The problem with computer random number generation, if it truly is just Random.next(), is that computer random number generation isn't really random.
The other problem, is I understand Gambler's Fallacy; I even read a few Wikipedia articles on it. Using a coin as an example of statistical results, we also have the concept of a fair coin, or unbiased coin. Each flip should give you a 50% chance of heads or tails. After 100 flips, you would normally expect approximately 50 heads and 50 tails, give or take. This does not mean that if you flip the coin 5 times and it's heads 5 times that it should be tails the next time or next 5 times. However, if you flip a coin 100 times and it's heads each time, you have to start to question whether this coin is really giving you random results, or whether it's weighted in one way or another.
During the Zaul meta, I specifically counted the times that Zaul dodged versus the times he got hit. The stated probably would expect a dodge about 47% of the time (2% base, 20% zeta leader, 25% if they took out a member of your team). I measured this during each battle and Zaul would dodge me 6 or 7 out of every 8 times. It was fairly consistent, even. Do I have enough data for a statistically significant analysis? I don't know. But not being able to look at their code, I have enough data to assume and be annoyed at CG for adding extra dodge chance to the defense during arena battles (possibly all defense).
Is this accurate? Still can't be said without being able to dig into their code. That's the fun thing about statistical analysis in a game. You'll never be able to really tell just by analyzing the results alone. You have to look at the code to know for sure. I have seen enough, however, non-biased data to make me annoyed at CG for that mechanic (non-biased = I recorded each dodge and hit for quite some time; I didn't just ignore hits).
But am I accurate in my assumption? I don't know. Could just be a bug; could just be annoyingly bad computer psuedo-RNG. As a programmer myself, I've dealt with computer randomness to know that it could both not be their fault, and be totally controllable/fixable on their end.
The problem with computer random number generation, if it truly is just Random.next(), is that computer random number generation isn't really random.
That's not a problem because the distribution of pseudo random numbers generated by modern software is by all means enough random for applications like this one. Moreover since the random numbers are server generated there are even more randomness factors.
If you notice any "anomaly" it's not because pseudo random generators are used, it's because yoru concept of "randomness" is fallacious or because there are bugs in the code (can't really dismiss the second one in swgoh)
The problem with computer random number generation, if it truly is just Random.next(), is that computer random number generation isn't really random.
That's very true, and the distribution of the random numbers and its general streakiness depends on the method used. I certainly hope that's the case, since if you search for "Activision patent" there is definitely some shady stuff going on in the world of microtransactions right now. I'm giving EA/CG the benefit of the doubt for now, since I've played for nearly 2 years now and I've gone through my share of bad and good streaks in all aspects of the game.
Ask any statistics professor to analyze lists created by human generated randomness vs computer generated randomness with the same p, and they will most likely be able to pick out the human generated list. Basically because humans avoid too much streakiness in probability, and when we see it in a computer generated event like this game, alarm bells go off, even if there isn't anything nefarious occurring.
I 100% believe this game is streaky. It's like they because it's designed to be. People are more likely to use crystals to sim more when they get a great sim (they feel "hot" and want to do it again) or when they have a horrible sim (they get frustrated and want to do it again to compensate for what they didn't get).
I tracked my drop rate a long time ago recording the number of cantina shard drops when simming 12 at a time. If it was truly random I would have seen a Bell shaped curve peaking somewhere around 4. The curve was much too flat and didn't resemble the curve it should at all.
There are just so many extremes in this game. Raid gear drops is another. Every single person I talk to has 1 specific raid gear that drops all the time. I'm positive I get furnaces over half of the time (I feel like it might be as high as 75%, but that's likely just my bias). For as long as I've been playing I shouldn't be getting a drop rate double what RNG tells me to expect.
This only applies to human players. The AI has a 100% chance to dodge and apply bonuses to all ships and characters every time. Imperial Tie Fighter, Luminara evasion, Dooku Counter Attack, Maul Counter Attack, Wedge bonus attack, Biggs calling assist.
This only applies to human players. The AI has a 100% chance to dodge and apply bonuses to all ships and characters every time. Imperial Tie Fighter, Luminara evasion, Dooku Counter Attack, Maul Counter Attack, Wedge bonus attack, Biggs calling assist.
Maul doesn't counterattack, Dooku *does* have a 100% counter rate, you can see it in his Unique ability. Biggs always calls Wedge to assist. These are all stated abilities that work for players and AI. Luminara most definitely does not have a 100% evasion, otherwise she'd be on every team so you could win on the defense.
What's the matter with you? Why are you like this?
Ask any statistics professor to analyze lists created by human generated randomness vs computer generated randomness with the same p, and they will most likely be able to pick out the human generated list. Basically because humans avoid too much streakiness in probability, and when we see it in a computer generated event like this game, alarm bells go off, even if there isn't anything nefarious occurring.
I 100% believe this game is streaky. It's like they because it's designed to be. People are more likely to use crystals to sim more when they get a great sim (they feel "hot" and want to do it again) or when they have a horrible sim (they get frustrated and want to do it again to compensate for what they didn't get).
I tracked my drop rate a long time ago recording the number of cantina shard drops when simming 12 at a time. If it was truly random I would have seen a Bell shaped curve peaking somewhere around 4. The curve was much too flat and didn't resemble the curve it should at all.
There are just so many extremes in this game. Raid gear drops is another. Every single person I talk to has 1 specific raid gear that drops all the time. I'm positive I get furnaces over half of the time (I feel like it might be as high as 75%, but that's likely just my bias). For as long as I've been playing I shouldn't be getting a drop rate double what RNG tells me to expect.
It wouldn't be bell-shaped. The expected value is to the left of the center, on the 0 side. The lowrr values would bunch up and the distribution would be skewed. Would probably look closer to a Chi-Square distribution.
Its good to see that others have experienced the same frustrations in their game play. I'm just happy that everyone is exposed too, and suffers the same RNG system - which in my mind is the definition of fairness.
I'm a math tutor, and seeing all these different attempts at applying statistics to this game is making my head spin, lol. Here is my take on this topic:
It is nearly impossible for any one person to figure out the exact odds and randomness of the game because their sample is way too small. It is also a bad idea to try and use aggregate data via each player self-reporting their results because that opens the door to all sorts of bias that you cannot account for. The only real way to figure out the odds of different things and how randomness is simulated would be data-mining (*hint, hint* people who know how to do that).
This brings me to the real issue here. The issue is not that we don't like the current odds (although that may be a separate issue, it is not the issue here), don't like the method of RNG used, or that we don't like streaks. The issue is that we don't know any of those things. In other words, we really just want to know what the odds are and how randomness is simulated in the game. That is it. It would make things a lot easier to understand for players, and there would be a lot less of these threads.
RGN is RGN. Its not SIYF ( sometimes in your favour )
I am having a tragic run with zeta drops on my main account. ive had 4 zeta mats drop in 13 weeks so... if ships are dodging your attacks lots of times , you are likley getting a greater zeta drop then i am ( eveeyone is getting a better zeta drop then i am )
In other words, we really just want to know what the odds are and how randomness is simulated in the game. That is it. It would make things a lot easier to understand for players, and there would be a lot less of these threads.
Agreed. Forecasting odds for swgoh is just fallacy.
Though, i would add, we want to know whether we're personal doing better then the next guy/girl. Knowing someone is getting it a lot worst puts things into perspective - unless you ARE that bottom person.
Ask any statistics professor to analyze lists created by human generated randomness vs computer generated randomness with the same p, and they will most likely be able to pick out the human generated list. Basically because humans avoid too much streakiness in probability, and when we see it in a computer generated event like this game, alarm bells go off, even if there isn't anything nefarious occurring.
I don't think anyone would think a human could realistically simulate randomness, but that doesn't mean humans can't (sometimes) detect bias when evaluating data.
The obvious bias when people pick numbers "randomly" is that if they choose a low number between 1-100, their next number will usually be something that divides the largest partition. Like students populating an empty classroom. The 2nd person won't sit right next to the first person, they will keep partitioning the remaining space pretty evenly.
Ask any statistics professor to analyze lists created by human generated randomness vs computer generated randomness with the same p, and they will most likely be able to pick out the human generated list. Basically because humans avoid too much streakiness in probability, and when we see it in a computer generated event like this game, alarm bells go off, even if there isn't anything nefarious occurring.
. Like students populating an empty classroom. The 2nd person won't sit right next to the first person, they will keep partitioning the remaining space pretty evenly.
I'm that guy that will sit right next to you. Huehue.
The problem with computer random number generation, if it truly is just Random.next(), is that computer random number generation isn't really random.
That's why it's technically called PRNG, pseudo RNG. There are many ways to simulate it, a common assignment in a CS course is to develop a RNG. The most commonly used method in distributed libraries is to use a very large randomly created static list of numbers with the entry point set by a seed. Most probably use the current time as the seed. I'm sure this is the method they employ because you can retrace GW battles after retreating. Reseeding with the same seed will result in the same list of numbers.
You can have true RNG in an app, but usually involves paying an outside company to be the source of the entropy (randomness) which uses truly random methods to obtain a numerical value.
What I find curious is that the fights seem to be kinda predetermined. If you retreat and repeat the same fight and do the same moves as last time, the fight goes exactly the same. Everything that is supposed to be RNG and every enemy move will happen exactly the same every single time as long as you have done the same moves before.
The problem with computer random number generation, if it truly is just Random.next(), is that computer random number generation isn't really random.
That's why it's technically called PRNG, pseudo RNG. There are many ways to simulate it, a common assignment in a CS course is to develop a RNG. The most commonly used method in distributed libraries is to use a very large randomly created static list of numbers with the entry point set by a seed. Most probably use the current time as the seed. I'm sure this is the method they employ because you can retrace GW battles after retreating. Reseeding with the same seed will result in the same list of numbers.
You can have true RNG in an app, but usually involves paying an outside company to be the source of the entropy (randomness) which uses truly random methods to obtain a numerical value.
GW was supposedly purposely created this way. It's a way for players to retreat after they lose and learn thier mistakes.
I 100% believe this game is streaky. It's like they because it's designed to be. People are more likely to use crystals to sim more when they get a great sim (they feel "hot" and want to do it again) or when they have a horrible sim (they get frustrated and want to do it again to compensate for what they didn't get).
They base many aspects of the game on the trillion dollar gambling industry. There's laws and regulations though with slots. Not saying even if they add variance it's bad, just that it's a rather naive stance to think that any company wouldn't do certain things when metrics show it's more profitable.
Humans, especially untrained humans, are notoriously bad at statistics. Even a lot of scientists are bad at interpreting data and are prone to confirmation bias and a whole array of other logical fallacies.
In other words, we really just want to know what the odds are and how randomness is simulated in the game. That is it. It would make things a lot easier to understand for players, and there would be a lot less of these threads.
I agree with this. Odds should be required to be posted.
Eventually this will get taken to court.. maybe not with this game. It's the wild wild west right now in the "microtransaction" realm. All sorts of tactics that are unethical at best and criminal at worst. Most of us turn a blind eye and just want to play the game at minimal levels of cash contribution. There are people out there that will latch on to this stuff and go after them though... likely rightfully so.
What I find curious is that the fights seem to be kinda predetermined. If you retreat and repeat the same fight and do the same moves as last time, the fight goes exactly the same. Everything that is supposed to be RNG and every enemy move will happen exactly the same every single time as long as you have done the same moves before.
This actually is the case for Galactic War, though not for anywhere else. (If you'd like to test it yourself, next time your guild does a raid you can retreat and retry multiple times and you'll get different results each time)
I think it's meant as a feature to prevent people from simply retreating if a GW battle doesn't go their way and try again and again and again. Ironically, it encourages that behavior because you can retreat, try again, make one change and see what happens. As a heads up, next time you do GW and retreat, on the retry get to wherever things went wrong and quickly double-tap Autobattle. This "eats" one of the numbers in the RNG seed and moves onto the next one (or there are separate manual and auto seeds, no one's quite sure exactly what happens) which will change the results. If my Poe whiffs and exposes no one, I retreat, retry, and auto his taunt and usually get a different result.
In other words, we really just want to know what the odds are and how randomness is simulated in the game. That is it. It would make things a lot easier to understand for players, and there would be a lot less of these threads.
I agree with this. Odds should be required to be posted.
Eventually this will get taken to court.. maybe not with this game. It's the wild wild west right now in the "microtransaction" realm. All sorts of tactics that are unethical at best and criminal at worst. Most of us turn a blind eye and just want to play the game at minimal levels of cash contribution. There are people out there that will latch on to this stuff and go after them though... likely rightfully so.
U gave me an opening I gots to take it" never tell me the odds" ha ha r.i.p solo
Humans, especially untrained humans, are notoriously bad at statistics. Even a lot of scientists are bad at interpreting data and are prone to confirmation bias and a whole array of other logical fallacies.
So Bob's observation is incorrect because people are often incorrect?
As a heads up, next time you do GW and retreat, on the retry get to wherever things went wrong and quickly double-tap Autobattle. This "eats" one of the numbers in the RNG seed and moves onto the next one (or there are separate manual and auto seeds, no one's quite sure exactly what happens) which will change the results.
My guess is that it's the target acquisition that consumes a number. The seed is toon related so another way you can change the results while keeping the same team members is to change the ability you use, like using one of the toon's basic instead of the special since they consume different numbers of random numbers. It won't work obviously in your Poe's first move case of course but say in a few moves you really need to have R2 stun CLS or something.
Replies
The other problem, is I understand Gambler's Fallacy; I even read a few Wikipedia articles on it.
During the Zaul meta, I specifically counted the times that Zaul dodged versus the times he got hit. The stated probably would expect a dodge about 47% of the time (2% base, 20% zeta leader, 25% if they took out a member of your team). I measured this during each battle and Zaul would dodge me 6 or 7 out of every 8 times. It was fairly consistent, even. Do I have enough data for a statistically significant analysis? I don't know. But not being able to look at their code, I have enough data to assume and be annoyed at CG for adding extra dodge chance to the defense during arena battles (possibly all defense).
Is this accurate? Still can't be said without being able to dig into their code. That's the fun thing about statistical analysis in a game. You'll never be able to really tell just by analyzing the results alone. You have to look at the code to know for sure. I have seen enough, however, non-biased data to make me annoyed at CG for that mechanic (non-biased = I recorded each dodge and hit for quite some time; I didn't just ignore hits).
But am I accurate in my assumption? I don't know. Could just be a bug; could just be annoyingly bad computer psuedo-RNG. As a programmer myself, I've dealt with computer randomness to know that it could both not be their fault, and be totally controllable/fixable on their end.
That's not a problem because the distribution of pseudo random numbers generated by modern software is by all means enough random for applications like this one. Moreover since the random numbers are server generated there are even more randomness factors.
If you notice any "anomaly" it's not because pseudo random generators are used, it's because yoru concept of "randomness" is fallacious or because there are bugs in the code (can't really dismiss the second one in swgoh)
That's very true, and the distribution of the random numbers and its general streakiness depends on the method used. I certainly hope that's the case, since if you search for "Activision patent" there is definitely some shady stuff going on in the world of microtransactions right now. I'm giving EA/CG the benefit of the doubt for now, since I've played for nearly 2 years now and I've gone through my share of bad and good streaks in all aspects of the game.
I 100% believe this game is streaky. It's like they because it's designed to be. People are more likely to use crystals to sim more when they get a great sim (they feel "hot" and want to do it again) or when they have a horrible sim (they get frustrated and want to do it again to compensate for what they didn't get).
I tracked my drop rate a long time ago recording the number of cantina shard drops when simming 12 at a time. If it was truly random I would have seen a Bell shaped curve peaking somewhere around 4. The curve was much too flat and didn't resemble the curve it should at all.
There are just so many extremes in this game. Raid gear drops is another. Every single person I talk to has 1 specific raid gear that drops all the time. I'm positive I get furnaces over half of the time (I feel like it might be as high as 75%, but that's likely just my bias). For as long as I've been playing I shouldn't be getting a drop rate double what RNG tells me to expect.
Maul doesn't counterattack, Dooku *does* have a 100% counter rate, you can see it in his Unique ability. Biggs always calls Wedge to assist. These are all stated abilities that work for players and AI. Luminara most definitely does not have a 100% evasion, otherwise she'd be on every team so you could win on the defense.
What's the matter with you? Why are you like this?
It wouldn't be bell-shaped. The expected value is to the left of the center, on the 0 side. The lowrr values would bunch up and the distribution would be skewed. Would probably look closer to a Chi-Square distribution.
It is nearly impossible for any one person to figure out the exact odds and randomness of the game because their sample is way too small. It is also a bad idea to try and use aggregate data via each player self-reporting their results because that opens the door to all sorts of bias that you cannot account for. The only real way to figure out the odds of different things and how randomness is simulated would be data-mining (*hint, hint* people who know how to do that).
This brings me to the real issue here. The issue is not that we don't like the current odds (although that may be a separate issue, it is not the issue here), don't like the method of RNG used, or that we don't like streaks. The issue is that we don't know any of those things. In other words, we really just want to know what the odds are and how randomness is simulated in the game. That is it. It would make things a lot easier to understand for players, and there would be a lot less of these threads.
on topic.
RGN is RGN. Its not SIYF ( sometimes in your favour )
I am having a tragic run with zeta drops on my main account. ive had 4 zeta mats drop in 13 weeks so... if ships are dodging your attacks lots of times , you are likley getting a greater zeta drop then i am ( eveeyone is getting a better zeta drop then i am )
Agreed. Forecasting odds for swgoh is just fallacy.
Though, i would add, we want to know whether we're personal doing better then the next guy/girl. Knowing someone is getting it a lot worst puts things into perspective - unless you ARE that bottom person.
I don't think anyone would think a human could realistically simulate randomness, but that doesn't mean humans can't (sometimes) detect bias when evaluating data.
The obvious bias when people pick numbers "randomly" is that if they choose a low number between 1-100, their next number will usually be something that divides the largest partition. Like students populating an empty classroom. The 2nd person won't sit right next to the first person, they will keep partitioning the remaining space pretty evenly.
I'm that guy that will sit right next to you. Huehue.
That's why it's technically called PRNG, pseudo RNG. There are many ways to simulate it, a common assignment in a CS course is to develop a RNG. The most commonly used method in distributed libraries is to use a very large randomly created static list of numbers with the entry point set by a seed. Most probably use the current time as the seed. I'm sure this is the method they employ because you can retrace GW battles after retreating. Reseeding with the same seed will result in the same list of numbers.
You can have true RNG in an app, but usually involves paying an outside company to be the source of the entropy (randomness) which uses truly random methods to obtain a numerical value.
GW was supposedly purposely created this way. It's a way for players to retreat after they lose and learn thier mistakes.
They base many aspects of the game on the trillion dollar gambling industry. There's laws and regulations though with slots. Not saying even if they add variance it's bad, just that it's a rather naive stance to think that any company wouldn't do certain things when metrics show it's more profitable.
Hahaha
I agree with this. Odds should be required to be posted.
Eventually this will get taken to court.. maybe not with this game. It's the wild wild west right now in the "microtransaction" realm. All sorts of tactics that are unethical at best and criminal at worst. Most of us turn a blind eye and just want to play the game at minimal levels of cash contribution. There are people out there that will latch on to this stuff and go after them though... likely rightfully so.
I think it's meant as a feature to prevent people from simply retreating if a GW battle doesn't go their way and try again and again and again. Ironically, it encourages that behavior because you can retreat, try again, make one change and see what happens. As a heads up, next time you do GW and retreat, on the retry get to wherever things went wrong and quickly double-tap Autobattle. This "eats" one of the numbers in the RNG seed and moves onto the next one (or there are separate manual and auto seeds, no one's quite sure exactly what happens) which will change the results. If my Poe whiffs and exposes no one, I retreat, retry, and auto his taunt and usually get a different result.
U gave me an opening I gots to take it" never tell me the odds" ha ha r.i.p solo
So Bob's observation is incorrect because people are often incorrect?
My guess is that it's the target acquisition that consumes a number. The seed is toon related so another way you can change the results while keeping the same team members is to change the ability you use, like using one of the toon's basic instead of the special since they consume different numbers of random numbers. It won't work obviously in your Poe's first move case of course but say in a few moves you really need to have R2 stun CLS or something.