Zero 5* in 29 pulls?

entrailbucket

I've had that conversation like 10,000 times...this is crossing quite close to an exploit that we're not allowed to talk about here, but what those players are seeing is not what they think they're seeing.

There's no "seed" per se, not in the way they're thinking there is. To be clear: I have not reviewed their code.  I don't know exactly how they've chosen to do this. 

But as of 2014 when the game launched, there is no need for anyone to write some bespoke seeded RNG function to do this stuff, and nobody would.  This game runs on AWS, I think?...you just use whatever random function AWS gives you.  If it's good enough for Amazon and like 100,000 other games, it's good enough for this one.

The details of their specific implementation might be interesting or relevant if you're trying to do something nefarious, but for the rest of us, it's random.

ThaRoadWarrior

Yeah I really didn't want to get into the exploit per se.

At any rate, I've been tracking my Latest pulls since december 2018, and my current 5* ratio (after removing 40 pre-shard bonus hero covers from the totals) is 403 organic 5* cover pulls out of 2,702 total Latest Lengends pulls, for a 14.91% hit rate in the last couple years. Almost never has it been one 5* every 7th pull, it's nearly always something like 10 4*s, then a couple 5*s in reasonably close succession, sometimes it'll be excessive dry spells over 30 followed by clumpy 5* pulls...anyway I do like tracking it because it gives you an idea of how long it has actually been since you landed a 5* in any given store. I track classic that way too. But if you're someone who is pulling from both stores, and you have say a 10 cover dry spell in each one, you "feel" like you're 20 pulls deep into that dry spell even though you're pulling across two different pools of RNG. So I find that it can be reassuring to know you're 36 dry in Latest, but only 4 dry in Classic for instance. Even though RNG doesn't "owe" you any 5*s, if you stay pulling and tracking long enough you'll see that your overall pull rate will balance out.

Since I've been pulling significantly less Classic across that time, having committed to staying current in lates by pulling as I go as an experiment for much of it, since 6/26/2018 I have pulled 56 5*s out of 419 Classic pulls (again removing 11 pre-shard bonus covers from that) for a slightly lower 13.37% hit rate in classic from a much smaller pool.

And actually come to think of it, some of those bonus covers were probably 4*s, but i don't have a clean way to extract those from my Google Sheets totals, so we'll just go with those numbers.

entrailbucket

So, you have a sample size of 2,700 pulls and your rate is close.  Daredevil has a similar sample size and his rate is close.  An 0/29 stretch somewhere in there might be noticeable at the time, but it'd barely even affect the overall rate.

The lunar new year store was really popular, and players I know pulled there hundreds of times.  If there were (let's say) 20,000 overall pulls on that store, I'd expect the number of total 5* given out was quite close to the 15% odds.  One subset of 29 pulls isn't anywhere near significant when you're talking about those kinds of numbers.

danielrandkai

It's unfortunate that they didn't go with gd Shang Chi for a Chinese New Year store. He's my favorite toon, and I love taking down meta teams with him. Because he's incredibly easy to take down on defense. Nicely balanced, as design should be. 

Bad

Truth is that I did always on the LL store.
However I don't think that the system will change for any other store.
It's easy to get a 5*. But as all of you know on stadistics everything can happen. 
After my experience I can say that the algorithm shuffles the 5* covers in a close to similar proportion on the 3 characters and their powers in order to be able to champ the 3 in a given number of pulls.
Because the numbers of the possibilities for me doing it 2 times on those pulls only would show that it was a miracle or a lie.
So it's as clear to me like that. 
Until the rng changes, of course. 

helix72

I recently pulled 760 tokens from the CNY vault. Twice I had runs of 35 4* without a single 5*, yet I still managed an above average 18.2% rate over those total 760 pulls including those two long dry streaks.

The last time a vault this good ran my pull rate was 8.2% over 100 pulls or so.

Yes it’s frustrating. Yes it sucks. But it doesn’t mean there’s anything bugged in the algorithm.

Multiple times people have suggested “streak breakers”, something like where if you go X pulls without a 5* you are guaranteed a 5* on your next pull. They’d have to lower the success rate a bit to keep the overall rate the same. Devs response was to give us bonus shards.

As for rate, the folks that have been tracking it the longest believe it is 15% not 1/7. This is based on years of collecting pull rates as well as the fact that if you look, the pull rate is shown not as 1/7 but as “~1/7”. That “~” means “approximately” and 1/7 is approximately 15%.

Daredevil217

helix72 said:

I recently pulled 760 tokens from the CNY vault. Twice I had runs of 35 4* without a single 5*, yet I still managed an above average 18.2% rate over those total 760 pulls including those two long dry streaks.

The last time a vault this good ran my pull rate was 8.2% over 100 pulls or so.

Yes it’s frustrating. Yes it sucks. But it doesn’t mean there’s anything bugged in the algorithm.

Multiple times people have suggested “streak breakers”, something like where if you go X pulls without a 5* you are guaranteed a 5* on your next pull. They’d have to lower the success rate a bit to keep the overall rate the same. Devs response was to give us bonus shards.

As for rate, the folks that have been tracking it the longest believe it is 15% not 1/7. This is based on years of collecting pull rates as well as the fact that if you look, the pull rate is shown not as 1/7 but as “~1/7”. That “~” means “approximately” and 1/7 is approximately 15%.

Sweet.  Hopefully I have some good luck coming!

Vhailorx

ThaRoadWarrior said:

That search is less about PRNG and more about when it is performed and how often, ie does every player get their RNG seeded one time, and then they are just redeeming along that string of pulls for the rest of their MPQ days, or was each pull a dip into the RNG uniquely. People used to speak on that topic with some authority but I haven’t located the source of that knowledge so I don’t really want to comment on it one way or the other.

It's definitely closer to the former (seeded once and a set stream forever) than the latter (each pull is uniquely generated at the time of pull).  Like most loot box games there were some early controversies over exploiting client-side RNG calculations, but things have been pretty stable for some time now.

RNG is all calculated server side and is seeded once per store and each player just pulls against that stream of predetermined outcomes out to forever. (The per-store seeding is why daredevil points out that the specialty stores are more like gambling.) I seem to recall  that the seeds for long-lssting stores (ie the latest store) have been reset once or twice over the past 4 years, but I can't find any documentation of that on the forums (nor would I expect to as it would be an invisible change to most users).

I do not know if anyone has tested the periodically repeating LT stores (e.g. fan favorites) to see if they reseed every year.

Vhailorx

Daredevil217 said:

helix72 said:

I recently pulled 760 tokens from the CNY vault. Twice I had runs of 35 4* without a single 5*, yet I still managed an above average 18.2% rate over those total 760 pulls including those two long dry streaks.

The last time a vault this good ran my pull rate was 8.2% over 100 pulls or so.

Yes it’s frustrating. Yes it sucks. But it doesn’t mean there’s anything bugged in the algorithm.

Multiple times people have suggested “streak breakers”, something like where if you go X pulls without a 5* you are guaranteed a 5* on your next pull. They’d have to lower the success rate a bit to keep the overall rate the same. Devs response was to give us bonus shards.

As for rate, the folks that have been tracking it the longest believe it is 15% not 1/7. This is based on years of collecting pull rates as well as the fact that if you look, the pull rate is shown not as 1/7 but as “~1/7”. That “~” means “approximately” and 1/7 is approximately 15%.

Sweet.  Hopefully I have some good luck coming!

Gambler's Fallacy Alert!

entrailbucket

Vhailorx said:

ThaRoadWarrior said:

That search is less about PRNG and more about when it is performed and how often, ie does every player get their RNG seeded one time, and then they are just redeeming along that string of pulls for the rest of their MPQ days, or was each pull a dip into the RNG uniquely. People used to speak on that topic with some authority but I haven’t located the source of that knowledge so I don’t really want to comment on it one way or the other.

It's definitely closer to the former (seeded once and a set stream forever) than the latter (each pull is uniquely generated at the time of pull).  Like most loot box games there were some early controversies over exploiting client-side RNG calculations, but things have been pretty stable for some time now.

RNG is all calculated server side and is seeded once per store and each player just pulls against that stream of predetermined outcomes out to forever. (The per-store seeding is why daredevil points out that the specialty stores are more like gambling.) I seem to recall  that the seeds for long-lssting stores (ie the latest store) have been reset once or twice over the past 4 years, but I can't find any documentation of that on the forums (nor would I expect to as it would be an invisible change to most users).

I do not know if anyone has tested the periodically repeating LT stores (e.g. fan favorites) to see if they reseed every year.

I don't think "seed" is the word you're looking for...that has a specific meaning when you're talking about random number generators.  You're just saying the outcomes are predetermined.

Again we're very close to discussing an exploit here, but if I roll a die 20 times and record each result in order, then read them to you one at a time, those individual rolls were predetermined but they are still random.

HoundofShadow

There's another similar phenomena which I see frequently in reddit MPQ. It has to do with pulling from Taco Vault: Spicy, which has 40 items, and in it, a random 4* is the most precious item. Unlike pulling from stores, pulling from vaults increases the probability of the item you want as you pull more.

The typical reactions when players pull the 4* on the 40th pull or close to the 40th pull are "the game is rigged."

However, I've never seen any player saying "the game is rigged" when they pull a 4* cover on the 1st pull. What they attribute to this is "luck", rather than saying the dev rigged the game in their favours.

KGB

entrailbucket said:

Vhailorx said:

ThaRoadWarrior said:

That search is less about PRNG and more about when it is performed and how often, ie does every player get their RNG seeded one time, and then they are just redeeming along that string of pulls for the rest of their MPQ days, or was each pull a dip into the RNG uniquely. People used to speak on that topic with some authority but I haven’t located the source of that knowledge so I don’t really want to comment on it one way or the other.

It's definitely closer to the former (seeded once and a set stream forever) than the latter (each pull is uniquely generated at the time of pull).  Like most loot box games there were some early controversies over exploiting client-side RNG calculations, but things have been pretty stable for some time now.

RNG is all calculated server side and is seeded once per store and each player just pulls against that stream of predetermined outcomes out to forever. (The per-store seeding is why daredevil points out that the specialty stores are more like gambling.) I seem to recall  that the seeds for long-lssting stores (ie the latest store) have been reset once or twice over the past 4 years, but I can't find any documentation of that on the forums (nor would I expect to as it would be an invisible change to most users).

I do not know if anyone has tested the periodically repeating LT stores (e.g. fan favorites) to see if they reseed every year.

I don't think "seed" is the word you're looking for...that has a specific meaning when you're talking about random number generators.  You're just saying the outcomes are predetermined.

Again we're very close to discussing an exploit here, but if I roll a die 20 times and record each result in order, then read them to you one at a time, those individual rolls were predetermined but they are still random.

Pretty sure he's saying the same thing you are. He's just using the word 'seed' to mean the game pre-rolls a massive number of random numbers and stores them in order server side (like your 20 die roll example). Then each time you open a token it gives you the next stored number. This was done because of the exploit a few years back which you and Vhailorx both mentioned and which many of us remember.

While seed has a specific meaning for random number generators it's also a reasonably common term that's often bandied about to denote the beginning of a sequence of random numbers.

KGB

P.S. The problem with the long run percentages equaling ~1/7 is that it's the short run that matters. First because of human perception (0-29 streaks suck, we've all been there if we've opened tokens long enough). Secondly, and more importantly, what you are drawing for isn't always the same (ie in Vegas, if you win, you always win the same amount of cash for the same bet when paying ****)  because as we all know the 5* are not equal. So having really bad luck in an important store like this one with 3 meta characters matters MUCH more than having bad luck in say the current latest legends where there are no meta characters.

entrailbucket

I know, I was just looking to avoid having the argument about the MPQ random number generator being "seeded" in a way that produces a non-random sequence.  Somebody was going to jump on that post even though that's definitely not what he meant.


Having bad luck sucks (I've gone 0/100 before, and 0/50 several times, but I've opened a LOT of tokens).  Complaining about bad luck is fine, and we all do it.  But bad luck is just bad luck -- it doesn't mean that the system is broken.

revskip

While I sure wouldn't mind seeing a pity counter enacted to reduce really brutal runs i don't know that 29 pulls would even trigger that.  

I've gone 0-50.  I've also gone 5 of 7.  I remember the former a lot more than the latter even though the latter was way more satisfying.  

Vhailorx

entrailbucket said:

Vhailorx said:

ThaRoadWarrior said:

That search is less about PRNG and more about when it is performed and how often, ie does every player get their RNG seeded one time, and then they are just redeeming along that string of pulls for the rest of their MPQ days, or was each pull a dip into the RNG uniquely. People used to speak on that topic with some authority but I haven’t located the source of that knowledge so I don’t really want to comment on it one way or the other.

It's definitely closer to the former (seeded once and a set stream forever) than the latter (each pull is uniquely generated at the time of pull).  Like most loot box games there were some early controversies over exploiting client-side RNG calculations, but things have been pretty stable for some time now.

RNG is all calculated server side and is seeded once per store and each player just pulls against that stream of predetermined outcomes out to forever. (The per-store seeding is why daredevil points out that the specialty stores are more like gambling.) I seem to recall  that the seeds for long-lssting stores (ie the latest store) have been reset once or twice over the past 4 years, but I can't find any documentation of that on the forums (nor would I expect to as it would be an invisible change to most users).

I do not know if anyone has tested the periodically repeating LT stores (e.g. fan favorites) to see if they reseed every year.

I don't think "seed" is the word you're looking for...that has a specific meaning when you're talking about random number generators.  You're just saying the outcomes are predetermined.

Again we're very close to discussing an exploit here, but if I roll a die 20 times and record each result in order, then read them to you one at a time, those individual rolls were predetermined but they are still random.

My understanding of RNG programs is that they use an arbitrary "seed" value (eg a player iD number) as the starting point for an algorithm that generates randomized numbers.  (And as mentioned earlier in the thread, if the algorithm isn't good enough, the generated numbers will repeat too quickly and are predictable. But I think that is less of a concern nowadays than it was, say, 15 years ago as the programs are much better). So the baaic principle is a (1) fixed, arbitrary value that differentiates player A from player B, and (2) an algorithmic black box that takes (1) as input and produces an infinite series outputs that are both random and also unique for each input  if I am wing about how these RNGs with I would love to be corrected. I am no expert.

So "seed" as I am using it means putting a fixed value into the algorithm. When a new mpq account is created, the latest store is seeded for that account and all LT pile from n=1 out to infinity are fixed no matter when they are used (though the character pool in LTs changes over time). And if the LT store is ever "reseeded" for any reason, then an entirely new set of pulls, starting from the next pull, would be generated.

Not trying to discuss any exploits.

Srheer0

Siv75 said:

Honestly, how does game mechanics work when you cant pull one 5* cover in 29 pulls? If odd are supposed to be 1 in 7, how does the game allow such outliers as 0 in 29 pulls? 

This entirely breaks the game as far as I'm concerned. It takes months of daily effort to save enough CPs to wait for one good store to come by. And when this happens... it is so demotivating I find it difficult to continue playing. 

Statistically, the chances of this happening by random is minuscule but this has happened to me 3 times so far with special stores. It is beyond ridiculous and creates such a negative experience I'm finding it difficult to find a reason to continue playing this game. 

Sure, I may not spend thousands to get  charachters to 550 but I have spent money being VIP...little that I can afford. All some of us would be contented with is a slow progress towards something...maybe championing a 5*. But what just happened means no progress for the last few months of effort. 

How can the game be designed to allow this? It is simply not good and you should do better. 

Absolutely ridiculous 

I personally think that special stores are rigged to not payout to me in the first bunch of pulls. I only have anecdotal evidence of this, but every time I do a small amount of pulls in a special store, I don't usually get many 5s out of it.  

I think my dryest spell record from LL was 0 in 36.  

As for your comment about no progress? Are you just selling the 4star covers you pulled? Every cover you earn is progress. Be is a 1star for 100 iso8, a 2star for champ rewards, or a 4star giving you a LL pull / 5star shards.  

Here's a challenge to you, save up enough CP for 300 pulls. By the time that happens, there should be some characters you like the look of. Get into the mindset of you have no cp to spend. Don't break early. Stay strong. 

@siv75 

LavaManLee

This discussion seems to crop up every now and then and the one thing I will add to this one is that the human brain has a hard time distinguishing "unlikely" with "impossible".  With a 1:7 chance, it is very unlikely you would pull no 5* in 29 pulls.  But it is possible.  It is definitely not "impossible" and it does not mean the game is cheating.

However, the human brain likes fairness and somehow it does not seem fair and therefore becomes more about the game cheating than the rational explanation that it is possible but not likely.

DAZ0273

It also depends on resources. If player has 300 pulls available and 29 of them in a row yield 4* they are bitterly disappointed but not bankrupt. However if they have 30 pulls and 29 of them are 4* then the reaction is going to be very upset indeed and more extreme because they have got *nothing* from what they were attempting, especially from a store that is enticing like this one.

So it can be a matter of perspective.

ThaRoadWarrior

Perspective on what this game is is important. The only thing you can effectively manage is your own mindset and your own expectations. Once you understand that this game is a roster management game where the core mechanic is opening tokens, you can frame your plans against the real final boss of MPQ: RNGeezus. If you want to progress on roster, there are only a few ways to do that effectively (and infinite ways to do it ineffectually, before the "everyone is right" crowd jumps in with their sophistry). The game would like you to be spending your resources suboptimally so you always feel behindthe 8-ball, but yet tantalized into wanting to take one more pull of the slot machine, so it generally gives you opportunities to chase desirable characters in short-run stores semi regularly alongside the opportunities to pull for longer periods of time in more persistent stores like Latest and Classic. If you are grinding for effectively all available progression CP, and only spending your resources into Latest Legends, it is possible to keep up with the release schedule barring some really bad luck. If you want to have specific characters extrememly highly leveled, then you are either going to have to go all in on full dilution in classic and arrive there sometime around the death of our sun, or you are going to need to save up resources until you have the  generally agreed upon amount that you'll need to overcome statistics, which i gather takes about a year of hard play? 

The "best" thing to do (in my opinion) to generally enjoy the game if you aren't a super competitive person is to try and get everyone on roster, and settle in for the long-haul of grinding the feeder system for the upper-tier characters you want. Whether you choose to commit to latest or classic is then up to you, but I would say that dumping all your resources into a single bottomless pit is going to give you the most consistent pull rate experience. This is in essence the digital version of collecting trading cards for a CCG - you can buy a case of let's say Pokemon cards, and not be gauranteed to get a whole set due to artificial scarcity and blind box packaging, and after opening 10-20 packs you're going to discover that an increasing percentage of every pack you open is a duplicate, and you'll have your hopes up higher and higher for that single card you really want. The advantage you have here in this game is that all those duplicates can be applied as champ levels that then feed into your ability to progress towards what you want through shards, tokens, cp, explicit covers, etc, so if you seriously intend to build a roster, you need to make sure that you have all the commons in place so that the bulk covers you are pulling in will mean something to you rather than getting burned down for a few hundred iso-8.

MegaBee

That reply from Customer Service is what I expected to see.