Question Your help on these observations....

[kktmd]

Sometimes I feel that MG's slots are almost too close to expected RTP Or that we always seem to hit a few more paying combinations than expected!

 

[SubzeroWins]

That data was very helpfull, thanks for sharing! Especially the true odds of hitting a monster 1000x payout

 

[SubzeroWins]

And btw Jufo, i'd be interested in reading that

 

[Jufo]

Sometimes I feel that MG's slots are almost too close to expected RTP Or that we always seem to hit a few more paying combinations than expected!

Yeah if you look at the graphs within the report I linked, the player is significantly above TRTP at first and then falls almost too conveniently close to it at the end. But he did hit each of payout magnitudes expected number of times so it is hard to complain with a result like that. It now becomes a question of how can one dig deeper beyond this and measuring the distances between large payouts was the first thing I could think of to separate a random slot from a fruit machine one. The problem is that even with those 200k spins of real money BDBA data there might not be enough big hits to get reliable analysis of their distances.

 

[Janek12]

Raising stake IS real play because that's what people do, as we've heard. If you want to do it, please do, because I certainly would if I had your program to play MG. I think it would be very interesting for you, me and everyone.... We could put any doubts to bed forever..

Several years ago, there was a poster here with the nick KIMSS (I hope I recall it correctly) and he posted extensive statistical data that proved, in his opinion, that he was more likely to lose on MG slots when placing a higher bet.
IIRC, a higher bet in his case was something like 6 EUR or perhaps even more. He admitted he sunk around 500,000 (yes, half a million) USD into slots over a period of several years.
He claimed he was a computer/software expert, based in Norway, with his own company I think, so it seemed he was able to collect and process the data from slots in quite an impressive way.
But I am no expert on statistics, so I was unable to draw my own conclusion from the charts and data he posted.
Overall, he seemed really a nice and intelligent guy, struggling with a gambling habit, which he seemed to have finally gotten rid of and left this forum for good.
The only conclusion I drew from that for myself was that I am in a "safe zone" anyway because I almost never bet higher than something like 2 EUR on slots.

 

[Jufo]

Several years ago, there was a poster here with the nick KIMSS (I hope I recall it correctly) and he posted extensive statistical data that proved, in his opinion, that he was more likely to lose on MG slots when placing a higher bet.
IIRC, a higher bet in his case was something like 6 EUR or perhaps even more. He admitted he sunk around 500,000 (yes, half a million) USD into slots over a period of several years.
He claimed he was a computer/software expert, based in Norway, with his own company I think, so it seemed he was able to collect and process the data from slots in quite an impressive way.
But I am no expert on statistics, so I was unable to draw my own conclusion from the charts and data he posted.
Overall, he seemed really a nice and intelligent guy, struggling with a gambling habit, which he seemed to have finally gotten rid of and left this forum for good.
The only conclusion I drew from that for myself was that I am in a "safe zone" anyway because I almost never bet higher than something like 2 EUR on slots.

Yes, I remember him very well. However most of his posts were ranting without any real evidence so I missed the part where he presented that extensive statistical data. I would think that if the data really proved something someone here would have taken a look at it and verified his evidence.

 

[kktmd]

Thats one of the aspects I'm interested in: how changing bet size affects RTP.

In my experience, a bonusround will occur 80-140 spins from the previous round. So, I did about 1.4 mio spins on Playboy, flatbetting at 0.3.

If I instead had raised my bet to 2.10 from spin 80 to spin 140 (and lowered after 140 - dont chase a bonus round!), my result would have looked like shown in the table. The last column is calculated using the exact same spin results, but with variable bets.

If I'm kicked off MG's server, my serie is interrupted - thats why some series end before reaching 0.

The text at the bottom of the table shows the Actual RTP and the calculated RTP (both under 1, sadly

NB: All numbers are show x100 too high (the decimal point is not placed correctly).

 

[Jufo]

Thats one of the aspects I'm interested in: how changing bet size affects RTP.

In my experience, a bonusround will occur 80-140 spins from the previous round. So, I did about 1.4 mio spins on Playboy, flatbetting at 0.3.

That can't be right. If the bonus round has the same 1 in 150 chance of triggering on every single spin it can come on the very next spin or on the 1000th spin. Or does Playboy slot have some kind of mechanism other than scatters to trigger the bonus round?

Since a span of 80th spin to 140th spin is 60 spins, there should be around 1/3 chance that the bonus round is triggered within that interval. You also need to miss it on the first 80 spins (probability of 58.6%) so overall the freespins will be triggered between spins 80-140 only 20% of time. If this isn't the case in your experiments that's an overwhelming evidence of non-randomness.

Do you think you could record (keep track of) the distances from one bonus round to next and show the results to us?

If I instead had raised my bet to 2.10 from spin 80 to spin 140 (and lowered after 140 - dont chase a bonus round!), my result would have looked like shown in the table. The last column is calculated using the exact same spin results, but with variable bets.

That sounds like a gambler's fallacy type of thinking: "If I had made that choice, I would have won" but the thing is it's very easy to say this after you already know the outcome.

 

[kktmd]

That sounds like a gambler's fallacy type of thinking: "If I had made that choice, I would have won" but the thing is it's very easy to say this after you already know the outcome.

haha....true....good thing I only gamble with pretend money!

 

[kktmd]

Ok, same as before, but this time we cash out if our balance hits 3*start balance or goes below 1/3*start balance.

Just playing around with the numbers here - I realize that fiddling with the numbers in retrospect is cheating ;-)

 

[kktmd]

OK - final scenario. This is the one I most often use when I play MG 243 slots with real money. Cash out at 3x startbalance or when balance goes below 1/3*start balance.

Spin 0-80: bet 0.3
Spin 81-130: bet 1.2
Spin 131-180: bet 2.1
Spin 181->: bet 0.3 (I try never to chase a bonus round)

 

[kktmd]

That can't be right. If the bonus round has the same 1 in 150 chance of triggering on every single spin it can come on the very next spin or on the 1000th spin. Or does Playboy slot have some kind of mechanism other than scatters to trigger the bonus round?

In my experience/opinion, if you dont hit a bonus round within the first 200 spins, you can wait a VERY long time before you finally hit it. These cases of über-long periods of horror, add a lot to the average IMO. I suspect the Median is somewhere in the 80-140 area (but I could be wrong - I know the average should ideally=the median in a uniform distribution).

 

[Jufo]

Hmm to confirm you should create a new 1M spins sample and see if the RTP by playing with the same strategy is still above 98% in the new data. Because now the RTP of 98% can be due to random variation (you don't hit the TRTP precisely in 1M spins, it can vary several % in either direction).

In my experience, if you dont hit a bonus round within the first 200 spins, you can wait a VERY long time before you finally hit it. These cases of über-long periods of horror, add a lot to the average IMO. I suspect the Median is somewhere in the 80-140 area (but I could be wrong).

The waiting time to next free spins round should still be completely random though. If you haven't hit it in 200 spins then, because the game doesn't have memory, you are still expected to play 150 spins until you see one. It shouldn't be due any sooner even if you have already been long time without. This is balanced by the fact that sometimes you should hit them within 10 spins of each other several times in row.

The number of spins between free spins should follow a well-defined probability curve. That's why it would be interesting to get data of this to verify that it is indeed random and not controlled somehow.

 

[kktmd]

I absolutely agree

Ideally, I would like to do 1 mio spins using real money, but......eeehhh.....not happening

"game doesn't have memory".....thats why I state in my profile that I think games are rigged ;-).....seriously, I've waited 2.000 spins on Hellboy...the odds of that are astronomical! I kept hitting their alternative bonusround every 80 spins - like clockwork. After a while, I started upping my bet dramatically after 70 spins, and ended up cashing 10*start balance out - but 2.000 rounds without 3 scatters??!

 

[Jufo]

I absolutely agree

Ideally, I would like to do 1 mio spins using real money, but......eeehhh.....not happening

"game doesn't have memory".....thats why I state in my profile that I think games are rigged ;-).....seriously, I've waited 2.000 spins on Hellboy...the odds of that are astronomical! I kept hitting their alternative bonusround every 80 spins - like clockwork. After a while, I started upping my bet dramatically after 70 spins, and ended up cashing 10*start balance out - but 2.000 rounds without 3 scatters??!

Well it's rare but can happen. I would say that if it was rigged then it would give the scatters at a predictable point after having lost specified amount rather than being random. Maybe you prefer non-random

 

[dunover]

It's interesting that clearly changing bet ups the RTP by nearly 3% (but only in your method) and even with the amount you have spun, it is something we shouldn't expect. We know that the bonus round is responsible for a large proportion of the RTP, and if in some way your raising IS getting that bonus round closer, then that could account for it. The thing is, if 100% unmanaged random, the other play would reduce the RTP you had 'kicked' back down to nearer the 95%. It hasn't.

 

[Jufo]

It's interesting that clearly changing bet ups the RTP by nearly 3% (but only in your method) and even with the amount you have spun, it is something we shouldn't expect. We know that the bonus round is responsible for a large proportion of the RTP, and if in some way your raising IS getting that bonus round closer, then that could account for it. The thing is, if 100% unmanaged random, the other play would reduce the RTP you had 'kicked' back down to nearer the 95%. It hasn't.

It's an interesting thought that you could somehow take advantage of software's non-randomness to get higher RTP (in this case extra 3%). However in 1M spins you can get 98% RTP by purely random chance so this needs further testing AND even if it works in fun-mode it's a completely different story whether it works in real money play as well. I remain sceptical.

However, measuring the distances between big payouts or bonus rounds should to be done to verify randomness.

 

[kktmd]

Exactly the same simulation on my Immortal Romance DB. I have removed series where I have less than 10.000 spins (like previously).

Spin 0-80: bet 0.3
Spin 81-130: bet 1.2
Spin 131-180: bet 2.1
Spin 181->: bet 0.3 (I try never to chase a bonus round)

Cashout at 3*startbalance or 1/3*startbalance.

Its probably not very rational, but its how I like to play MG243 anyway

 

[Jufo]

Hmm only 1% diff though. Too little to draw conclusions.

 

[kktmd]

Aaaaaand...The Finer Reels of Life....

 

[dunover]

Hmm to confirm you should create a new 1M spins sample and see if the RTP by playing with the same strategy is still above 98% in the new data. Because now the RTP of 98% can be due to random variation (you don't hit the TRTP precisely in 1M spins, it can vary several % in either direction).



The waiting time to next free spins round should still be completely random though. If you haven't hit it in 200 spins then, because the game doesn't have memory, you are still expected to play 150 spins until you see one. It shouldn't be due any sooner even if you have already been long time without. This is balanced by the fact that sometimes you should hit them within 10 spins of each other several times in row.

The number of spins between free spins should follow a well-defined probability curve. That's why it would be interesting to get data of this to verify that it is indeed random and not controlled somehow.

Yes and no. If the game is totally random, which it is? then yes in theory. If kktmd has demonstrated an average amount of spins between each bonus round, then no, because the likelihood of NOT getting it at a certain point becomes a rarer and rarer event with each spin. I.E. over a million spins the game only fails to produce a bonus round say 3 times in any period of 1000+ spins so as your count rises the chances of hitting it increase. There is a mathematical effect (I can't remember the guy's name) that does confirm even though one event is not related to the previous in random things, balancing does occur.

 

[Balthazar]

Errrrmmmm hurrrummphhhh, cough *Finsoft* is a name we won't mention here, and there have been rumblings about some Playtechs in fun mode if you look around on here, but no the slots are the same. The only ones I know COULD be done this way are RTG as they can be operator set. Trust me, your fun play and real engine is the same on major reputable softwares.

They admitted it was an error though and it has been fixed. Unless MGS is straight out vile and dishonest with a clear intent to fool its players, free play and real play are the same. Also, kktmd's results in free play are spot on with the TRTP.

I don't understand that "change your stake and it will change the result of the spin" thing. Why would you guys believe such thing? That's part of the "the slot remembers how much you have put in it" dementia.

Look at the Playboy results in the screenshot threads...people hitting big ass wins left and right. Someone even got the full screen of wilds twice. The slot is new..surely they didn't put more in it than they have won did they? When I hit my $900, I had lost maybe $100 in it, tops. Since then I'm +$60. The slot is doing a hell of a bad job at remembering me and trying to get even with my RTP.

 

[Jufo]

Yes and no. If the game is totally random, which it is? then yes in theory. If kktmd has demonstrated an average amount of spins between each bonus round, then no, because the likelihood of NOT getting it at a certain point becomes a rarer and rarer event with each spin. I.E. over a million spins the game only fails to produce a bonus round say 3 times in any period of 1000+ spins so as your count rises the chances of hitting it increase. There is a mathematical effect (I can't remember the guy's name) that does confirm even though one event is not related to the previous in random things, balancing does occur.

Yes, obviously the more spins you do the more likely you are getting the feature because you simply have more shots at it. If the odds for freespins are 1 in 150, then you have 49% chance to get it by 100th spin, 73.8% by 200th spin, 86.6% by 300th spin and so on.

But on each individual spin the odds to get it should be the same. Random game doesn't have memory so now matter how many spins you have done already your odds of getting the feature soon have not increased at all. This is no different than if you see 10 Blacks having been rolled on Roulette, the odds for Red on next spin are still the same as before.

 

[Balthazar]

Random game doesn't have memory

Dunover never understood this concept.

 

[dunover]

Dunover never understood this concept.

You are thinking one-dimensionally. Plus I've agreed with you that MG slots are same in free/paid play. Add in the dimension of TIME. Using your example, I state in advance that I want you to get 10 consecutive blacks. 1024-1. I then ask you to get 11 blacks on the bounce. 2048-1. Both unlikely, but if I had a gun against your head and told you to choose ONE option to live, which one would you choose as the best bet? Exactly.

I fully understand an individual go is unrelated numerically, but in time it is not. As we expect the TRTP to play out over time, then ignore that factor at your peril.

Remember what we are doing here - establishing what the pool of values contains and the frequency of them over an extended period of play/time. kktmd has spent many hours getting the stats, and has established the average frequency of many results, and has also been able to demonstrate the way collective spin results (i.e. going 1000+ without a bonus round) become extremely rare - something worth thinking about next time you take that next session of '3000 unrelated random spins'.........

Roulette is a poor example as it is physical wheel with no outside controlling programme/software. Unlike a slot it cannot be made to pay a TRTP.

And the disingenuous remark ' the slot is having a hell of a time remembering my RTP' - you said yourself you haven't played that many spins on it, bit like me throwing a die once, getting a 6 and proudly shouting 'I win all the time!' This thread is concerning play over millions of spins. As you well know. And something you do know, whether you admit it or not - carry on playing it for a million spins and you certainly ain't gonna be $60 up at the end.....


"Unless MGS is straight out vile and dishonest with a clear intent to fool its players, free play and real play are the same. Also, kktmd's results in free play are spot on with the TRTP.

I don't understand that "change your stake and it will change the result of the spin" thing. Why would you guys would believe such thing? That's part of the "the slot remembers how much you have put in it" dementia."

I've maintained that throughout the thread - it's kktmd who doubts this may be the case - go back and check..........(fun play real money play)

We aren't saying whether we believe it yet Balt - again check. We are testing the idea, and so far kktmd has produced interesting figures changing stakes, don't count your chickens yet...

 

[Balthazar]

Roulette is a poor example as it is physical wheel with no outside controlling programme/software. Unlike a slot it cannot be made to pay a TRTP.

Hmm? Roulette was made to meet its TRTP (aka house edge). Both Roulette and slots are exactly the same in the sense that they will meet their expected RTP over time without the need to cheat, remember previous results or taking stakes in consideration.

What you guys are implying is that MGS slots are rigged and influenced by outside factors (such as previous results and/or stake) when you press the spin button. Sure MGS could be run by cheating crooks but I'd rather not believe that (otherwise I'd just stop playing).

 

[mathsboy1975]

Yes, obviously the more spins you do the more likely you are getting the feature because you simply have more shots at it. If the odds for freespins are 1 in 150, then you have 49% chance to get it by 100th spin, 73.8% by 200th spin, 86.6% by 300th spin and so on.

But on each individual spin the odds to get it should be the same. Random game doesn't have memory so now matter how many spins you have done already your odds of getting the feature soon have not increased at all. This is no different than if you see 10 Blacks having been rolled on Roulette, the odds for Red on next spin are still the same as before.

Yes that is true under the assumptions of constant probability of 1/150. We can observe that empirically over a huge number of spins, but if, as was theorised the game is able to alter these probabilities periodically in controlled amounts in an oscillatory manner such that the theoretical RTP is still satisfied we cannot assume that this is the case. This thread has quite a lot of activity and data posted and I am going to take a look tomorrow in more detail when I sober up a bit. I still cannot think of a way to statistically verify whether the probability does change - not possible anyway if it is constant. I have never attempted to contact slot manufacturers or casinos directly but I am guessing that they would be reluctant to divulge any information on the matter. I personally don't think it is the case in reality and if it were we would never learn about it from the industry as it would give credibility to the "I win a lot when I open my account but soon after the wins dry up" theorists. I still think it is an interesting exercise though as even if it were true, provided the enhanced payout periods matched the reduced payout periods it would in no way make any difference to the games fairness which at the end of the day is the most important aspect.

 

[dunover]

Hmm? Roulette was made to meet its TRTP (aka house edge). Both Roulette and slots are exactly the same in the sense that they will meet their expected RTP over time without the need to cheat, remember previous results or taking stakes in consideration.


What you guys are implying is that MGS slots are rigged and influenced by outside factors (such as previous results and/or stake) when you press the spin button. Sure MGS could be run by cheating crooks but I'd rather not believe that (otherwise I'd just stop playing).

Nope, please don't put words in our mouths. Who mentioned 'outside' factors? The mechanisms come from within the slot programmes. Who mentioned 'cheating' and 'crooks' - you're being inflammatory now. We've already said they achieve their advertised TRTP's, many times.

So you're assuming the slots are totally unfettered, and nothing exists in the algorithms to prevent a player staking $5 and getting 100x stake, raising immediately to $20 a spin and hitting 1000x stake, then raising to $100 a spin and a few spins later hitting 4000x stake, staying at 100 stake and then 20 spins later getting a 2000x hit and pretty much (from a low-stakes start) busting a casino out? i.e. there is not management of risk or better put the amount a player can go over the RTP line by in relation to amount gambled. I don't mean stopping a player getting 20x on his first 1 unit spin and thus having a short term spike in RTP of 2000%, but I mean that sort of result over an extended period which would involve a potentially huge hit for the casino?

 

[Balthazar]

So you're assuming the slots are totally unfettered, and nothing exists in the algorithms to prevent a player staking $5 and getting 100x stake, raising immediately to $20 a spin and hitting 1000x stake, then raising to $100 a spin and a few spins later hitting 4000x stake, staying at 100 stake and then 20 spins later getting a 2000x hit and pretty much (from a low-stakes start) busting a casino out?

I'm not assuming this, this is how it's supposed to be. If you think that it's not like that then you think it's rigged.

When you play your $100 spin, what difference does it make for the slot if that $100 comes from previous wins or from a fresh deposit? None. You've got $100 in your account and you use it to buy a spin. The casino would rather see you play that $100 on a single spin with a 95% TRTP than to withdraw it. They don't have to "protect themselves" since they already have the odds on their side.

 

[Jufo]

Yes that is true under the assumptions of constant probability of 1/150. We can observe that empirically over a huge number of spins, but if, as was theorised the game is able to alter these probabilities periodically in controlled amounts in an oscillatory manner such that the theoretical RTP is still satisfied we cannot assume that this is the case.

No, the probability is supposed to be constant all the time, it only varies from one slot to next based on their reel strips. A standard slot has five reelstrips and draws random numbers to determine the stopping position for each strip. See the reelstrip layout for example here:

You do not have permission to view link Log in or register now.

The only mechanism at work is to randomly draw the stopping positions on the reelstrips. The TRTP will be satisfied long-term statistically by the law of large numbers and without no other control. This means that pure randomness dictates whether you have long droughts of losing sessions or tightly packed winning sessions.

Now the above is what we have to assume. It's the only configuration that makes the slot fair. Then we can collect real data and do statistical tests based on the hypothesis that the above is true. If data is conflicting then we will have evidence that the slot is non-random and perhaps we can even figure out how it is non-random. However as an initial assumption or null hypothesis we have to assume the above fair configuration.

This thread has quite a lot of activity and data posted and I am going to take a look tomorrow in more detail when I sober up a bit. I still cannot think of a way to statistically verify whether the probability does change - not possible anyway if it is constant. I have never attempted to contact slot manufacturers or casinos directly but I am guessing that they would be reluctant to divulge any information on the matter. I personally don't think it is the case in reality and if it were we would never learn about it from the industry as it would give credibility to the "I win a lot when I open my account but soon after the wins dry up" theorists.

When you have time, please read the report I linked from a previous analysis of one player's real-money play. There is also 300,000 spins of real-money play data if you come up with some statistical test idea

I still think it is an interesting exercise though as even if it were true, provided the enhanced payout periods matched the reduced payout periods it would in no way make any difference to the games fairness which at the end of the day is the most important aspect.

No, it would still make a huge difference to game fairness if the RTP was kept constant in some controlled, non-random fashion. This is because the software would control how much you can win and your wins would be spaced too far from each other to lower the probability that you are ever able to get ahead, so it would make hell of a lot of difference to the player, and it would not be a fair game at all.

There are games that are openly like that and they are called fruit-machines or AWPs. However those games clearly state how they operate. The player has the right to know whether he is playing a genuine random game or an AWP.

 

[Jufo]

So you're assuming the slots are totally unfettered, and nothing exists in the algorithms to prevent a player staking $5 and getting 100x stake, raising immediately to $20 a spin and hitting 1000x stake, then raising to $100 a spin and a few spins later hitting 4000x stake, staying at 100 stake and then 20 spins later getting a 2000x hit and pretty much (from a low-stakes start) busting a casino out? i.e. there is not management of risk or better put the amount a player can go over the RTP line by in relation to amount gambled. I don't mean stopping a player getting 20x on his first 1 unit spin and thus having a short term spike in RTP of 2000%, but I mean that sort of result over an extended period which would involve a potentially huge hit for the casino?

When you play your $100 spin, what difference does it make for the slot if that $100 comes from previous wins or from a fresh deposit? None. You've got $100 in your account and you use it to buy a spin. The casino would rather see you play that $100 on a single spin with a 95% TRTP than to withdraw it. They don't have to "protect themselves" since they already have the odds on their side.

Actually, they still might feel like they need to protect themselves. Even at 5% statistical disadvantage a lucky player can make a small casino bankrupt. He only needs enough variance to overcome the house edge temporarily. It's worth remembering that the combined bankroll of all players is larger than that of the casino and given that casinos hand out bonuses which can shift the RTP in player's favour, it's actually very possible that casinos have some risk-mitigation measures built in within the software to not go bankrupt due to a player's lucky streak.

I have seen a Casino Rewards rep spill out once: "We manage the outcomes to a certain degree" which looks like it is not completely random.

Often it feels like there is some sort of invisible ceiling that caps how much you can win from a single deposit. For example: if you play a game with 100% RTP (like Classic Blackjack at MG) then you should be able to turn a $100 deposit into $10000 balance at odds of 1 in 100. However, I once tried this and despite 100+ attempts I struggled to ever reach a balance higher than 10x of my deposit, which made me feel like there was a hidden risk-mitigation control in place to prevent me from reaching too high balance. It seemed that I always started losing at the same point. Obviously if I deposit $10000 in one hundred separate $100 chunks and the software prevents me from ever reaching $10000 balance then I cannot be playing a 100% RTP game.

 

[JuicyFruit]

Yes, thanks, but I don't think 60@30p then 10@£15 will be right - simply because the 60 spins at 0.30 only add up to £18 before you hit the £15 spins. The results would be meaningless I think?
That's why I suggested a long multi-thousand run at .30 then bumping up to £15 at say 20,000 spins. This would show (any) contrast better.

Wow! Slapped down with a vengeance. That will teach me not to read through each and every single post, before just responding to the OP's original question.

 

[dunover]

Actually, they still might feel like they need to protect themselves. Even at 5% statistical disadvantage a lucky player can make a small casino bankrupt. He only needs enough variance to overcome the house edge temporarily. It's worth remembering that the combined bankroll of all players is larger than that of the casino and given that the casino also hands out bonuses which can shift the RTP in player's favour it's actually very likely that the casino has some risk-mitigation measures built in within the software to not go bankrupt due to a player's lucky streak.

I have seen a Casino Rewards rep spill out once that "we manage the outcomes to a certain degree" which looks like it is not completely random.

Sometimes it feels like that there is some sort of invisible ceiling that caps how much you can win from a single deposit. For example: if you play a game with 100% RTP (like Classic Blackjack at MG) then you should be able to turn a $100 deposit into $10000 balance on every 100th attempt. However I once tried this and despite 100 attempts I struggled to ever reach a balance higher than 10x of my deposit, which made me think that there was of hidden risk-mitigation control in place to prevent me reaching too high balance. Obviously if I deposit $10000 in one hundred $100 chunks and the software prevents me from ever reaching $10000 balance then I cannot be playing a 100% RTP game, can I?

Exactly - the casino's exposure cannot be unlimited. I know they set max stakes etc. but even so imagine some lucky nutter betting £45 a spin on BDBA hitting the 5 wilds/gems in freespins? 187,500 pounds and then goes on to hit monsters on other slots? If we remember VWM hitting over 50k (AWP) he said when they came to pay they only had 27k available in Neteller so he had to be paid in 2 instalments. And this was from a bigger MG group.

This is not a foil-hat thread, and I originally actually asked 6 questions but as usual everyone seizes on the one that could prove all is not as it seems. kktmd has been a great help and once he accepts the freeplay mode is the same as paid games on MG, he can have the same confidence as us in his figures. The dirty work is already done; MG slots meet their TRTP over a surprisingly low amount of events, i.e. 1-2 million. This makes sense as IF the TRTP went true at say 10,000,000 spins then some players could experience dramatic beatings (bear in mind during the life of an account at any casino the average player would unlikely exceed a couple hundred thousand spins on any particular slot) while some could hit excessive winning streaks.

My idea was to dramatically raise stake to see if it somehow triggered any changes in the slots' behaviour. I suggested a few hundred K spins at lowest stake, then whacking it up by say 50x - kktmd did this and as we saw the RTP was over 98%, as opposed to 95-97 playing normally. Obviously the longer this goes on the RTP would no doubt settle back again or would it? The results aren't conclusive thus far as the sample is too small, but the clues are there to keep investigating the matter.

When meeting the RTP the slots are random, and if we look at BDBA in the million spins+ his highest win was it appears 5 gold bars/wilds on freespins, but he never hit the gems/wilds on freespins. So, the slot hit it's TRTP but in doing so never needed to produce the highest win. So, the variance can actually be slightly different on the same slot for a different block of 1 million spins.

 

[mrmark21]

You lost me when you started speaking large numbers. Very interesting thread. Just confirming what I already thought

 

[dunover]

You lost me when you started speaking large numbers. Very interesting thread. Just confirming what I already thought

Well, that's just it, we haven't confirmed much at all really thus far, except the TRTP is pretty much bang-on on MG slots with a surprisingly low number of spins. We've had people conjecturing in the past 100,million and similar. Clearly it's not the case and if it were the variance could be staggering on some games, assuming the RNG pool values are pro-rata.

We've had a few clues about changing stake significantly and it's short term effects, but we (kktmd lol...) need to check this over a far bigger sample amount of spins.

 

[Jufo]

When meeting the RTP the slots are random, and if we look at BDBA in the million spins+ his highest win was it appears 5 gold bars/wilds on freespins, but he never hit the gems/wilds on freespins. So, the slot hit it's TRTP but in doing so never needed to produce the highest win. So, the variance can actually be slightly different on the same slot for a different block of 1 million spins.

It's normal that you don't hit the highest payout in 1M spins, because the probability for that payout is usually less than 1 in million. I simulated a batch of 1 Million spins on BDBA slot and wrote down the highest win out of those 1M spins, and repeated this 5 times. The highest win for each batch of 1M spins was (with 9 lines, 1 coin per line):

Batch 1: 25125  (2791.67x Bet)
Batch 2: 37870  (4207.78x Bet)
Batch 3: 37810  (4201.11x Bet)
Batch 4: 37850  (4025.56x Bet)
Batch 5: 37575  (4175x Bet)

So it looks like that in 1 Million spins the highest win is close to ~4200x Bet, but in one of the five batches it was only 2790x Bet.

 

[Nifty29]

Exactly - the casino's exposure cannot be unlimited. I know they set max stakes etc. but even so imagine some lucky nutter betting £45 a spin on BDBA hitting the 5 wilds/gems in freespins? 187,500 pounds and then goes on to hit monsters on other slots? If we remember VWM hitting over 50k (AWP) he said when they came to pay they only had 27k available in Neteller so he had to be paid in 2 instalments. And this was from a bigger MG group.

This is not a foil-hat thread, and I originally actually asked 6 questions but as usual everyone seizes on the one that could prove all is not as it seems. kktmd has been a great help and once he accepts the freeplay mode is the same as paid games on MG, he can have the same confidence as us in his figures. The dirty work is already done; MG slots meet their TRTP over a surprisingly low amount of events, i.e. 1-2 million. This makes sense as IF the TRTP went true at say 10,000,000 spins then some players could experience dramatic beatings (bear in mind during the life of an account at any casino the average player would unlikely exceed a couple hundred thousand spins on any particular slot) while some could hit excessive winning streaks.

My idea was to dramatically raise stake to see if it somehow triggered any changes in the slots' behaviour. I suggested a few hundred K spins at lowest stake, then whacking it up by say 50x - kktmd did this and as we saw the RTP was over 98%, as opposed to 95-97 playing normally. Obviously the longer this goes on the RTP would no doubt settle back again or would it? The results aren't conclusive thus far as the sample is too small, but the clues are there to keep investigating the matter.

When meeting the RTP the slots are random, and if we look at BDBA in the million spins+ his highest win was it appears 5 gold bars/wilds on freespins, but he never hit the gems/wilds on freespins. So, the slot hit it's TRTP but in doing so never needed to produce the highest win. So, the variance can actually be slightly different on the same slot for a different block of 1 million spins.

It will even out in the long run. Unless there has been some new form of mathematics invented recently, it has to.

If the software/games are somehow influenced by suddenly increasing the bet, then there would be casinos everywhere going under because any AP worth their salt would be all over it like a cheap suit.

The problem is that either the RTP increases EVERY time you up your bet, or it NEVER does. So, it has to be mean that EVERY time you suddenly jack up your bet you will be a winner.....and we know for a fact that this is not the case.

It may be, in simple terms, that you start with $200, spin 100 times @ 0.50c, have $190 left (for instance), then jack it up to $25 a spin...and hit a 50xbet win. Well, your RTP is going to be HUGE at THAT point...if you cash out RIGHT there, you're a winner. However, what happens the times when you do NOT hit anything with the big spins? Your RTP plummets, of course. So the numbers will even out in the LONG run.

I know from experience that I LOSE far more often than WIN when I up my bet. The problem is that we only tend to remember the times we did it and "Yoohoo!!" hit a monster....we don't recall all the times we hit something nice on our normal bet and blow it all back using bigger bets.

I put it to you that IF it WAS a way to "beat" the casino, every operator would have a term stating that players cannot INCREASE their bets above a certain percentage....otherwise they would asking for it, surely?

I'm not a stats expert, but from what I read of the numbers kkmtd came up, I can't see how this phenomenon can be accurately measured using any kind of autoplay function etc. As players, we generally do not play until we hit 3xbankroll or -1/3 of bankroll and then start over etc, or increase our bet at that point until the same event occurs either way. The only way I can see the being proven via stats is actually using the REAL stats via playcheck and looking at what happens to the game/s at the point where the bet is raised...and this is going to be VERY difficult and time consuming for anyone. I just don't see how you can some up with accurate numbers based on computer simulations of what is a totally unpredicatable human behaviour. IMO, it is not the same as running 1.4 million spins @.30c...this IS an accurate measure.

If the experiment were run so that the bet was raised exactly every 100 spins (for e.g) and ran for another 100 spins at the higher bet, and was rinsed and repeated tens and thousands of times over, then I think we would find that the number for each bet size would be very similar in terms of RTP. I just don't see raising the bet at random intervals AFTER the data has been gathered using min bet really telling us anything at all....like the example I have earlier where you could have an amazing RTP if you got lucky and hit a reasonable win as soon as you raise the bet, but most times you will end up with an awful one.

Well I've confused myself now...but at least I know what I mean

I mean, a group of us could commit to trying this out (in fun mode even), and record what happens each time we raise the bet on a slot i.e. win/loss, and effect on current RTP. If anyone could be bothered of course lol.

In any event, it is a good discussion, and I am ignoring the tinfoil hat parts, as they are fairly mild compared to most I've heard, and at least those casting the theories are actually willing to look at real stats and data.

 

[mrmark21]

Well, that's just it, we haven't confirmed much at all really thus far, except the TRTP is pretty much bang-on on MG slots with a surprisingly low number of spins. We've had people conjecturing in the past 100,million and similar. Clearly it's not the case and if it were the variance could be staggering on some games, assuming the RNG pool values are pro-rata.

We've had a few clues about changing stake significantly and it's short term effects, but we (kktmd lol...) need to check this over a far bigger sample amount of spins.

Well I definitely agree with Jufos post #87 But there's allot of other interesting information on this thread. I noticed you're original post Dunover and wrote a response but before I could post my stupid Nexus10 ran out of battery

 

[dunover]

It will even out in the long run. Unless there has been some new form of mathematics invented recently, it has to.

If the software/games are somehow influenced by suddenly increasing the bet, then there would be casinos everywhere going under because any AP worth their salt would be all over it like a cheap suit.

The problem is that either the RTP increases EVERY time you up your bet, or it NEVER does. So, it has to be mean that EVERY time you suddenly jack up your bet you will be a winner.....and we know for a fact that this is not the case.

It may be, in simple terms, that you start with $200, spin 100 times @ 0.50c, have $190 left (for instance), then jack it up to $25 a spin...and hit a 50xbet win. Well, your RTP is going to be HUGE at THAT point...if you cash out RIGHT there, you're a winner. However, what happens the times when you do NOT hit anything with the big spins? Your RTP plummets, of course. So the numbers will even out in the LONG run.

I know from experience that I LOSE far more often than WIN when I up my bet. The problem is that we only tend to remember the times we did it and "Yoohoo!!" hit a monster....we don't recall all the times we hit something nice on our normal bet and blow it all back using bigger bets.

I put it to you that IF it WAS a way to "beat" the casino, every operator would have a term stating that players cannot INCREASE their bets above a certain percentage....otherwise they would asking for it, surely?

I'm not a stats expert, but from what I read of the numbers kkmtd came up, I can't see how this phenomenon can be accurately measured using any kind of autoplay function etc. As players, we generally do not play until we hit 3xbankroll or -1/3 of bankroll and then start over etc, or increase our bet at that point until the same event occurs either way. The only way I can see the being proven via stats is actually using the REAL stats via playcheck and looking at what happens to the game/s at the point where the bet is raised...and this is going to be VERY difficult and time consuming for anyone. I just don't see how you can some up with accurate numbers based on computer simulations of what is a totally unpredicatable human behaviour. IMO, it is not the same as running 1.4 million spins @.30c...this IS an accurate measure.

If the experiment were run so that the bet was raised exactly every 100 spins (for e.g) and ran for another 100 spins at the higher bet, and was rinsed and repeated tens and thousands of times over, then I think we would find that the number for each bet size would be very similar in terms of RTP. I just don't see raising the bet at random intervals AFTER the data has been gathered using min bet really telling us anything at all....like the example I have earlier where you could have an amazing RTP if you got lucky and hit a reasonable win as soon as you raise the bet, but most times you will end up with an awful one.

Well I've confused myself now...but at least I know what I mean

I mean, a group of us could commit to trying this out (in fun mode even), and record what happens each time we raise the bet on a slot i.e. win/loss, and effect on current RTP. If anyone could be bothered of course lol.

In any event, it is a good discussion, and I am ignoring the tinfoil hat parts, as they are fairly mild compared to most I've heard, and at least those casting the theories are actually willing to look at real stats and data.

Nope, we fully understand what you have just imparted. The best point there is you observing the fact that few players would hammer away at a slot at the same stake for million+ spins. I suppose it's not reflective of real-life playing, but as you say the end result is the same - TRTP being met. It's what happens in between that interests us, and it costs us nothing to satisfy our curiousity as to what if anything can be positively observed. In fact when the data is in and analysed it could be a reference for any real foilers that appear on the forum in future - send 'em here and by the time they've read the first 2 pages they'll be so baffled and lost they'll give up...

 

[mathsboy1975]

No, it would still make a huge difference to game fairness if the RTP was kept constant in some controlled, non-random fashion. This is because the software would control how much you can win and your wins would be spaced too far from each other to lower the probability that you are ever able to get ahead, so it would make hell of a lot of difference to the player, and it would not be a fair game at all.

There are games that are openly like that and they are called fruit-machines or AWPs. However those games clearly state how they operate. The player has the right to know whether he is playing a genuine random game or an AWP.

I do not quite agree here. I am not saying that the amounts you can win changes, just the probability could move in such a way that in a "dry" mode the low payout reel combos have higher than usual probability of coming in and vice versa. Over a long enough number of spins, provided that the probabilities spend equal times in "dry" and "generous" modes the TRTP would remain unchanged.

This would in no way impact the fairness - in fact you could view it as an enhanced method of gameplay (in my opinion) that could mean that when you are in a good streak it really is happy days time. Also, the probabilities would still be low, for the big payouts, just tweaked a little bit. I suppose what I am saying is something like a law of conservation of probability but taken over a large number of spins over which the probabilities do vary.

I am not saying that I think this IS the case at all, and I have no evidence to suggest otherwise - just a theory. It just seems possible that this could be implemented in such a way that the games are still completely fair and with the same TRTP.

I will definitely take a look at your report into the real money stuff though along with all the other good info that this thread has generated.

 

[kktmd]

The only way I can see the being proven via stats is actually using the REAL stats via playcheck and looking at what happens to the game/s at the point where the bet is raised...and this is going to be VERY difficult and time consuming for anyone. I just don't see how you can some up with accurate numbers based on computer simulations of what is a totally unpredicatable human behaviour. IMO, it is not the same as running 1.4 million spins @.30c...this IS an accurate measure.

...ehhh....I'm not sure I understand? Just to clarify: the 1.5 mio spins I logged were real spins, done using MG's servers (but done using Free Play Mode - not Real Money Mode). Do you think Free Play Mode is different from Real Play Mode?

 

[dunover]

...ehhh....I'm not sure I understand? Just to clarify: the 1.5 mio spins I logged was real spins, done using MG's servers (but done using Free Play Mode - not Real Money Mode). Do you think Free Play Mode is different from Real Play Mode?

No, he's not saying that. He means that a normal player will play different stakes at different times during his 'career' on any given slot. This will mean the wins are in different amounts and values to what we get at 30p continuously. He still gets the same RTP in the end assuming he plays long enough, but he might quit when on an RTP+ curve to take profit or an RTP - curve to stop a bad run. Therefore he is unlikely to see the process to conclusion which will always be (over the sheer amount of spins you make) around TRTP.

 

[Jufo]

I do not quite agree here. I am not saying that the amounts you can win changes, just the probability could move in such a way that in a "dry" mode the low payout reel combos have higher than usual probability of coming in and vice versa. Over a long enough number of spins, provided that the probabilities spend equal times in "dry" and "generous" modes the TRTP would remain unchanged.

This would in no way impact the fairness - in fact you could view it as an enhanced method of gameplay (in my opinion) that could mean that when you are in a good streak it really is happy days time. Also, the probabilities would still be low, for the big payouts, just tweaked a little bit. I suppose what I am saying is something like a law of conservation of probability but taken over a large number of spins over which the probabilities do vary.

I am not saying that I think this IS the case at all, and I have no evidence to suggest otherwise - just a theory. It just seems possible that this could be implemented in such a way that the games are still completely fair and with the same TRTP.

I cannot see any possible reason for a slot developer to implement it like that because the exact similar behaviour (winning and losing phases) are generated by pure randomness as well. You suggest that someone would program an extra layer into the game for no difference whatsoever to the player.

We are lead to believe that slots work simply by drawing random positions on the reel strips to stop at. I have seen the exact slot design blueprints from one casino software and their slots have been implemented exactly like that. Remember also that gaming license regulations require each outcome to be independent from previous outcomes, something which would not hold in your example.

Therefore we should assume this to be the case and run statistical tests from this hypothesis. When you do statistical tests you have to have some default assumption to base the tests upon.

I will definitely take a look at your report into the real money stuff though along with all the other good info that this thread has generated.

Yeah please do, but only after your hangover goes away, it may be too heavy stuff to read in hangover. I am interested to hear if you have any ideas on how to test the randomness.

 

[Nifty29]

Nope, we fully understand what you have just imparted. The best point there is you observing the fact that few players would hammer away at a slot at the same stake for million+ spins. I suppose it's not reflective of real-life playing, but as you say the end result is the same - TRTP being met. It's what happens in between that interests us, and it costs us nothing to satisfy our curiousity as to what if anything can be positively observed. In fact when the data is in and analysed it could be a reference for any real foilers that appear on the forum in future - send 'em here and by the time they've read the first 2 pages they'll be so baffled and lost they'll give up...

Yes....it kinda shows what I've always said.....that TRTP is really about the operator and is of real tangible use as a statistic only to them to enable them to run their business and marketing with some degree of certainty and stability.

TRTP is really only rough indicator to a player, since most players never play enough spins to be in the realms of the LLN. Its a bit like music notes with no Clef....they will give a a rough idea of how much higher or lower the tones should be but are not accurate I.e. not precise or guaranteed.

In the end, just because you're playing a higher TRTP game, doesn't mean that you're going to win more than you might have on a lower TRTP. What really matters to the average Joe is their bankroll and making it last, and that's why it is just as important, or even more important, to know the variance of a game. I think too many players put too much thought into TRTP and don't think about other equally important aspects of gambling.

 

[mathsboy1975]

We are lead to believe that slots work simply by drawing random positions on the reel strips to stop at. I have seen the exact slot design blueprints from one casino software and their slots have been implemented exactly like that. Remember also that gaming license regulations require each outcome to be independent from previous outcomes, something which would not hold in your example.

Ah I was unaware of both of these aspects. From everything here I have managed to read so far on the forum I was lead to believe that the established best guess on the software operation was not to do with reel positions at all but via a pure RNG that decides the WIN amount then the slot front end just selects which reel symbols to display that agrees with the designated win amount.

As for the gaming licence issue of outcome independence then that really does put the nail in the coffin on the theory of non-constant probabilities.

 

[mathsboy1975]

Yeah please do, but only after your hangover goes away, it may be too heavy stuff to read in hangover. I am interested to hear if you have any ideas on how to test the randomness.

Just read your report. To be honest that is precisely how I would go about testing fairness of the slots. Stratify the price ranges into a workable yet diverse enough amount and do a Pearson chi-square test on observed to expected values.

I assume that the reason that the expected loss curve on your plots is non-linear is simply due to the players variation in staking pattern?

Also, how did you arrive at your expected values? Did you use empirical probabilities from free play spins or did you base it on uniform probability distribution applied to a known reel symbol layout?

It is a pretty conclusive proof that the slots are fair. That would be a really good article to educate conspiracy theorists.

 

[dunover]

Ah I was unaware of both of these aspects. From everything here I have managed to read so far on the forum I was lead to believe that the established best guess on the software operation was not to do with reel positions at all but via a pure RNG that decides the WIN amount then the slot front end just selects which reel symbols to display that agrees with the designated win amount.

As for the gaming licence issue of outcome independence then that really does put the nail in the coffin on the theory of non-constant probabilities.

You are probably right, as the reels graphics can be generated ex post facto on online slots after the actual win amount is selected by the RNG. In fact on MG I've always suspected this is the case, and on WMS games I'm pretty certain. This doesn't mean that they don't appear to act like Jufo says, like mechanical slots with separate reel stops - the result is exactly the same after all; if you get the reels strips and analyse the values and probabilities they exactly match the paytable/TRTP of the game.

 

[Jufo]

Ah I was unaware of both of these aspects. From everything here I have managed to read so far on the forum I was lead to believe that the established best guess on the software operation was not to do with reel positions at all but via a pure RNG that decides the WIN amount then the slot front end just selects which reel symbols to display that agrees with the designated win amount.

You have read too many posts by conspiracy theorists over here if they actually made you believe that. It wouldn't work well for 5-reel slots to simply draw win amount from RNG and then make the reels match that payout just for eye candy because there are so many possible positions the reels may land in (for example with 24 symbols on each reel there are 24^5 = 7962624 possible unique reel positions) and also thousands of possible win amounts from a single spin. Drawing a win amount from RNG and then make the game display that win amount is probably something that is used with online scratchcards though.

As for the gaming licence issue of outcome independence then that really does put the nail in the coffin on the theory of non-constant probabilities.

Well yes, but we have seen many cases where these license regulations have been violated anyway (like in the recent Finsoft incident) so we shouldn't also take for granted that every licensed software is random and fair.

 

[Jufo]

Just read your report. To be honest that is precisely how I would go about testing fairness of the slots. Stratify the price ranges into a workable yet diverse enough amount and do a Pearson chi-square test on observed to expected values.

I assume that the reason that the expected loss curve on your plots is non-linear is simply due to the players variation in staking pattern?

Yes, the player varied his bets from $0.09 per spin to $18 per spin so he is betting big where the expected loss curve drops steeper and vice versa. The expected loss curve is simply plotted in Excel as: (100% - 95.46%)*Total_wagered, in other words I keep track of player's cumulative total wagered and multiply this by the house edge of the game (4.54%) to get his expected loss at any point.

Also, how did you arrive at your expected values? Did you use empirical probabilities from free play spins or did you base it on uniform probability distribution applied to a known reel symbol layout?

I have programmed an off-line simulator for several Microgaming slots based on their known reel layouts and paytables. With this simulator I can generate millions of spins and estimate the probabilities for each payout. So the results I get from my simulator are the ones you would get from a fair and random slot.

If you want to have a look how the simulator works I uploaded a small demo version to:

You do not have permission to view link Log in or register now.

Just download the Zip-file, extract it, browse to either slot's folder and launch 'play.bat' either by double-clicking in Windows explorer or from Command prompt. The simulator is made with Java so you will need Java 7 runtime installed to run the simulator. You can get the latest Java runtime from here:

You do not have permission to view link Log in or register now.

In the demo you get to play text version of the slot. Each character represents a slot symbol which are declared in the attached .jpg file.

It is a pretty conclusive proof that the slots are fair. That would be a really good article to educate conspiracy theorists.

Well, my analysis of this player's data over 200,000 spins established that MG slots A) meet their theoretical RTP B) deal out payouts that are completely in line with the payout distribution you would get from completely random reels. However there are still a couple of things:

- In the bankroll graph the player ended up suspiciously close to expected return, as if the software forced him there.
- The chi-square test for comparing payout distributions gave a value of 0.999.. which to me seems almost too good match. So yes, you will get payouts that are completely in line with expectations but it seems that they are almost too close to expectations.

Therefore I am not entirely convinced of MG game fairness yet and I feel like measuring the distances between high payouts that we discussed yesterday, might be a good idea.

 

[dunover]

Yes, the player varied his bets from $0.09 per spin to $18 per spin so he is betting big where the expected loss curve drops steeper and vice versa. The expected loss curve is simply plotted in Excel as: (100% - 95.46%)*Total_wagered, in other words I keep track of player's cumulative total wagered and multiply this by the house edge of the game (4.54%) to get his expected loss at any point.



I have programmed an off-line simulator for several Microgaming slots based on their known reel layouts and paytables. With this simulator I can generate millions of spins and estimate the probabilities for each payout. So the results I get from my simulator are the ones you would get from a fair and random slot.

If you want to have a look how the simulator works I uploaded a small demo version to:

You do not have permission to view link Log in or register now.

Just download the Zip-file, extract it, browse to either slot's folder and launch 'play.bat' either by double-clicking in Windows explorer or from Command prompt. The simulator is made with Java so you will need Java 7 runtime installed to run the simulator. You can get the latest Java runtime from here:

You do not have permission to view link Log in or register now.

In the demo you get to play text version of the slot. Each character represents a slot symbol which are declared in the attached .jpg file.



Well, my analysis of this player's data over 200,000 spins established that MG slots A) meet their theoretical RTP B) deal out payouts that are completely in line with the payout distribution you would get from completely random reels. However there are still a couple of things:

- In the bankroll graph the player ended up suspiciously close to expected return, as if the software forced him there.
- The chi-square test for comparing payout distributions gave a value of 0.999.. which to me seems almost too good match. So yes, you will get payouts that are completely in line with expectations but it seems that they are almost too close to expectations.

Therefore I am not entirely convinced of MG game fairness yet and I feel like measuring the distances between high payouts that we discussed yesterday, might be a good idea.

This is the sort of paragraph that will get foil-hat names thrown at you, or 'it's rigged I tells ya' accusations........
But there are observations that we can make safely as long as we have the facts/data to back them up. This is one of them. Your line about 'payout meeting expectations, as if it forced him there' is the whole reason I posted my now infamous question 3 in my original post. This is why I requested kktmd (as you may have guessed) to run tests with big changes in stake. Your example has a change of 200* from lowest to highest and then you saw the slot 'forced' the player to the TRTP over a quite short period. So it's almost certainly going to transpire that there is going to be a middle ground between the '100% random' advocates and the foil-hatters.
The insinuation you make, and I spotted, are that the way the slots behave mean there HAS to be a correlation between certain changes in stake and short-term outcomes. This isn't surprising to me, but may be to others.
Your example shows the player played and lost hundreds of thousands over hundreds of thousands of spins, but ended up spot on the TRTP. This HAS to mean that had the player been aware of what we are discussing here, that say his RTP was only 93% at a point near the end of the data, he could (and I hate this word) almost predict the gain he would make from the next cycle of play before he terminated the sequence. In other words, if he had lost 300k and was showing 93% RTP, then by dividing the losses by 100 and multiplying them by 95 and projecting that over his next stakes that led to the end of the cycle, he could work out that his net gain on THAT session would be 2% of the total losses - before he even played it.
This shouldn't occur on a totally random slot; the randomness is there but the only way I can see of the programme managing this is that pool of values alters in some way in order to 'manage' the final outcome at TRTP (in other words 'force' him there as you said) by say removing more 'zero' events and adding in more 'premium win' outcomes. Do you agree this could happen?