Showing Posts For Antonio Cappello.1806:
I’m doing several times each weeks zerg world boss, for example Melandru (100+ players). On champions, if you’re not grouped, with a zerk warrior, hitting the champion as much as you can with your best skills, you have 0 loot (no chests, nothing at all).
If you are in a group, you always get your loot (champion’s chest).
If you want to toss around anecdotal evidence like it means something, I’ll do the same and point out that I run champion trains solo quite often, and I’ve never missed out on loot provided I managed to tag the mob. Sure, tagging requires doing enough damage to remove some percentage of the mob’s health, but once tagged, loot no longer depends on DPS.
I have never experienced this issue you speak of; nor has anyone else I know. Without some real non-anecdotal evidence supporting this particular claim, I see no reason to accept your anecdote over mine.
EDIT: Also, in the long term this can certainly be RNG, and it probably is. Not many people win the lottery either, but one could argue that the ones that do are exceptionally lucky. Some people even win more than once. That doesn’t make the lottery any less random.
technically, they do it indirectly – by helping with tagging. Now, once you tagged the mob, however, the drop chances are exactly the same and dps or group stops having any effect at all.
Ah, true. As you say though, once the mob is tagged, it’s all the same. I’m sure there’s some optimal number of participants and dps that maximizes number of mobs successfully tagged with minimum effort, but as you say, that’s not really the RNG system that’s causing the issue, per se.
Edit: come to think, your mathematician friend might have been saying that for a pseudorandom generator to produce true random numbers, the seed would have to be infinite. This is actually correct – still doesn’t change anything though.
If that’s what his mathematician friend was getting at, that would indeed make more sense. I think it would be great if probabilities and some of the basic concepts of random numbers could be treated more often in secondary school education. Where I’m from it’s not common, but it’s such a useful topic.
I’ve talked about this with a phd mathematician, in order for a number to be truly random the upper bound would have to be infinity and so the average size of a truly random number is one half infinity and so no computer system will ever be able to use truly random numbers.
FWIW this guy did work on RNG algorithms for industry and he said they don’t really want something approximating real randomness because they don’t like long stretches of unlikely events. He said what industry is in the market for is an algorithm with few long stretches of repeating results but overall is close to random and that’s what he had to program to please the customer.
The argument being made by that poster was that mathematically there are no truly random numbers. That’s wrong, however. What you’re speaking of is a technical restriction, which is a different ballgame altogether.
Yes, for true randomness a random real variable must have infinite range. This is obviously not achievable from a technical standpoint. Then again, the principle of RNG isn’t really realizable on a classical computer anyways, since it relies on a seed (pseudo-random instead of random). The point also still remains that for all intents and purposes, and over any appreciable time scales, a good pseudo-RNG system is statistically very close to a true RNG system.
(edited by Antonio Cappello.1806)
There were multiple threads on this evidence less than a year ago, just because people don’t like that evidence because it clearly shows that there is still to this day an issue doesn’t mean it’s anecdotal nor does it dismiss this issue just because people think that it’s proper to restrict items in this fashion. This game as well as all others that have practices the systematic restriction on rewards have had the problem of continuous player loss, don’t think the surge of players coming back during this particular patch will in any way be lasting if the rewards system isn’t addressed properly in the open world. There’s also been evidence of drops being particularly imbalanced when players go to the WvW areas compared to the PVE open world, as well as dungeons compared to PVE open world and yes part of it is to make more money, however, most games that have done this don’t last very long.
There’s also the factor of DR which you ignored in your response, that also factors into the equation and DPS based rewards that most games have used early on but then the developers realized how restrictive these processes are and changed for the better. WoW is a great example of this changing over to a system of currency for rewards completely whereas even for crafting materials in this title one can use currency from completing dailies and monthlies however, even those rewards are still RNG instead of being able to transfer funds by buying these items directly. (the RNG bags from laurels for example)
I would like to see links to said evidence. The few sources I’ve seen (including drop rate investigations on the wiki) seem to suggest that drop rates are… Well, what they are. I’m not sure how one could find any evidence that RNG isn’t working as intended, given that ArenaNet isn’t exactly providing the exact drop rates for their items.
I’m also dubious about your claims that DPS or group size contribute to drop rates. Again, I would like to see a set of controlled trials (at least a few thousand to ensure statistical accuracy) that could support this. Before you suggest I go looking these threads up myself, I’ll point out that the burden of proof is on you; you’re making the claims, so you need to provide the evidence.
I didn’t comment on the DR because I have no position on that, and it’s not entirely relevant to the question at hand, which is (loosely speaking) whether MF works as advertised. DR sits as a blanket on top of all systems, including MF.
EDIT: The purpose of my selective response to your comments was not to cherry-pick ideas, but to point out that you made an error in your claim that pure randomness doesn’t exist. I don’t really have a position on whether the loot system that currently exists is good or bad, but I have an interest in ensuring that people have the correct understanding of things like random processes.
(edited by Antonio Cappello.1806)
“It’s RNG deal with it” is no longer a valid argument. It’s well known that mathematically speaking complete and utter randomness is impossible, it is also a well known fact that the typical algorithms used by all mmo companies are skewed. So that being said let’s look at the problem.
Mathematically speaking, complete and utter randomness is quite possible, so I’m not sure what you’re talking about there. Take any example of a noise process with a Dirac delta correlation function and you have a noise process that isn’t predictable. Now, the spectral density of the noise function may not be “random”, but that doesn’t mean the process isn’t random. “Random” doesn’t mean “something for which you can’t construct a distribution function.”
As for the rest of your comments, I don’t think it’s of much use speculating on it without hard evidence. Anecdotal evidence is, for many reasons, not terribly useful in determining whether RNG works. Sure, MMOs (and computers in general) use pseudo-RNG systems, but I would be very impressed indeed if you were capable of completing actions in game such that you could actually see the difference. The time scales we’re talking about are miniscule, here.
Give someone item Y after X amount of hours guaranteed and it’s called grinding. Make an RNG based system and people cry out that certain people are tagged for luck. They are tagged for luck, just like the guy that wins the lottery. It’s all percentages and probabilities.
Pro Tip: Best way to get stuff in games is still do what you love to do, save up the money and then buy the thing you want.
Indeed. I don’t think there’s any real winning when it comes to designing loot systems. Personally I follow your “pro tip” while playing as well. I set some long-term goals, figure out the parts I can reasonably achieve without wasting my life away, and then work towards them by doing things I find fun in game. Along the way, I inevitably pick up items or materials that I can use to further the seemingly less achievable parts of my long-term goals, and everything seems to work out just fine.
That’s a good point. There was someone earlier in this thread who literally suggested there was a conspiracy to give some players more luck than others. I’d imagine that when this person examines their drop rates, they are going to be interpreting them in context of the supposed conspiracy, which will make them see a pattern that supports the conspiracy, which will give them a stronger incentive to interpret future drops in that context, and so on. I doubt that creating a “random pattern” RNG will stop people from doing this. It’s just such a strong part of our nature to be awful at probability.
My thoughts exactly. Making changes to make the RNG “appear” more random is ultimately arbitrary. Any system that doesn’t guarantee Drop X after at most Y hours is going to be subject to this very same issue. That’s why I don’t think there’s really a viable solution within the RNG system itself.
Note: That’s not to say viable solutions don’t exist. Every method that can be used to obtain an item through a non-RNG source helps reduce that feeling (e.g. if I don’t feel like farming for Item X I can just go get it by doing Event Y). Within the RNG system itself though, I feel it’s not really feasible.
That’s probably what I’m suggesting, yeah. The system I describe is likely weighted so as to limit the possibility of noticeable streaks. But, as you say, you lose statistical randomness in the process.
I’m not sure how one would implement that either, especially considering players’ goals can vary significantly. I know some people who couldn’t care less about getting exotic drops, but if getting that last green means unlocking a skin they don’t yet have, that would be much more important.
I also believe that with things like confirmation bias, this would just shift the problem elsewhere. A “bad luck” streak of three or four poor rolls can bias someone’s opinion enough to make them forget, for example, their previous string of rares a week ago.
Yes, but this isn’t about statistics so much as it is about perception. Since RNG can’t exactly be random, it has to be built in a way so as to appear as close to random as possible (most RNG accomplishes this stunningly convincingly).
I’m not sure how you could do this. Whenever there is a statistical probability of multiple events occurring, RNG will always produce streaks. Even if all drop rates were normalized, people looking for Specific Item X will still be disappointed and feel themselves unlucky when they don’t get it.
GW2 has broader and more general loot tables, so the chance of getting a specific item is much lower than it would be doing specific activities in some games (where, for example, a boss drops an item out of a list of ten). That’s not the same as saying the RNG feels more random, however, and it doesn’t make streaks any more or less likely. In fact, killing the same boss 20 times and not getting that single item with a 10% drop rate can feel more unlucky than GW2’s precursor system, where at least everyone knows the drop rate is incredibly low.
I’m simply suggesting that the appearance of randomness is important to successful RNG, and that I define that success as neither feeling lucky or unlucky—it should simply feel random.
But this is part of the problem. Confirmation bias will make it impossible for a system to truly “feel” random to people, because humans just have a terrible ability to grasp what “random” is. There’s no such thing as a system that doesn’t have lucky streaks.
As an anecdotal example, I was running a dungeon with some PuGs the other day. I got an exotic and a few rares out of the deal, and one of the PuGs did not. This, to him/her, was a shining example of why the system isn’t random. If a player can be so woefully incorrect after a single short term trend, how do you design a system that feels random to such a person?
10 heads in a row is unlikely (1/1024). Successful RNG should present at least the appearance of randomness; you should neither feel lucky or unlucky.
Which is the exact same probability as any other sequence of ten coin flips (see Alex’s point above). This is how a series of ten coin flips works. It is indeed random, and any exact sequence of ten flips is just as likely as any other.
Noticeable streaks are a sign of bad RNG.
That’s not entirely correct. Noticeable streaks at the statistical level are a sign that there’s some trend that isn’t random. Noticeable streaks on smaller scales happen all the time and are a natural result of RNG. If you flip a coin ten times and get ten heads, that doesn’t mean the process isn’t random or that it’s biased.
Given that they aren’t changing anything else about unidentified dyes (still tradeable, etc.), I imagine they will remain on the laurel vendor. The post mentioned that drops were stopping, which (to me at least) implies that otherwise the status quo will remain.
I imagine you’re right, at least for the time being. It will be interesting to see what happens to the value of the dyes after the patch starts to settle. I wouldn’t be surprised to see speculators hoarding unidentified dyes to try to make a profit come patch time, which could artificially lower prices for a time. We’ll see I suppose.
Of course demand has increased; they announced they were removing dye drops. I believe the demand for dyes would, in fact, decrease if they didn’t make that move. As it is, you’re comparing two entirely different situations.
The post also didn’t imply that they were removing dyes from other sources like laurel vendors, so I don’t think it’s going to be so bad.
I’m just going to point out that electric field lines are simply a pretty visual representation of the actual physics going on. In reality, electric fields are just solutions to Maxwell’s equations, which in the absence of moving charges happen to be solutions to Poisson’s equation. In the absence of charge density, these solutions further simplify to become solutions to Laplace’s equation. All solutions to Laplace’s equation share the same property: they do not allow discontinuities. As such, to get a discontinuity, there has to be a charge density present. In the pretty picture of field lines, if we assume only one electric field is present, field lines “crossing” represent a discontinuity.
So to sum up, “this is probably like electric field lines” is a weird statement that doesn’t actually mean anything useful in this context. I will further add that if you superimposed two electric fields, you could indeed draw field lines that crossed. We don’t do this simply because it’s easier to consider the net field, and draw the lines accordingly (the net field is more interesting to us).
Anyways, now that the physics digression is over, I’m interested to see why Scarlet is gunning for the leylines and what she’s hoping to do with them. I hope this isn’t anything as “simple” as releasing an Elder Dragon. I share the opinion that somebody probably should have noticed if there was that much magic pouring out of the environs of Lion’s Arch, and I also do believe that there is a marked difference between the Artesian Wells and the two leyline crossings we know about. I would be a bit surprised if an Elder Dragon popped out of that hole.
As someone who has been compromised before in another game, I can say that if things happened as you say, it’s very likely that your email has indeed been compromised by some method. There are many attack vectors that can be used to strike a computer, anti-virus/firewall or not, including automatically executing trojans that latch themselves into advertisements on other websites.
There are plenty of reasons why your other accounts may not have been touched. The bank account is an obvious one; most banks would still require your PIN or something of the like in order for you to log in. As for the rest, perhaps the hacker didn’t know about them, or targeted GW2 specifically.
More importantly though, if you haven’t done so yet, you really need to ensure that your email (and computer, of course) is secure. That’s the only real way they could have intercepted the password change request and, perhaps more strikingly, the authentication emails you receive when you try to log in from an unrecognized location.
EDIT: As an addendum, you should strongly consider getting a smartphone; not just for GW2, but for other applications that allow for two-factor authentication. Remember that the whole point of two-factor authentication is to split authentication into two different categories: something you know, and something you own. It’s the closest you can get to being “safe” while connected to the Internet.
(edited by Antonio Cappello.1806)
You’ll want to log in using the email address you registered the game with. That’s what’s referred to as your account name, not your display name (different things).
Ah, I had assumed the poster was the friend who can’t play, simply posting using the other account. After re-reading the original post, I can see that it’s not really clear who is actually sitting at the keyboard typing. :P
If it’s your account in question, you’re the one who has to put your details into the support request. Don’t use your friend’s details for that, even though he was the one who posted on the forums for you (I think Inculpatus thought you meant your friend needed help).
… and in most fictional works, anti-matter is meant to be the opposite of matter (but still physical) that would result in mutual destruction if coming in contact with matter.
I feel the need to point out that this is somewhat close to the definition of anti-matter in real world physics, so it’s not terribly surprising that many fictional works use it as such. Wikipedia’s article on anti-matter is half-decent.
I know that Blizzard hasn’t always offered character restorations so liberally, and for about half of the time I played WoW (from launch until the end of the second expansion), this feature was very limited. The customer service forums used to be rife with complaints about restorations not occurring.
I wouldn’t be surprised if the problem here was related to the gem-to-gold conversion (and vice versa). That, plus a few other things that I won’t mention (for fear that scammers could try to use them) make it fairly easy in my mind to use a restoration system to duplicate items. To counter some of these methods, ArenaNet would have to track the history of a deleted character in great detail to root out any potential wrongdoing. Perhaps WoW suffered from the same issue, but Blizzard just didn’t care as much because of the inability to legitimately buy gold with real money.
TL;DR version: I doubt it’s anywhere near as simple as ArenaNet just rolling back the character if they’re also looking to prevent abuse and protect the economy. Sure, all it takes is some employee scanning through logs and such, but time isn’t free.
probability only works if you know the exact chance, a 1/4 chance is to rough a number.
if, for instance, you get 1 in 5 runs and then 2 in 5 runs then the 2/5 or 1/5 is way to rough to measure. (that would be 1/30 btw)
Well first of all, that’s not true. The problem is that you’re using sample sizes that are far too small to draw any kind of statistical data. Also, as I mentioned in my post above, using the Bayesian update rule, one can generate reliable statistical data without knowing the “exact chance”. One only needs to be in the right ballpark. In this case, that wouldn’t be hard to accomplish, since with four fractals in that tier, the probabilities are not likely to be terribly complicated, and we can safely assume that if they’re not exactly 1/4, they’re close.
… you need to observe events on a large scale thus connecting separate events.
I also feel the need to point out that the individual trials are indeed separate, and looking at the data across multiple trials does not “connect” them.
Consider four coin flips. The probability of getting HHHH is exactly the same as the probability of getting TTTT, or HTHT, or THTH, or TTTH, etc.
Why is it then more probable to get two heads and two tails? Because the number of specific distributions in which there exist two heads and two tails is larger than the number of specific distributions in which there are one, two, or three Hs. The probability of each distribution occurring is still identical.
First of all, simulated randomness =/= true randomness. If there is an algorithm deciding on an order, then, by definition, it’s not random. No matter how random it might appear.
Second, of course they aren’t connected and the dice doesn’t remember the previous roll. However, for the expected result(what probability tells us) to be close to the real outcome, you need to observe events on a large scale thus connecting separate events. That’s why your chance of rolling a 20th tails in a roll isn’t 50%.
The only difference between pseudorandom and random is outlined in the Wikipedia article I linked. Notably, none of the statistical differences between RNG and PRNG would have any bearing on the game.
No, your chance of flipping a 20th tails in a row is indeed 50%. Your chance of flipping 20 tails in a row is not 50%, but once you’ve flipped those 19 tails already, the last flip has a 50% probability of being heads and a 50% probability of being tails. There’s nothing questionable or debatable about that statement in this case, since we’re describing a system without memory of its previous outcomes.
Lastly, a large part of Bayes’ theorem relies on guess work. You assume certain conditions to be true and calculate the probability based on that. That’s far from accurate now, is it? The rest of it relies on extensive testing. Sure, if somebody was to run thousands of fractals and collect the data from those, the resulting distribution would be rather accurate. Except that’s not how people make the game forum probability calculations. They go off of the assumption that the data is pristine and fair. I’m not saying it’s not, I’m saying we don’t know whether it is or isn’t. Which is why those calculations are ultimately pure speculation and waving them around saying “you had enough time this maybe correct maybe very wrong probability calculating post says so” is puppy.
If you believe that a large part of Bayes’ theorem relies on guesswork, you need to research a bit more on the subject, as that is entirely incorrect. Bayesian statistics is a staple in statistical mechanics, and with a good prior, is incredibly accurate, especially when bias is involved. Picking a good prior is vital, but it should not involve guesswork, nor does it need to in this case. We know the fractals are tiered. We can assume that the fractals within each tier are randomly generated, with the caveat that no fractal may appear twice. That’s a good prior to start with. Then we test it and run experiments (running fractals), and update our prior using the Bayesian state update rule. If there’s bias, it will show in our updated prior, and our statistical model will converge with reality. The only way this would not be the case is if our original prior was widely off base (e.g. if ArenaNet imposed a massive bias towards one or two fractals). However, why would they do this? Occam’s Razor suggests that this wouldn’t be the case.
You do realise that true randomness doesn’t exist in programming, right?
Right, so this isn’t true. The most basic form of RNG done using a computer is pseudorandom number generation (see http://en.wikipedia.org/wiki/Pseudorandom_number_generator). The principle of PRNG is that a seed is generated based on some kind of deterministic value (say, the system clock), and that seed is used to generate a field of random numbers. However, if you peruse the Wikipedia article, you’ll find that this makes the numbers effectively random. If, for example, fractal generation is seeded by the server clock (which is likely accurate to the millisecond), to even guarantee two groups got the same seed would require that both groups step into the fractal at exactly the same moment (accurate to the precision of one millisecond). Network latency alone would make this a nigh impossible task.
I’m not sure why you feel there needs to be only one die, but that is provably false. The seed generation may indeed be deterministic, but the rest of the process is not. In a way, the seed determines which die you get.
It’s the best word I know to describe it. That’s how you get higher probability over multiple tries. Otherwise your chance of rolling a 3 would always be 1/6.
Right, except your chance of rolling a 3 is indeed always 1/6. You’re falling into the (quite common) gambler’s fallacy (see http://en.wikipedia.org/wiki/Gambler%27s_fallacy). In the example of a coin flip or die roll, the system has no memory of its previous flips. If you flip 20 heads in a row, the chance of your next flip being a head is still 1/2. The chance of you getting 21 heads in a row is ~9.54 × 10^-7. That certainly seems small! However, the chance of you rolling 20 heads and then a single tail is also ~9.54 × 10^-7.
In fact, if you determine that you’ll be doing a set number of fractals (say, 10 runs), then each time you don’t get the fractal you want, you actually have a smaller probability of getting the right fractal (check the article for the math behind that).
Yeah, I’ve seen those. They tend to also have people who say “it’s all in your head”. Because, again, that’s a shared dice.
That has nothing to do with having a “shared die”. The probability distribution of your rewards (and your fractals) is independent of the next guy’s. If there’s someone out there getting all of the swamp fractals, that doesn’t make it any more or less likely that you’ll see it. You seem to have a misunderstanding of what a probability distribution actually is. Small samples may or may not conform to the distribution, but at statistically large sample sizes, they will. There is also no rule in probability that states that a streak of good luck must be followed by a streak of bad luck or vice versa. Probabilities do not balance out that way, and that’s just more of the gambler’s fallacy.
… snip for the rest …
Sure engineers use probability. That’s basically what “tolerance” is. For every batch of sheet metal produced, there’s an allowable amount that the sheet thickness is allowed to waver, where this conforms to some probability distribution. In a very bad design process, the probability distribution is constant; in other words, each sheet has a completely random chance of being any thickness within the tolerance. Of course, with engineering, we do testing to ensure that our numbers are correct, and we account for those tolerances in design. The same goes with medical issues. Your comment that “we don’t use the probabilities” is somewhat nonsensical, because a probability only tells us the likelihood. If you have a 0.5% risk of developing a particular illness and you’re displaying symptoms that could either be that illness or some other set of diseases, it’s probably better to check for the other more common diseases first.
You may find it interesting to look up Bayes’ theorem (see http://en.wikipedia.org/wiki/Bayes'_theorem). We do not need to know the initial probability of the fractals appearing in order to determine the probability distribution, because the update rule for Bayes’ theorem allows us to build it from the ground up! We can even predict whether there’s bias in the system. We already know such a bias exists, since the fractals are tiered. If one wished to do so, one could run a few thousand fractals to build the probability distribution. Or, by Occam’s Razor, it’s probably fine to note that it’s easier for ArenaNet to program a random distribution within each tier, as opposed to biasing fractals for some arbitrary reason.