| Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Tunak
   |
Posted - 2007.08.30 15:02:00 -
[1]
Assuming a perfect random number generator would the results your are seeing be possible or impossible?
For example: If the random number generator were perfectly random would it be possible or impossible to receive x kill missions in a row from a 95% courier 5% kill agent?
|
Tunak
|
Posted - 2007.08.30 18:32:00 -
[2]
 Originally by: Etho Demerzel
Originally by: Tunak
Assuming a perfect random number generator would the results your are seeing be possible or impossible?
For example: If the random number generator were perfectly random would it be possible or impossible to receive x kill missions in a row from a 95% courier 5% kill agent?
Well, it is the same of wining the lottery (the real one) several times in a row. The chances are so small that you can suppose that the person cheated with certainty.
If the random generator was reasonable, not even perfect, the chances of having the same event repeated more than 4 or 5 times in a row, for a probability of 50% or less is very slim. And it gets so small that you can suppose it won't happen if you go far enough (20 repetitions for example).
What I experienced are several occurrences of these unlikely phenomens. Which definitely shouldn't happen.
It is like throwing dice. Get 4 dice (normal 6 sided ones). Throw them 100 times and check how many times, if any, ALL the 4 will have results lower than 4 (3 or less). You will get around 6. Then try again with 8. You won't get a single one, most likely. I can bet real money with you on this game if you don't believe me. ;)
That's a nice speech but you didn't answer the question.
Is it possible or impossible to receive x kill missions in a row from a 95% courier 5% kill agent?
|
Tunak
|
Posted - 2007.08.31 11:10:00 -
[3]
Originally by: Etho Demerzel
Everything is possible. If the random functions are at least reasonable the chances of these event you described happening are of 0.00003125%. Which are considerably lower than the chances of you getting shot the next time you walk to the street. Or about the same of a person being stroke by a lightning. For it to happen several times to the same person well, how many cases you know of a person striken by multiple lightnings...
Straw Man. It's impossible for there to be a fixed chance of being shot or by being hit of lightning. Neither of these are random events. If I'm at home sitting in my chair reading I'm not going to be hit by lightning. If I'm out of the golf course holding a club while lightning is happening I would expect to be hit.
Quote:
There is a clear mathematical definition of random. Random is "Without any pattern or preference".
I'm confused. You give this definition but you started this thread because your observations didn't fit the pattern you expected. Which is it? Pattern or no pattern?
Quote:
Simply put, there is no motive at all to make anything chance based NOT random. I highly doubt it is intentional. It is clearly a bug, and should be fixed. It has been bringing a fair amount of frustration to the community for a long time. And if it extends to modules it can make things that should have a reasonable reliability in a fair amount of attempts completely unreliable.
Random is often frustrating because it doesn't fit a pattern.
Quote:
Bottom line, 10 repetitions of the same result for a 50% chance event should be an extremelly rare event. It is better for everybody this way and it is automatically guaranteed by implementing correctly the random generator. Five repetitions for a 5% chance event should be even more rare...
I agree. It should be a rare event. How do we know it's not a rare event? The rng doesn't operate on a per character event. It's quite literally impossible to evaluate the mission selection system, as an end user. The 50/50 example you give is across all missions given. If you're getting 5 kill in a row and some else gets 100 courier with no kill we may still be at 50/50.
How many missions are run in a day? I run out of a station that has 6 distro agents. If you use your cancel on long range missions or to help group the destination along one path of travel the volume of missions you can do is staggering. Given the volume of missions given by the server as a whole is getting 5 kill in a row a rare event?
There is absolutely nothing in mathematics that states _your_ distribution will approach the average. This is known as the gambler's fallacy.
I'd also like to point out that you admit everything you describe is possible.
Since people here are willing to discuss things intelligently I'd like to bring up a related topic. Why would they even be using a rng? A rng is a waste of CPU cycles. There are less intensive ways to go about it.
Say all your game mechanics operate on a 1-100 roll. You create a large array, fill it with an equal distribution of each value 1-100 and then shuffle it. Each time the server starts up fold the array some amount of times to redistribute the order of the numbers.
When an event happens simply grab the next number and increment the index. It's less CPU cycles than calculating the next number and it guarantees a given distribution of results. How ever it is not random.
This raises the question. Do you really want random or just a predictable distribution?
|
Tunak
|
Posted - 2007.08.31 11:14:00 -
[4]
Edited by: Tunak on 31/08/2007 11:19:31
Originally by: pandymen
It also seems as though when you decline a mission, you tend to get the "other" kind of mission.
This was in the patch notes a while back. When you decline a mission your next mission is not supposed to be the same mission you declined. However I'm not sure what their definition of 'same' is. I've received the same named mission but against a different type of NPC after canceling. Technically it is a different mission or is the mechanic broken?
EDIT: Fixed bad quote formatting.
|
Tunak
|
Posted - 2007.08.31 12:10:00 -
[5]
Originally by: Chloey Rans
Or in
other words the probabilty density function of the underlying random generator
does not generate an even distribution for the random variable but something like a beta PDF. This would result in the same mean values but different higher moments, e.g. two events would be more likely than others.....
What I describe is not a random number generator. It is a method of generating a perfect distribution. This is what the OP is asking for. The OP does't want random. They want predictability which as the OP has stated is the opposite of random.
|
Tunak
|
Posted - 2007.08.31 13:35:00 -
[6]
Edited by: Tunak on 31/08/2007 13:35:35
That's 2 personal attacks in 2 posts. Please stop and maintain the civility of the thread.
Originally by: Chloey Rans
slowly, slowly: a random generator is a method of generating a distribution of
a random event (say roll on d6: event = result of one dice roll). if the distribution is even, every event hase the same probalility (e.g 1/6) and voala, you have the perfect random generator for a even distribution (if thats what you wanted). And yes, you have than a 'predictability' in a statistical sense.
But if your random generator generates: 6 (event) = 2/6 (probability),
5 = 1/12, 4 = 1/12, 3=1/12, 2=1/12, 1 = 2/6 you have the same expectational
value of your dice-experiment (E=3.5) but an uneven distribution where it
is more likely that you roll a 1 or a 6, and thus, several (lets say 3) 6 in a row are more likely (=0.037037) compared to a even distribution (=0.0046296).
Yes that's accurate but not applicable to the situation. You would never evaluate a rng by looking at the mean. The only way to evaluate it is by frequency.
EDIT: Typo
|
Tunak
|
Posted - 2007.08.31 13:41:00 -
[7]
Originally by: Scordaf
This may very well be the case, but it merely serves to illustrate that the current system doesn't work. Even if server-wide the mission system is still at 50/50, for the lucky/unlucky end-user, this is not the case - which suggests that the system needs checks in place to help insure that each individual player gets a more 'random' distribution of missions from within their own subset of missions received.
But that would not be random. That would be deciding on an outcome before hand and then altering your results set to match it.
Additionally this would be a non trivial increase in CPU and storage load on the server with no gain to the end user.
|
Tunak
|
Posted - 2007.09.04 16:25:00 -
[8]
Originally by: Etho Demerzel
Tunak, for a large amount sample a random sequence result is predictable and it is perfectly even. The chances against it are so low that you can simply ignore then. Anomalies happen, but they are not likely.
So is a random sequence predictable or is it "without any pattern or preference" and thus unpredictable as you stated in a previous post?
If it is predictable then when is the next time I'm going to get a streak of 4 kill missions in a row from one of my distribution agents?
Quote:
Now the data:
Marketing missions executed: 1543
Number of Kill missions: 79
Number of courier missions: 1464
Percentage of kill missions: 5.12%
Percentage of courier missions: 94.88%
So your overall numbers are matching up you're simply concerned with the streaks. As I said before this is a case of "Gambler's Fallacy."
Let's talk about looking at anomalies and streaks. For example let's take this sequence.
HTHTHTHTHTHTHT
If we pull a random subset and get HHHHHHH we'd expect that it's impossible to get a T when in fact it is equally as likely to get a T as an H.
What if we pull a random subset and get HHTT? What about HTHT? Which of these three subsets (HHTT, HTHT, HHHH) correctly describes all coin flips ever made and why?
Sure you can evaluate subsets but you cannot apply those results to the whole. That's the definition of independent events.
The 95/5 is a limit. As the number of missions pulled approaches infinity the distribution between courier and kill approaches 95% courier and 5% kill. That's it. There is absolutely nothing in mathematics that states a randomly selected subset will meet the 95/5 split. Hell nothing in math says the complete set will ever reach a 95/5 split.
Quote:
Oh, and one last thing. I have received the same mission after refusing a mission (exactly the same, against the same NPC etc) several times...
I was simply repeating what the patch notes said. If you think it's broken submit a bug report.
|
Tunak
|
Posted - 2007.09.04 20:39:00 -
[9]
Originally by: Etho Demerzel
You seem like a junior Calculus student thinking you know a lot about math, when you don't even have adequate knowledge about logics.
And yet you can't debunk them and must resort personal attacks.
|
Tunak
|
Posted - 2007.09.06 18:43:00 -
[10]
Originally by: Etho Demerzel
I am an engineer as well
This explains a lot.
You started this whole thread over what is basically 1.85 missions.
For a 95/5 split with the number of missions you quote you should see 77.15. You saw 79.
You obviously like finding fault with others. It's the only thing you've done in this thread. Fault with CCP. Fault with every responder. You participate in personal that have no point in a discussion.
The straight up split didn't provide enough fault so you had to go digging and find the streaks.
I don't contest that the streaks are highly unlikely to happen. However answer me this. What says your 1543 sample set has to have a perfect 95/5 split? What says your sample set has to have the perfect number and length of streaks?
The streaks did happen. Probability is not prophecy. I'm sorry you had to do/turn down one extra kill mission.
|
| |
|