,500 and then somebody won $10,000, the rate of return would be 80%. So, with that said, you could easily place bets of even $12,000 and not win a single bet. Not one dollar. And you could assume a ZERO percent rate of return at that point. Bad assumption.

 

Again, if the only payout was the $10,000, the math says you have to bet $12,500 to win 80% of the bets or $10,000.

 

To make matters worse, in reality, it makes lots of small payouts along the way. So, to eventually get to the $10,000 payout, people might have to bet $20k or maybe $30k to get there.

 

The assumptions about a rate of return based on the small sampling are downright wrong!

 

(by the way, I do believe shipboard payouts are very tight)

[Dadmech]

That's why they're called "Games of Chance" I love to gamble, not as much as my wife, but I go to enjoy my time there, play as long as I possibly can, and any winnings over what I have broght with me go in my pocket for another day! Just my 2cents worth...;)

[sjbdtz]

No, I am not excluding jackpots at all. My method is

 

• How many total pulls were made?
  
• How much $$$$ total came out of the machines - I don't care whether it was all a single jackpot, or $1/pull--I just want the sum of the winnings
  
• How much $$ was put into the machine?
  

That's all! $out/$in = % returned.

 

 

So by extension, the fact that I jaywalked successfully twice yesterday means I can extrapolate the anyone can jaywalk anytime?

 

What you heard is that for the math you want to do ($$ out / $$ in), you need a larger sample.

 

that's it, that's all. 700 pulls is not significant. Heck, to contextualize it.... it was 2 machines for less than one hour...with 40 interruptions each of 10 - 20 seconds while the next puller sat down.

 

So.... rounding the time to 1 hour, we had 3600 seconds of playing time. Reducing that by the 40 seconds x 40 players for changing & seating (1600 seconds) that's 2000 seconds of playing (or 33 minutes).

 

Have you ever sat in a casino which had Wheel of Fortune machines and seen them go for 33 minutes without a SPIN? I have. Often.

 

On those PARTICULAR MACHINES, the SPIN element is where the payouts are.....NOT INCLUDING the jackpot.

 

So the 90% average payout includes those $40 - $1000 wins on the SPIN, as well as the occasional $20k jackpot....yet the time interval was very short to be evaluating the frequency with which one can expect to win.

 

 

Ellen, here's the thing I have the biggest issue with:

 

Statistics are most valuable when used predictively. Knowing that for the 33 minutes we played, we got a 28.5% return means nothing to anyone but the 80 of us.

 

To everyone else, including T2C1 who immediately won on the same machine, the rate of return has differed.

 

It seems to me that you are trying to extrapolate data which should be seen in a vacuum, to be meaningful across the casino.

 

The reason I think this, is that I clearly recall you saying immediately after the slot pull: "Everyone should post that NCL's Slot machines only return 30%". Had you said "Everyone should post that WE ONLY GOT 30% during the 33 minutes that we played the Wheel of Fortune machine PORT SIDE in the smoking section" that would have been FAR more accurate.

[erdoran]

So by extension, the fact that I jaywalked successfully twice yesterday means I can extrapolate the anyone can jaywalk anytime?

 

What you heard is that for the math you want to do ($$ out / $$ in), you need a larger sample.

 

that's it, that's all. 700 pulls is not significant. Heck, to contextualize it.... it was 2 machines for less than one hour...with 40 interruptions each of 10 - 20 seconds while the next puller sat down.

 

So.... rounding the time to 1 hour, we had 3600 seconds of playing time. Reducing that by the 40 seconds x 40 players for changing & seating (1600 seconds) that's 2000 seconds of playing (or 33 minutes).

 

Have you ever sat in a casino which had Wheel of Fortune machines and seen them go for 33 minutes without a SPIN? I have. Often.

 

On those PARTICULAR MACHINES, the SPIN element is where the payouts are.....NOT INCLUDING the jackpot.

 

So the 90% average payout includes those $40 - $1000 wins on the SPIN, as well as the occasional $20k jackpot....yet the time interval was very short to be evaluating the frequency with which one can expect to win.

 

 

Ellen, here's the thing I have the biggest issue with:

 

Statistics are most valuable when used predictively. Knowing that for the 33 minutes we played, we got a 28.5% return means nothing to anyone but the 80 of us.

 

To everyone else, including T2C1 who immediately won on the same machine, the rate of return has differed.

 

It seems to me that you are trying to extrapolate data which should be seen in a vacuum, to be meaningful across the casino.

 

The reason I think this, is that I clearly recall you saying immediately after the slot pull: "Everyone should post that NCL's Slot machines only return 30%". Had you said "Everyone should post that WE ONLY GOT 30% during the 33 minutes that we played the Wheel of Fortune machine PORT SIDE in the smoking section" that would have been FAR more accurate.

...and that's why I posted exactly what I did, stating it was the mad hatters slot pull! And came back with the info from NCL about the hundreds of thousands of samples before having a valid result.

 

My instincts still say that 700 samples are significant, but I've also posted that I don't know much about slots!

[sjbdtz]

...and that's why I posted exactly what I did, stating it was the mad hatters slot pull! And came back with the info from NCL about the hundreds of thousands of samples before having a valid result.

 

My instincts still say that 700 samples are significant, but I've also posted that I don't know much about slots!

 

 

fair enough.... I guess I err on the side of 700 samples are not significant. :p

[garycarla]

fair enough.... I guess I err on the side of 700 samples are not significant. :p

 

700 samples are very insignificant. Do the math of what the machine must take in to pay back the 80% or whatever.

[Another ship trip]

700 pulls is incredibly insignificant. Hundreds of thousands of pulls may start to show results that could be considered statistically significant but probably needs to be looked at over a very long time. Why do you think folks come back every day from Vegas saying Casino "X" was great and paid out alot and the next person says the same Casino "X" was tighter than Fort Knox? It is because slots pay on a variable reinforcement schedule that cannot be timed nor predicted. (that is also what makes them addictive). If you pull the aggregate of all the NCL fleet over a year of all their slots, then maybe you could get a MEANINGFUL statistic at that point. The short term is meaningless for determining long term payout. The stats listed earlier about Vegas in the post are accurate and are computed over the entire region and probably over a very extended amount of time. Believe me, the casinos know what their overall average is on their various denominations of slots.

While cruise ships in general are tight, I have seen dozens and dozens of jackpots of 1000.00 and more be hit at sea over the many years I have cruised. Luck of the pull!

(When video poker is concerned, there is also something called "perfect play" which enters into how much the payback can be that does not affect regular slot machines.)