"
anubite wrote:
As a business person, you might know, that it's better to piss off half the community and to gain the favor of the other half, than to try and "please them all". This is what POE is designed to do - it is a niche market. It captures the niche market. And when I see someone like Kripp play (I don't necessarily like kripp, but he's a good example because he's popular), I see his face light up when he gets a string of good luck in race! And when he gets a string of bad luck? Well, nobody likes bad luck, and Kripp will whine he was cursed, but it's bad luck that defines good luck; once you've had some bad luck sprees, good luck sprees become fun. They become exciting. And they give unskilled players ways to compete and gain entry into the market.
When a new player picks up a mirror in twilight strand at level 1, and he's instantly very rich, POE has probably secured a potential customer. Altering RNG potentially could destroy that phenomenon, by altering the value of certain items or preventing the circumstance like that, at all.
If I were designing a game to make money, I would make a good game first, and a streamlined skinner box second.
As a business person, you know its not going to be half half, its going to be closer to 5%, 95%
The population of the hardcore games is an extreme minority
|
Posted bydeteego#6606on Feb 26, 2013, 1:51:39 AM
|
"
Not a good idea for a variety of reasons. Perhaps the simplest is that the more significant the perturbance is, the more it becomes a disincentive. Meaning the larger the number of things found by the increased factor "N+" the less things are found from the normal factor "N-". So once you find a drop that is distributed in this way, that's a pretty good time to call it a night and stop because the odds that you will get anything else are very poor until you accumulate enough failures that you've upgraded back up to the N+ probability.
I can see what you are saying there.
What you appear to be saying is that if an event is 1:1000, but we have decided that the p drop should approach 1:1 at count 2000, that if someone has a bad run of luck and the regulator kicks in at 2000, and a nice item drops (unique, map, exalted, whatever), that when it does, they will finally decide to go to bed.
Have you considered, perhaps, that maybe they should? :-P
So, I am kidding around a bit, but have you considered it from the opposite perspective. Not when they quit, but rather that it keeps them playing?
Yeah. So I am totally on board with what you are saying, and agree with every word, except for the opening sentence "not a good idea."
*grin*
|
Posted byCourageous#0687on Feb 26, 2013, 6:52:10 AM
|
"
Khazrad wrote:
I'm sorry to say, but anubite, I think you haven't grasped at all what kind of system the OP is proposing and the foundations of maths behind the 'randomness' of strings of independent events.
Ah sorry, you're completely right. I'll go back to practicing my multiplication tables.
My Keystone Ideas: http://www.pathofexile.com/forum/view-thread/744282
|
Posted byanubite#0701on Feb 26, 2013, 11:10:15 AMAlpha Member
|
"
anubite wrote:
"
Khazrad wrote:
I'm sorry to say, but anubite, I think you haven't grasped at all what kind of system the OP is proposing and the foundations of maths behind the 'randomness' of strings of independent events.
Ah sorry, you're completely right. I'll go back to practicing my multiplication tables.
Oh, let me take a quote from you that illustrates what I mean:
"
So far, everyone including me has proposed solutions to the proposed problems of "randomness not being fun" but we have failed to justify why we deserve 5/6 linked items, why we deserve to easily get the right color sockets on our gear, why we deserve more uniques and more/better drops. I don't think we can justify that. It's up to GGG to determine what they think is fair because we as players will always want more than that.
The whole topic isn't about more, it's about more even distributed. That means cutting the probabilities at the 'very lucky' end (fuse once, get 6l) and the 'very unlucky' end (fuse 1000 times, don't get a 5l) and move the distribution slightly towards the expected outcome. The overall chance of getting something is the same. The point is, and I agree with the OP in that, that the (simple to implement and light in terms of resources) system in place is not a good one., I suggest that if you lower the base chance but implement a factor based on the context of the event, you'd get better results.
My main reasoning is that the number of events in this game are very low compared to the chances which leads to a huge standard deviation. You have written that over time, things will even out, and while that is true for 'very long' periods of time (infinity) I doubt that it is true for average game time to get anywhere near the expected value for rare events. Particularly Map drops would benefit big time from normalization because in that system, streaks in both directions amplify the chance of continuing the streak which (as seen in the big thread somewhere) makes some people very happy and others very mad (both rightfully so, I guess).
|
Posted byKhazrad#5396on Feb 27, 2013, 7:37:45 AM
|
|
to the Original Poster: there is no need to change probability.
if all drop chances are truly equal, the Law Of Large Numbers (LLN) theorem ensures you would get the expected value, given a large enough amount of attempts K.
K being very large if you are on a real bad-luck streak like I am.
the only questions are, does the Expected Value of the current Random Number Generator go near success instead of near failure? and are the chances truly equal with each "roll", and unaffected by non-relevant things like character class?
if the answer to any of the above two questions is "no" - then the RNG's odds are bad and need to be changed.
the LLN theorem applies even in Casinos. although Casino owners use other (possibly dubious) methods to ensure the house always wins.
for example: lose $1 20 times, then win 15$ in the 21st roll = $5 profit to the house.
expected value here is somewhere around $15. expected value for a fair Path Of Exile RNG should be around "$20".
Alva: I'm sweating like a hog in heat
Shadow: That was fun Last edited by johnKeys#6083 on Feb 27, 2013, 8:03:32 AM
|
Posted byjohnKeys#6083on Feb 27, 2013, 7:56:59 AM
|
|
Hi, John, you have characterized my OP correctly. A subtle nuance was that I emphasized smoothing values and not peaks; hence my comment that such a system could in effect increase positive outcomes slightly. And hence my comment about not just one bad outcome, but also history, as this latter alteration will amortize to zero over time. The historical system is best at nullifying bad first impressions, such as the poor chap in the other thread who has used chance orbs 37 times in a row, with no rares. Using a historical system, after a few interventions, it would be unlikely there would be any more, because after there are enough samples, a user will be balanced.
|
Posted byCourageous#0687on Feb 27, 2013, 9:36:06 AM
|
"
Courageous wrote:
"
Not a good idea for a variety of reasons. Perhaps the simplest is that the more significant the perturbance is, the more it becomes a disincentive. Meaning the larger the number of things found by the increased factor "N+" the less things are found from the normal factor "N-". So once you find a drop that is distributed in this way, that's a pretty good time to call it a night and stop because the odds that you will get anything else are very poor until you accumulate enough failures that you've upgraded back up to the N+ probability.
I can see what you are saying there.
What you appear to be saying is that if an event is 1:1000, but we have decided that the p drop should approach 1:1 at count 2000, that if someone has a bad run of luck and the regulator kicks in at 2000, and a nice item drops (unique, map, exalted, whatever), that when it does, they will finally decide to go to bed.
Have you considered, perhaps, that maybe they should? :-P
So, I am kidding around a bit, but have you considered it from the opposite perspective. Not when they quit, but rather that it keeps them playing?
Yeah. So I am totally on board with what you are saying, and agree with every word, except for the opening sentence "not a good idea."
*grin*
Haha, great response!
|
Posted byPolarisOrbit#5098on Feb 27, 2013, 9:39:28 AM
|
"
Courageous wrote:
Hi, John, you have characterized my OP correctly. A subtle nuance was that I emphasized smoothing values and not peaks; hence my comment that such a system could in effect increase positive outcomes slightly. And hence my comment about not just one bad outcome, but also history, as this latter alteration will amortize to zero over time. The historical system is best at nullifying bad first impressions, such as the poor chap in the other thread who has used chance orbs 37 times in a row, with no rares. Using a historical system, after a few interventions, it would be unlikely there would be any more, because after there are enough samples, a user will be balanced.
that poor chap who failed 37 times with the Orb Of Chance is me :(
but my point is, I could have gotten a unique item if I tried 40 times. or 50 times. maybe. assuming RNG is "fair" (in terms of expected value) and with equal odds on every roll. that's what LLN theorem says.
I just rage-quitted using Orbs Of Chance entirely, so I'll never know, but if fairness and equality of odds holds true - current RNG needs no change. if not, it needs a complete overhaul with entirely new parameters.
besides, implementing a system such as the one you suggest would be next to impossible, since the code would have to keep track of every character's every item drop, and transmutation/enchantment attempt.
a big, messy data structure for each character x tenths of thousands of characters. lots of both runtime and space complexity.
Alva: I'm sweating like a hog in heat
Shadow: That was fun
|
Posted byjohnKeys#6083on Feb 27, 2013, 10:22:44 AM
|
|
Not impossible at all, no. It does require some accounting for each individual thing you want to regulate. Each would be decided case by case. For uniques, you could keep a record on the char. For fusings, on the item. And so forth.
|
Posted byCourageous#0687on Feb 27, 2013, 10:53:03 AM
|
"
Khazrad wrote:
"
anubite wrote:
"
Khazrad wrote:
I'm sorry to say, but anubite, I think you haven't grasped at all what kind of system the OP is proposing and the foundations of maths behind the 'randomness' of strings of independent events.
Ah sorry, you're completely right. I'll go back to practicing my multiplication tables.
Oh, let me take a quote from you that illustrates what I mean:
"
So far, everyone including me has proposed solutions to the proposed problems of "randomness not being fun" but we have failed to justify why we deserve 5/6 linked items, why we deserve to easily get the right color sockets on our gear, why we deserve more uniques and more/better drops. I don't think we can justify that. It's up to GGG to determine what they think is fair because we as players will always want more than that.
The whole topic isn't about more, it's about more even distributed. That means cutting the probabilities at the 'very lucky' end (fuse once, get 6l) and the 'very unlucky' end (fuse 1000 times, don't get a 5l) and move the distribution slightly towards the expected outcome. The overall chance of getting something is the same. The point is, and I agree with the OP in that, that the (simple to implement and light in terms of resources) system in place is not a good one., I suggest that if you lower the base chance but implement a factor based on the context of the event, you'd get better results.
My main reasoning is that the number of events in this game are very low compared to the chances which leads to a huge standard deviation. You have written that over time, things will even out, and while that is true for 'very long' periods of time (infinity) I doubt that it is true for average game time to get anywhere near the expected value for rare events. Particularly Map drops would benefit big time from normalization because in that system, streaks in both directions amplify the chance of continuing the streak which (as seen in the big thread somewhere) makes some people very happy and others very mad (both rightfully so, I guess).
Well, you've confused my points. I was talking about more in the context of buffing drops, which is the more natural thing to do instead of completely changing how 'random' is calculated. The desire to change how 'random' is calcualted stems from a desire to have more loot, not less. If you were getting tons of loot, you wouldn't ask them to change the way their rng works.
I understand perfectly, I think, what the OP desires and describes as a solution to his proposed problem.
Over infinite periods of time per user, things average out, but it's more than that. If we have 500k people playing POE, that's a large enough population that on average, drops are random. People get random amounts of wealth that average out to be fair. There are people on the low end and people on the high end. And provided we ignore certain other factors (wealth is a question of exponential growth and playtime; the rich get richer), over an infinite period of time, people get 'fair' amounts of wealth.
My Keystone Ideas: http://www.pathofexile.com/forum/view-thread/744282
|
Posted byanubite#0701on Feb 27, 2013, 11:05:25 AMAlpha Member
|
