Payout percentage in MG in the long run...

What would the payout % be in + 250.000 spins?

  • 90% or lower

    Votes: 13 35.1%
  • 91%

    Votes: 3 8.1%
  • 92%

    Votes: 3 8.1%
  • 93%

    Votes: 1 2.7%
  • 94%

    Votes: 7 18.9%
  • 95%

    Votes: 8 21.6%
  • 96% or higher

    Votes: 2 5.4%

  • Total voters
    37

steinhaug

Senior Member
Joined
Feb 21, 2006
Location
Norway
It happends to be I have wasted alot of time with "practise-play" the last month(s), and I have buildt a huge dataset with playcheck data for the entire period. I have passed 239765 spins as we speak, most on maximum wager. So since funmode is identical with real mode, I thought I'd poll the payout percentage for you all. I will post the results when there are some votes for what you think, ;)
 
I'm guessing in the 80's :D

Just a thought... if they were all max lines, could that be multiplied? i.e. 239765*number of lines = 3,596,475 if all were 15 line slots this is a large sample imo
 
Oh yes, it's a large sample - and it SHOULD satisfy the hardest people in here screaming for lucky streaks and rules of large numbers and chaos theory, :) It's a few 100 hours atleast of gameplay here, so yes - it has eaten some work time for sure, hehe.

But no use in sticking with the data myself - after all it's no secret data, atleast I never saw an EULA or other small print stating that you are never to share playcheck data with other people. It will be fun to see what the poll will show in some days.

At the moment the machine is spinning in auto-play for full machinery, :D Collecting data, collecting data... Ahhh... Sweet... If I could only gain what I have lost in this experiment I would at the moment be -1,817,071.96 richer... The sweet smell of money you could have lost!
 
I've finally given up on MG. Too much money lost, too many unforgiving losing sprees. I've had some good, exciting wins there too, but they still don't compensate for the amount I've lost. The biggest problem is that losing runs like that are absolutely no fun, so the whole process isn't as addictive as the business model needs it to be. So I've found giving up on them and closing my accounts pretty easy. I've also given up on Inetbet for the same reason.

3dice on the other hand - fun slots (though not enough of them), good payouts so far, excellent customer service... The test will be the payout percentage. Of course I don't expect to win in the long run, but if I get entertaining game play, some good wins and the chance of some big wins, I'll be a customer for ages. Long, barren tedious losing runs and I'm off...
 
I've finally given up on MG.

Watching +250.000 spins you see alot, and I have seen some "strange" patterns, however it seems they shift after you get a 5 symbol win. A slot can be "hot" for a week, and over night it's another slot which is hot and the one which were hot is suddenly ice cold - it seems the rabbit hold is endless...

So my experience with this experiment is bye bye online gaming aswell, I will not put money more in theese casinoes! You _DO_ win when you need to it seems, however nobody in here believes this. Most likely there will be a stat fanatic in here which will have a field day diving into the data.

Having lost, surely, +50.000 in cold cash in this game I am not amused either! If I were you I'd probably stay away from 3dice aswell, since I have a feeling the "online community" shares their flaws. (Again, non existing flaws).

And for all of those winning shots, there are a HUGE difference in wagering 0.09-0.9 bets and 9-250 bets! Surealy this has to do with the fact that you cant win anything on low wagering...

Again - my personal opinions, hit me!
 
It happends to be I have wasted alot of time with "practise-play" the last month(s), and I have buildt a huge dataset with playcheck data for the entire period. I have passed 239765 spins as we speak, most on maximum wager. So since funmode is identical with real mode, I thought I'd poll the payout percentage for you all. I will post the results when there are some votes for what you think, ;)

I think the payout would be ~88.03% ;) . Seriously, 1 SD in % return over 250,000 spins should be Standard Deviation per spin / 5. With an SD per spin of 15 (may be reasonable for high variance slots) and a fixed bet size, the expected 2 SD range is 89-101%. Zoozie's simulator could probably give more precise information.
 
Seriously, 1 SD in % return over 250,000 spins should be Standard Deviation per spin / 5. With an SD per spin of 15 (may be reasonable for high variance slots) and a fixed bet size, the expected 2 SD range is 89-101%. Zoozie's simulator could probably give more precise information.

This is damn accurate!
I did 1000 simulations of:
250000 spins on thunderstruck (freespins not counting as these 250000)

Of these only 4 had payout a little over 100%. Highest was 1.0074301819113503
The lowest payout was 0.9135081857338577
But the avg payout% was 0.956 which a little higher than was expected, so it was a slightly 'lucky' simulation.

The data with the payout from the 1000 simulations is included. And here are two histograms of the results with low and high interval resolution.
 
Great to see some interest in this, I should also add (in case this was misunderstood), my gameplay is in all the slots. So +250.000 spins all over the place in total. So some slots will have a high payout and some will have a lower. I'll prepare the results and also post the data tomorrow.

Always nice to have something to look forward to, :)
 
This is damn accurate!
I did 1000 simulations of:
250000 spins on thunderstruck (freespins not counting as these 250000)

Of these only 4 had payout a little over 100%. Highest was 1.0074301819113503
The lowest payout was 0.9135081857338577
But the avg payout% was 0.956 which a little higher than was expected, so it was a slightly 'lucky' simulation.

The data with the payout from the 1000 simulations is included. And here are two histograms of the results with low and high interval resolution.
Based on the simulation data, the SD per spin was about half of my guess. This is surprising to me since Wizard of Odds listed a greater degree of variance than this in a Reno 3-reel slot, as described at
You do not have permission to view link Log in or register now.
. It also seems that 250,000 spins is not enough to produce a good gaussian distribution, as the curve is still skewed a bit on the higher side due to some sims hitting rare high payouts.

Based on this data, I expect you'd need ~2 million spins, to have a 95% chance of being +/- 1% of expected payout.
 
Based on the simulation data, the SD per spin was about half of my guess. This is surprising to me since Wizard of Odds listed a greater degree of variance than this in a Reno 3-reel slot, as described at
You do not have permission to view link Log in or register now.
.

I forgot to write I was playing the full 9 lines at Thunderstruck.
If I played online 1 line the variance would be higher.
 
Based on this data, I expect you'd need ~2 million spins, to have a 95% chance of being +/- 1% of expected payout.

Would this work on an overall 2 million spins on different games or would you need 2 million spins on one particular game to be scientific sure?
 
Would this work on an overall 2 million spins on different games or would you need 2 million spins on one particular game to be scientific sure?
If playing a game with a large progressive jackpot, like Major Millions, then 2 million spins is not enough. If you are playing games that do not have a greater variance than Thunderstruck and using a constant bet size, then 2 million spins probably would be enough. The result would be the expected return with that fraction of play on each game, rather than the expected return for any one game.
 
To be a valid sample, it would have to be all on the same slot, at the same betsize. There are so many possible outcomes on the likes of Thunderstruck that even 2 Million spins might not be enough due to the rare and high hitting possibilities, such as 5 wilds in the free spins, or lesser, but still fantastic, the combo of 4 wilds + 5 wild hammers on a single spin within free spins. There is also the 5 scatter retrigger during free spins. These rare hits pay very well indeed, and if not properly represented in the sample would drag down the payout percentage.

Rather than looking through all that playcheck data, you could just set autoplay to run 5000 spin sessions, and log the total won or lost each time. You could build up a database of 250,000 or even 2 million spins this way, and can keep adding to it.
To see if these are patterns of hot or cold in time, you could do a set amount of these sessions each day, and see if there is any pattern (there shouldn't be).
I have noticed times when scatters appear frequently, and then long spells where they hardly appear. There is a strong link between which spell the slot is in, and the frequency of bonus rounds. I also notice long spells where almost nothing is hit, followed by spells where alot of small hits keep me going, even without any bonus rounds.

A simpler case might be to study a low variance slot, without free spins or bonuses, such as 5 reel drive, and then Double magic (to see if 3 reel slots are better or worse than 5 reel slots).
 
Man, so it's really useless since MG, RTG and all other software provieders and their proprietary software can never truly be busted unless someone - attached with a camera - comprises their property and analyzes the code line by line...

Who has the possibility to do +2 million spins on one slot? And more interstingly, do it on lowest bet and then on highest bet? You also mention that 2 mill is probably to little, you would need maby 10 mill?

This is suerly good news for everyone, but sadly it means that chaos and law of big numbers is the casino makers biggest friend... hrmf!

If you remember, this is also why I asked in another thread - how many times would you need to see a pattern, to scientifically know it's infact a pattern - and not variance or whatever (I am not good with formulas and such, logistics and patterns however I work with).

Anyways, just had to respond to what I found sadening comments on the amount of spins! It took weeks for 250.000 spins... I can't imagening doing 2 mill on every game...
 
...
Based on this data, I expect you'd need ~2 million spins, to have a 95% chance of being +/- 1% of expected payout.

Unless the Wiz data set was ~2mm*15 I already asked, and showed that the sample was over three million... or did I? hmm.
 
Man, so it's really useless since MG, RTG and all other software provieders and their proprietary software can never truly be busted unless someone - attached with a camera - comprises their property and analyzes the code line by line...

Who has the possibility to do +2 million spins on one slot? And more interstingly, do it on lowest bet and then on highest bet? You also mention that 2 mill is probably to little, you would need maby 10 mill?

This is suerly good news for everyone, but sadly it means that chaos and law of big numbers is the casino makers biggest friend... hrmf!

If you remember, this is also why I asked in another thread - how many times would you need to see a pattern, to scientifically know it's infact a pattern - and not variance or whatever (I am not good with formulas and such, logistics and patterns however I work with).

Anyways, just had to respond to what I found sadening comments on the amount of spins! It took weeks for 250.000 spins... I can't imagening doing 2 mill on every game...

I wrote 2 million spins for a 95% chance of being within 1% of expected payout. That doesn't mean you need to do 2 million spins to prove something is wrong. It just means with fewer spins the expected range of return will be larger. BAsed on Zoozie's simulation with 250,000 spins on Thunderstruck, there is a 95% chance of being within ~3% of the the expected payout. So there is about a 1 in 100,000 chance of the measured payout being 88% over 250,000 spins.
 
So this would mean:
If the proprietary software would supress large wins, and give them when you need then instead (ladning on an overall ~95% payout), this would in practise be impossible to proove! I assume this, since the lack of the random jackpot win would need such an redicculous large dataset, and given chaos theory the jacpot just aswell mey hit 3 times in a row...

From what I understand here, lojo, zoozie, aka23 and me would be able to develope the most succesful RNG in the world for the owner - and anyone could check it and not be able to find the cheating mechanism.

This is really astonishing in my opinion, or should I say - depressing. However, totally besides the point of this thread so I will back off now. :)
 
Didn't mean to derail Kimss (and thanks for putting me in a class of intellect I don't hold a candle to:))
Don't bow out, just try to state another way if you need to. How long of a run would we need to show what?
Please remember the RNG and the program shouldn't be looked at in the same light.:)
 
Last edited:
Datafile is stored here in 14 days from now (17mb uncompressed):
You do not have permission to view link Log in or register now.


The results are in, and the payout are 92.55% in 250.000 spins.

Attached are the payouts for the different games and wagers. Any stat fanatic will - by looking in segments of the stats from day to day - see some strange "artifacts". Also, for those interested - something strange happends after the "big" win in Glory of Rome which I am on the stage of a thesis on. I will need to reset this experiment and do 3 casinoes at a time to confirm(?) my findings. Anyways, enough of that. Time for me to pause this and do some real work, I do not get paid digging into statistics, :)
 
The results are in, and the payout are 92.55% in 250.000 spins.

You are using very different bet sizes. And in terms of payout only the results from the high bets will basically effect the total payout%

So to say you got 92.5% payout from 250.000 spins(which was a 1/100 event or so from my simulation) is actually not quite correct. So the effective number spins is much less.
 
You are using very different bet sizes. And in terms of payout only the results from the high bets will basically effect the total payout%

So to say you got 92.5% payout from 250.000 spins(which was a 1/100 event or so from my simulation) is actually not quite correct. So the effective number spins is much less.

There is a lesson to be learned from this dataset, and that is however poorly you play, however you loose or win, your overall payout will be 92-95% after passing ~15.000 spins. This I also tested in a much lower dataset with 2 other casinoes. That - should be mindboggling, as it would assume there are some "devine" powers at play here - even if the dataset are to low. From your understanding I might aswell end up with <80% or high >110%, but I dont - you get pinned in the 92-95 land.

As you know from earlier conversations Zoozie, playing one slot is not interesting (since I assume the system is broken) - looking at how the slots and payouts work together - _that_ is interesting.
 
You are using very different bet sizes. And in terms of payout only the results from the high bets will basically effect the total payout%

So to say you got 92.5% payout from 250.000 spins(which was a 1/100 event or so from my simulation) is actually not quite correct. So the effective number spins is much less.

An idea might be to normalize data, so ignore the bet size, but considering the very different bet sizes that would be a non-trivial job.
Other idea might be to simply forget the individual win/loss amounts, just sum up the payout ratios and calculate the average. Another approach would be to calculate the median of the payout%s. I mean sort the payout percentages, and find the middle element in the sorted list.

I'm quite busy nowadays, but it would be nice if you could calculate the above described average and median of payout%s.
 

Users who are viewing this thread

Meister Ratings

Back
Top