Imperia Online International
March 28, 2024, 16:04:41 PM *
Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length
 
   Home   Help Search Login Register  
Pages: [1]   Go Down
  Print  
Author Topic: Rumours accuracy - question for IO experts and REAL statiticians  (Read 2363 times)
0 Members and 1 Guest are viewing this topic.
AddisMakem
Pointillism is pointless
Lieutenant
***

Karma: +12/-5
Offline Offline

Posts: 191


Nil carborundum batardi


« on: September 01, 2009, 19:40:03 PM »

I know rumours only return answers with a 50% accuracy, so two questions:

1.  For IO experts: Assume that if real resources or army are 100,000 the rumour range is between 50,000 and 200,000?  Or is that wrong and if so why?

2. For proper statos: How many rumours must I run before the average result gives me an answer within 1. 10% 2.  5%   3.  2%  of the real number.  Fairly sure this calculation can be done but it's more than 20 years since I did stats at uni and even then it was part of an economics degree  *hihi*
« Last Edit: September 01, 2009, 19:42:18 PM by AddisMakem » Logged
AddisMakem
Pointillism is pointless
Lieutenant
***

Karma: +12/-5
Offline Offline

Posts: 191


Nil carborundum batardi


« Reply #1 on: September 01, 2009, 20:05:13 PM »

Makes sense - kind of - but 50% of 200,000 is 100,000 so it depends on what IO mean by a 50% error margin.  If your answer is definitive then I'm happy the first part of my question is answered.

Anyone care to hazard an answer to the second part?
Logged
tiger
Community Manager
*

Karma: +694/-194
Offline Offline

Posts: 3319


« Reply #2 on: September 01, 2009, 21:05:39 PM »

i wont go into discussion but i think in case of 100 k you are between 50k and 200k. Because you can get bad luck and 100 k is 50% of something. what? 200k. and it could be -50 % so arrond 67 k.  %0 % of 67 is 33 + 67 = 100 k. My opinion, but i can ask programers if somebody really wish.
Second part.
With few rumors( 2 is minimum) you can clearly guess within some % difference. But it depends on luck. so it cant be given in form: do 7 rumors and you are in +3% accuracy.
Example. If you get one time 200 k and next time 80 k you know minimum is 100 k and max is 160 k. the bigger difference you get, more accurate you are. but mostly with few rumors you see if there is enough or not enough.
Logged


tiger
Croatian Community Manager
AddisMakem
Pointillism is pointless
Lieutenant
***

Karma: +12/-5
Offline Offline

Posts: 191


Nil carborundum batardi


« Reply #3 on: September 01, 2009, 21:17:39 PM »

Thanks Tiger

Re first part: yes I'd like a clear answer from the programmers about what the rumour range actually is - otherwise second part of question becomes impossible to resolve.  Would really appreciate this *Smiley*

Fairly sure there is a statistically valid answer to the second part of the question and, hate to say it, but doubt that sample size of 7 would get us to within 2% accuracy.  Trouble is all my stats notes are back in the UK while I continue my slow waltz thro Africa  *xxx*  Kind of guessing but would think it must be nearer to a sample size of 30 but it is a hell of a long time since I did any real stats.  Point is if I can resolve the question then obviates most of the need for spying - given I chuck a couple of suicide attacks in - which gives me another thought..........

So, again, if any statos can remind me of the formulae would be really grateful

Logged
Loboscuro
Private
*

Karma: +1/-0
Offline Offline

Posts: 22


« Reply #4 on: September 02, 2009, 19:01:53 PM »

Second question: Like you, hell if I remember since I studied statistics 10 years ago. OK, I know statistics is not the name of a exotic fruit, like the manual says, but at least is a exotic part of maths  *crazy* Hope you get the answer from a real brain, mine is wasted  *freak*

First question: I have to disagree here with Tiger. As far as I understand the explanation, the 50% applies to the real number of troops/resources in the province, not to the number you get with one rumour. So, if the real number is 100k, you can get answers in the range of 50k-150k.

Ok, it is possible the programmers made it in the way that the 50% applies to each and every answer you get with a rumour, but it would be really weird. I mean is much easier for the program to read the real number (100k) and then randomize a number applying a +/- 50%, than "creating" a number and then checking that number is +/- 50% of the real one...dont you think?
Logged
chegewara
Captain
****

Karma: +62/-58
Offline Offline

Posts: 407


« Reply #5 on: September 02, 2009, 19:18:04 PM »

in this situation we can make 2 algorithms. first is real value +/-50% and this is 50%-150% or second, real number divide or multiple by 2 and this give us range 50%-200% like said Tiger. i really think that second option is real
Logged
AddisMakem
Pointillism is pointless
Lieutenant
***

Karma: +12/-5
Offline Offline

Posts: 191


Nil carborundum batardi


« Reply #6 on: September 02, 2009, 19:32:08 PM »

Thanks all *Smiley*

Be great to get clarification on the rumour range from the programmers as there is a real diversity of views from people whose opinions I really respect

Ran 5 rumours against 2 targets who I'm fairly confident I know the real numbers on from recent battle reports and there was a big diversity - the first average was within 6% but the second was 23% wide of the mark.

Will run more rumours later on each to see if I can get it to converge a bit more.  But if there are any real statos out there perhaps they can ease mine and Loboscuro's aging minds *Smiley*

Logged
tiger
Community Manager
*

Karma: +694/-194
Offline Offline

Posts: 3319


« Reply #7 on: September 02, 2009, 21:23:57 PM »

because rumors are based on luck you cant predict that so easy. In large number of rumors you would really get average, but you cant make 200 rumors on 1 player prov.
First question: I have to disagree here with Tiger. As far as I understand the explanation, the 50% applies to the real number of troops/resources in the province, not to the number you get with one rumour. So, if the real number is 100k, you can get answers in the range of 50k-150k.
yes, 50 % applies to real number, not to rumored one. So if player have 100k exactly, then rumors can show 50k to 150k. i agree on that.
but when you get rumor, its a random number in +-50% of some real resource number which you wish to figure out. with one rumor you can only get min and max resources in some interval.
To get boundries you must assume you missed maximum 50% in + or -.
Rumors showed 100 k and thats 50 % of real resource because it has shown only half of it. so maximum is 200k, he cant have more.
Rumors showed 100 k and thats 150 % of real resource because it has shown maximum value you can miss. so 100 k is 150 %. real resources is 66.7 k
thats minimum and maximum in the way i understand math.
for example if you get once 100k and second time 290k you can be sure real resources are close to 200 k. those far numbers in rumors are most precise.
whenever i attacked, that worked fine for me, unless i was to late because calculating min and max. *Smiley*
I will really ask programmers, but after the maintinence will be over and i will gave you answer.
Logged


tiger
Croatian Community Manager
higrm
Sergeant
**

Karma: +3/-0
Offline Offline

Posts: 77


« Reply #8 on: September 03, 2009, 10:09:32 AM »

In Realm 1, where you can rumor all day long without limit, just rumor yourself.  you know the exact amount of resources you have there.  put 10k gold in a province and write down the numbers you get when you rumor yourself.  You will soon see how accurate or inaccurate you are after 20, 30, 100 tries. 

As far as statistics goes, this is a random distribution, (any number between -50% to +50% has an equal likelihood of being chosen) and not a Gaussian  or Normal distribution. Therefore, you will need a few more tries to get the min and max of the range.  The true value should be the average of the min and max values, once you have identified them.

I did the test on my own province and you can see the average approaching the real value fairly quickly with just gold in the province.  I'm not sure whether having a mix of resources would give you a wider spread, though the average of the min and max should still be close to the real value. (is the total gold value randomized +/-50% or is each resource amount randomized 50% and then converted to gold?)

higrm
Logged
Loboscuro
Private
*

Karma: +1/-0
Offline Offline

Posts: 22


« Reply #9 on: September 03, 2009, 12:59:03 PM »

Yep, Tiger is right. I cant position myself to start the maths from the real number cos I dont know that figure. I have to position myself from the number I get on the first rumour and then apply the +/- 50%, so the range is 50-200 k. I see it now.

Anyhow, you have to love risk way too seriously to go on an attack based on the info obtanied with one rumour. I do 3-5 on potential victims and, based on that info, I decide to send spies. Except at the beginning of the game, when I dont have enough rumours a day...but if I ran out of rumours and I dont want to send spies blindly, goldmines become my close friends  *hahaha*
Logged
Pages: [1]   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.12 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!