Online Slot stats May 2010!

[takethemoney]

Once this was said to me and it took awhile for it to sink in, but I was told that I could be on a machine with an RTP of 110% and still lose. Anyone believe this? It's true.

 

[3Dice]

Enzo and 3Dice casino may run a tight ship but transparency is not in their vocabulary. And with no regulation even 3Dice is only a stones throw from being the next rogue. It must kill them to watch all these other casinos raking in the dough with all the rogue tactics being used online.

If there was any type of regulation what so ever in the slightest form being enforced with online casinos there would be no need for good people that care and feel that need to go through such excruciating efforts , even if in vein for the sake of some sort of gauge to judge which casino is better. The fact that Enzo even entertained this thread gives some credibility to his site, but where are all the rest of the so called veteran advocates or reps? Can anyone defend their RTP’s?

Not to be insulting here Enzo, but why should we believe the RTP’s you printed? Because you say we should? Or because some alleged independent lab (hired by your company) says that’s what they are? All these independent firms are in bed with the casinos. I’m sure your firm is just like Technical Systems Testing lab the company that handles RTG and offers their seal of approval for honest RNG’s. So please send me a copy of yours so in case I run short on toilet paper I have some backup.

Hi 4 of a kind,

Players are not the only party here that wishes that better regulatory bodies would exist for online casinos. Given the global aspect of this business, I personally (and this is my personal opinion) believe that it is going to be difficult for any single nation to setup a valid body for that. Perhaps the answer will ultimately be in the form of an open-source, non-paid and community regulated system, a thing for which the audience may not yet be ready.

As a casino operator, I believe 3Dice has done everything we can to make sure our operation is as transparent as possible. That includes :

- selecting our third party auditing firm with great care. Please review their customer list.
- afaik 3Dice is the only casino publishing live RTP information (updates every 10 minutes on our Old / Expired Link).
- afaik 3Dice is the only casino publishing all big wins live as they happen, both in the chatroom as well as on the zeitgeist.
- We also publish live banking statistics, and just about every other bit of information that may be remotely interesting to the player.

I personally feel that this strategy of publishing live information is very beneficial to the player. You can verify this system easily yourself, and if you hit a big win you will see it show up in the chat, on the zeitgeist, and you may even see the effect it has on today's RTP.

A lot of care has been spent building the zeitgeist page, if you haven't seen it yet then be sure to check it out. If other casino's would follow suit on that, it would be an endless source of information for the players.

Enzo,
Can you please clarify this statement for me?

"Final Outcome Distribution Tests were performed using confidence intervals between 95% and 98%, which are documented intervals of confidence for such statistical analysis."

In addition to what I said above, please don't consider anything I said as a direct attack towards you. You are one of the few with credibility.

A confidence interval is a way to statistically describe how reliable test results are. Say we have a sample of size x, and we measure a parameter in that sample, say RTP. The statement the RTP is y with a confidence interval of 95% .. then we mean that we are sure that for the entire population (not just the sample of size x) the RTP will be no further than 5% from y.

One of the biggest factors involved in calculating the confidence interval is that sample size x. Typically if the sample size x is low, the confidence you should have that the sample is a good representation of the whole population is low. The bigger the sample size the higher the confidence interval. If your sample is the entire population, your confidence becomes 100%.

It's exactly these types of things shyguy needs to urgently look into. I've explained all this to him in pm, and urged him to remove the bad advise from his post and update his tables with the required numbers as e.g. sample sizes and confidence intervals. He blatantly refused to do so. I would've prefered him being the one setting things straight as opposed to me.

I wouldn't be so strict if this were a player posting, but lets keep our eyes on the track here. Is the average player (of which I do not require a mathematical background) really helped with a site called slotistics, ran by someone who doesn't know the first thing about statistics, filled with numbers that are meaningless, giving out bad advise that will harm the player. And do we as an informed community really accept it that someone is trying to make money off players by doing that ?

Would we accept this blatant type of mistake and ill advice from e.g. a casino rep ?? I think not - and if you wont accept it from the support department then why on earth would you accept it from the marketing department ? (i.e. affiliate). It's unethical.

Once this was said to me and it took awhile for it to sink in, but I was told that I could be on a machine with an RTP of 110% and still lose. Anyone believe this? It's true.

10 players make a $10 bet on this 110% machine. 9 lose. 1 wins $110. Unfortunately, takethemoney, you were one of those 9.



Cheers,

Enzo

 

[PNWGirl38]

Hell, I'm still of the mind set that.............. 2+2= 5




I hated math, but the teacher was hot!

 

[4 of a kind]

A confidence interval is a way to statistically describe how reliable test results are. Say we have a sample of size x, and we measure a parameter in that sample, say RTP. The statement the RTP is y with a confidence interval of 95% .. then we mean that we are sure that for the entire population (not just the sample of size x) the RTP will be no further than 5% from y.

Would a sample size of spins which give you a 95% confidence interval be acceptable for determining that the Slot is 'fair'? That only means you're 95% certain, correct? But there's a 5% chance it's not? Why would you only be able to establish a 95% confidence interval in a sample size of Slot spin outcomes? Is it merely that the variance is so insane, that you'd simply need so many tens of millions of spin results to establish a higher confidence interval?

For example land based casino slots data is broken down monthly with nothing more then retrieving 100% data stored.

 

[3Dice]

Would a sample size of spins which give you a 95% confidence interval be acceptable for determining that the Slot is 'fair'? That only means you're 95% certain, correct? But there's a 5% chance it's not? Why would you only be able to establish a 95% confidence interval in a sample size of Slot spin outcomes? Is it merely that the variance is so insane, that you'd simply need so many tens of millions of spin results to establish a higher confidence interval?

For example land based casino slots data is broken down monthly with nothing more then retrieving 100% data stored.

Hi 4 Of a Kind,

Dont get confidence interval confused with confidence level. Both can be expressed as percentages which can be confusing. It may be more clear if you express the confidence interval as an absolute interval.

So for example a confidence interval of 95% on a measured value of 100 translates to an interval [95,105]. Now for this interval we could also calculate the confidence level. The confidence level describes the odds that another sample will lie within the confidence interval.

Typically the wider the confidence interval, the higher the confidence level.

Say we consider a sample of people of which we measure the length. A sample like this would typically follow your standard bell curve distribution. The shortest person is 125cm, the longest 225cm and the average is 175 cm. In this setup we could have a high confidence level that even for a sample of size 1 the average length fits within 175cm +- 50cm .. If we want to keep that high confidence level but be more specific in the interval then we'll have to measure more than 1 person .. so as the sample size grows we can keep our confidence level high, but reduce our confidence interval.

Now you see these three variables, confidence level, confidence interval and sample size are closely related. If we want a high confidence level on a small sample size then we will have to accept a big confidence interval. As the sample size grows we can reduce the confidence interval without reducing the confidence level.

I don't want to turn this thread any more into a math course lol, wiki has some good articles on this.

So to correct your original statement, a 95% confidence interval doesn't mean your 95% sure. It means your [fill in confidence level] sure that the entire populations value is no further than 5% of the measured sample one.

If you measure _all_ the data, and you give your numbers the correct name, then the confidence interval is reduced to zero and the confidence level 100%. Thats why landbased you can perfectly say hotel x in month y has had an average RTP of Z. If you measure all the data you don't have to specify the sample size, confidence interval and level.

Cheers,

Enzo

 

[Lucky Loser]

Enzo,

Love your casino. Are there any new slots in the works ?

 

[Bencuri]

Online Slotistics May 2010

Amazing how many of you were kind enough to send me your slot stats and logs. They have made very interesting reading and now it is all collated I am able to provide you with some very interesting stats, some you will like, others you will not.

The RTP's are a bit lower than expected. The whole stats for May were however taken from other logs and sessions prior to the beginning of May but I decided to use them as it has given me something to work with.

Something I did learn from the majority of logs is the popularity of certain slot machines, but more importantly when you all tend to start losing.

Many of you tend to reach about 40-60% above your initial deposit but carry on. From what I have seen, those that do this then lose approxiamately 50% of that total.

Simply put, player deposits $100, wins up to $140, continues to gamble, walks with $70!!

With that 70% the majority of you then played at the same casino but at a different slot, same thing happens.

From what I have observed only this month, it would seem logical for anyone who manages to win 40-60% above their initial deposit, to move onto a totally different casino with a different software platform. I will assume many of you play at different casinos so this is an easy thing to do, the problem is you cannot transfer your money from one account to the other.

Hopefully you will have funds in at least two different accounts. Now CM offers quite a lot of casinos with different software platforms and I would strongly advise that you mix at least two during your play time. IE Wager $100 at RTG and then $100 at MG.

Sessions that ended and re-started the following day looked to decrease at a slower rate of 12-17%. Therefore, if on the Monday you leave with your $140 (40% profit) you could expect to lose 12-17% of that on the Tuesday. Once that initial 12-17% is completed, the rise was slow but measured up against the standard RTP's I have recorded.

I would recommend all players to look to change software platforms when/if reaching 40-60%. As a fail safe you may even wish to use 25%+ as a guide.

Below are the figures I have calculated for May. A total of 832 sessions were collected. I only received 9 sessions from Boss Media platforms. Remember they are sessions I have received from the players and not what the actual software companies may or may not provide.

I decided to change the "Free Spins per dollar" to "Free Spins per 100 spins". The same goes for "Bonus Games". Surprisingly perhaps RTG offered more free spins per 100 but even that just broke the 1 Free spin per 100. This is alarming.

Many of those stats that I had collected showed many players to receive a lot more free spins after spinning at least 250+ during the one session, a higher stake per spin also showed to be more productive at getting more free spins and bonus games than playing at the lowest stake. I will be looking to add more to this by the end of June.

More stats can be found on the website, please do keep sending them to me on slotistics@slotistics.com The more I have the more I can find out.

If you like what I am doing there is a little "Donate" button on the site, all donations kindly accepted! Best of luck and will hopefully have more for you at the end of the June!

I am hoping to provide more on specific slot machines, actual casinos, IE Rushmore versus Cherry Red, and most popular slots on each platform and the average amount of spins taken before winning free spins, bonus games and a win over $50 in one spin.

Please remember I cannot do it without your help so please take a minute or so to jot some of your notes down!

Thanks once again to all who sent me their's and Keep those stats coming!

Software RTP:

Playtech: 88.59
Micro: 97.21

Free spins in 100 spins:

RTG: 1.158
Playtech: 0.059

I hardly believe this. It can be true in case of a special selection of games, but if you consider every game, and make an overall statistic, I am sure Playtech will overcome Micro and RTG. And if you involve particular Rival casinos, that may also.

Or maybe your casino selection caused the problem. Accoarding to my experience the CM accredited Playtechs are more tight than non accredited ones. Try to do this statistic at Prestige Casino, Noble Casino and Sega Casino. I am sure you will get different data. I am not a slot fan, but most of the time when I play in these casinos, I can break even at least most of the time. Compared to that Micro and RTG is a money vacum. They have games on that I haven't seen the free spins yet, how ever I am playing them for years now.

 

[Rusty]

Rusty,

Here's a sample split in groups of 100k spins of tut. The paytable of tut is exactly the same as that of mg's isis game. Yes its a high variant machine, maybe shyguy can tell us how much isis he has in his logs .. It's one of the popular mg games so I wouldn't be surprised it is one of the games in there.

At 3Dice it is our most popular game (see Old / Expired Link). I feel these two reasons make it a good choice for examination. Feel free to suggest another 3Dice game to me.

The data below is an overview of 15 groups of 100k spins of data from may and a bit in april. I've listed the average RTP for each group, and for the 3 groups of 500k spins you can make from the same data.

100k spins : 96.47
           : 96.59   
           : 91.62
           : 93.11
	   : 86.16    92.79
		
   
           : 96.16
           : 94.60
           : 93.46
           : 94.77
           : 100.94   95.99

           : 95.70
           : 94.35
           : 101.69
           : 94.14    
           : 96.02    96.38  

100k spins  : 86.16 to 101.69 = 15.53 % = +- ~8%
500k spins  : 92.79 to 96.38  = 3.59  % = +- ~2%
1.5M spins  : 95.05
 10M spins  : 94.89

As you can see, groups of 100k spins in just this sample of 15 range from 86.16 to 101.69 .. almost 16% .. if shygy displays any RTP's on groups of data as small as that size, then my estimate of 5% of in either direction is still a conservative number. In other words with samples of this size he could print 95 but it could just as well be 90 or 100 .. big difference if you ask me ..

Even if you group 500k spins, the range is still 3.6% .. still a big error .. at the very least shygy needs to display how much data was used for the calculations so that the numbers can be read.



Lol Rusty, I didn't mean to come on so strong. Your sample only proves my point tho .. I mean you say it yourself .. you add a single spin to it and it jumps 2.5% .. not exactly very stable now is it ?

A big win that occurs one in a million has a 1 in 1000 chance of re-occuring within 1000 spins after it was last hit. So one in 100 sessions of 10000 will have two of those jackpots .. that's 5% right there ..

If _any_ single spin still makes that much of a difference then you have no stable sample at all. You've said it yourself .. a jackpot win like that should only be a fraction of the RTP.

Your example only proves my point. 10.000 spins is simply not enough to do any type of RTP statistics .. and you would need no more information than the fact that there are events with frequencies that range into the millions. Many of the events that make up a big part of the pay of a slot occur less than 1 in 10.000 spins naturally .. so a sample of that size simply cannot be used to measure that. Most samples wont have those events and they will have lower than expected RTP's and the sample's that do will have much higher than expected RTP's.



agreed. The idea of measuring the number of freespins/bonuses received per 100 spins is absolutely valid. And to determine that number, using the general rule of thumb (30 times frequency), and say a worst case scenario of a slot that only hits theoretically once every 250 spins, you can analyze with reasonable certainty on a sample of 250x30=7500 spins. But you should not publish an RTP for that - it is meaningless. You can publish the average number of bonuses received and if their expected frequency is 1 in 250 or higher then that number will be a valid number.

However, the numbers published by shygy are obviously not correct, not the RTP's and none of the other numbers. For example, and as I stated in my first post he claims 1 feature (bonus or freespin) once every 1500 or so spins on MG .. to measure that type of frequency he would need a sample of 1500*30 = 45000 .. so either he doesn't even have that many spins .. or he has made mistakes in his analysis/calculations.



lol. define 'negatively weighted' .. many 3 reel slots are pretty straightforward .. i.e. 100 slots/reel .. jackpot occurs 1 in (100*100*100) .. or one in a million.

Kindest Regards,

Enzo

Hi,
Sorry forgot all about this thread and was just reminded by shyguys new post.

It seems we are only disagreeing on a point of scale here.

You have chosen stats for one of the highest variance games there is as I knew and predicted you would.
If you had chosen a game with much lower variance you can be sure that my argument stands up even better though I will explain why these figures still support my argument in a moment.
Again I can not talk for 3Dcie slots as I just have not played them much.

Even with figures you have chosen on a very high variance slot you can see that 12 out of 15 of the 100 spin sets of data are in fact covered by about only 5% spread. This should of made you blink at least.


I have already stated that when we know certain data is anomalous we can discard it and I would say your figures are in fact a strong vindication of my views. I think we could even make a reasonable statement on the probable RTP of the slot even though that is not really my argument as I have stated.

Also a jump of 2.5% is not much when you are talking about 1 in 50 million+ event occurring within 100,000 spins(lines). 5 wilds for example and if you spent the rest of your life looking for an occurrence of 5 wilds twice on the same slot in 100,000 spins I can promise you that you will not find it. (Unless it is TopGame where 5 wilds pay zero!:clap
That is the reality of how unlikely such events are and even such a hugely anomalous data set on a small sample only gives a 5% swing!

We pretty much agree on the rest but I still say you are wrong to consider sample sizes of less 500,000 spins a joke - even your carefully chosen data suggests that - we just need to apply a some foreknowledge, common sense and critical thinking to the figures.

The reason I am coming on so strong on this is because similar arguments are often used in an attempt to undermine or discredit my findings (Not you, I know you are genuinely arguing your corner and what you believe to be correct) so it is important I make it known that my views stand up.

Regarding negatively weighted 3 reel slots.
I am referring to slots that award a bonus game when a bonus symbol stops on 3rd reel. Obviously the stop position is negatively weighted or the bonus symbol would occur every 20 spins or so. (These slots do not have 100 symbol+ reels but quite small reel strips and apply weighting to the reel stop positions instead to effect probability of outcome. A natural consequence of this is that the bonus symbol will appear much more often 1 above or 1 below the payline.
I am sure you must be aware of this but perhaps my terminology was poor.

We are taking up way too much of each others time on this though and I know you will still see certain things differently. I am glad though that you agree relatively small sample sizes can be useful which is my main point.

PS
I agree with your comments about Shyguys data and if he takes your advise on board (disclaimers etc) I think it is a worthwhile thing he is attempting to do.

All the best.