Graviton
Well-Known Member
That's the same thing. Each time you have a 25% chance, and over a sufficient amount of time and attempts you'll have won that prize 25% of the time. What it doesn't mean is that every fourth time you'll win what you want.
-
We are looking for you! Always wanted to join our Supporting Team? We are looking for enthusiastic moderators!
Take a look at our recruitement page for more information and how you can apply:
T.farm level up discussion
- Thread starter voidsource
- Start date
That's the same thing. Each time you have a 25% chance, and over a sufficient amount of time and attempts you'll have won that prize 25% of the time. What it doesn't mean is that every fourth time you'll win what you want.
voidsource
Member
heres that thing with what im trying to be clear about
" The fallacy of composition arises when one infers that something is true of the whole from the fact that it is true of some part of the whole. "
in this case that bc theres a 25% chance of me winning this prize, that means everyone will get a 25% chance of winning. And specially over x period of time. And i keep mentioning that isn't how this work. U have 25% chance every time u try this encounter but it does not mean u will be on the 25% if given x amount of time.
So by telling ya that after 13/14 weeks of no upgrade prize, it does not mean ill get a prize next week simply bc x time has passed. You gotta remember that my last upgrade was about in week 20 (3 months ago ) and that meant my % was higher than 25%. Hence why that overall x time period did not work. Faulty.
That being said, my original message of maybe having it tweaked so that players dont go excessively into long wait times to get the higher/better reward out of the options provided, should be something to take into consideration. Afterall, it is a game which tries to overall keep clientele happy.
" The fallacy of composition arises when one infers that something is true of the whole from the fact that it is true of some part of the whole. "
in this case that bc theres a 25% chance of me winning this prize, that means everyone will get a 25% chance of winning. And specially over x period of time. And i keep mentioning that isn't how this work. U have 25% chance every time u try this encounter but it does not mean u will be on the 25% if given x amount of time.
So by telling ya that after 13/14 weeks of no upgrade prize, it does not mean ill get a prize next week simply bc x time has passed. You gotta remember that my last upgrade was about in week 20 (3 months ago ) and that meant my % was higher than 25%. Hence why that overall x time period did not work. Faulty.
That being said, my original message of maybe having it tweaked so that players dont go excessively into long wait times to get the higher/better reward out of the options provided, should be something to take into consideration. Afterall, it is a game which tries to overall keep clientele happy.
CaptainKirk1234
Active Member
I have had 9 without a upgrade I dont complain, infact I have never had a upgrade.
wolfhoundtoo
Well-Known Member
If you play for eternity you will get really close to 25% of the upgrades that you want. It's based on probabilities which is a science and in fact it does say that you can conclude things about the whole population based on what happens to the individual parts. There's a whole host of reasons for that but clearly you only pay attention to those parts that support your desires. That's how probabilities work. You are complaining because in the short term you don't get what you want and thus it's broken. What you want isn't a random generator but (I forget the term) is basically a loot table that is weighted to give you prizes the longer you don't win that prize. That is entirely different than the method Inno uses. Yes if you look around hard enough you will find where they've explained or announced how it works.
CaptainKirk1234
Active Member
Makes sense just because you havent got it before doesnt increase the chances of getting it next time.
Vger
Well-Known Member
But that doesn't apply here. We already have visibility to the whole. There is a 25% chance of getting the TF upgrade. That is the whole. Knowing just that, you can use basic math from probability theory to understand how the system will behave if it is working as advertised.
You are doing the opposite here. You are looking at a very small number of parts and expecting them to look like the whole.
Turns out there's a term for that too:
" A fallacy of division is the error in logic that occurs when one reasons that something that is true for a whole must also be true of all or some of its parts "
Algona
Well-Known Member
tl;dr OP's complaint that there is a problem with the TF Upgrade drop rate is not supported by any evidence given by the OP or any other poster. OP's own experience as reported in this thread supports that the RNG is working correctly. OP does not understand probability.
There are damn good reasons folk talk about meaningful sample size in discussions of probability.
Can't ague with that, that's a matter of personal choice. So no problem if you choose not to. But you have to live with the consequences of your choice, namely the futility of trying to make a point about probability with no reliable data.
I spent 7 months collecting and daily annotating here on the forums 10K WW/FoY collections to find out the Diamond drop rate. But I'm sick like that.
The reason you need a big sample size and have to record each instance is that the actual data is susceptible to statistical analysis for reliability of the data, pattern detection, and to support any conclusions to be drawn.
No hard data means any conclusions drawn are baseless guesses.
Your choice.
As far as asking your Friends what their TFU drop rate is?
Same reasoning applies, no hard data just their memories. Add in observer and respondent bias, namely by asking the question you prompted them to answer in a way that would support your preconceived notions.
Umm, well, since your results over time are about what can be expected I don't rightly see what the problem is, maybe even worked in your favor?
Excluding your 14 week run of bad luck you garnered 7 TFUs in 18 weeks, above average. Since those all came before the run of bad luck you got the use of all those lvl 2 TFs for more time then if they had been evenly distributed thus garnering approx 2000 extra FPs.
[glass half full]Seems lucky to me.[/glass half full]
You're right, a 14 week run of misses is a reason to ask what is going on.
So you came to the forum and asked about it. Reasonable action on your part.
Your response to the answers you got isn't as reasonable. It ain't a matter of folk being a clique ganging up on you, just folk who understand probability better then you telling you what you don't want to hear 'cause it conflicts with what you want and what your 'common sense' is telling you.
Your problem here isn't a 14 week long run of bad luck. Your problem was caused by you selecting that 14 week run out of context of and ignoring your total experience of collecting 7 TFUs in 32 weeks.
Selecting part of your data to analyze is a no-no in the probability game. Lead you to the bad conclusions that prompted this thread.
Understandable that you would make the mistake of selecting the data you want given the following exchange:
I did a double take and laughed hard and long.
Typo? Did you mean to say probably?
'Cause you nailed it dead on with what you did type.
Everything you said in this thread and from September screams you don't understand probability.
No problem there as long as you don't bet money.
Here's what you do if you want to gain some understanding of probability. Buy your insurance (doesn't matter what kind of insurance) agent a cup of coffee (dinner night be more appropriate, a bottle of their favorite spirits moreso) and ask them to explain how probability influences insurance companies.
Their livelihood depends on that interaction. To wit, individual interactions don't matter, just the collective results.
Unless you have a specific question, I'm done here. No point in discussing probability with someone who doesn't grasp the fundamentals.
There are damn good reasons folk talk about meaningful sample size in discussions of probability.
Can't ague with that, that's a matter of personal choice. So no problem if you choose not to. But you have to live with the consequences of your choice, namely the futility of trying to make a point about probability with no reliable data.
I spent 7 months collecting and daily annotating here on the forums 10K WW/FoY collections to find out the Diamond drop rate. But I'm sick like that.
The reason you need a big sample size and have to record each instance is that the actual data is susceptible to statistical analysis for reliability of the data, pattern detection, and to support any conclusions to be drawn.
No hard data means any conclusions drawn are baseless guesses.
Your choice.
As far as asking your Friends what their TFU drop rate is?
Same reasoning applies, no hard data just their memories. Add in observer and respondent bias, namely by asking the question you prompted them to answer in a way that would support your preconceived notions.
Umm, well, since your results over time are about what can be expected I don't rightly see what the problem is, maybe even worked in your favor?
Excluding your 14 week run of bad luck you garnered 7 TFUs in 18 weeks, above average. Since those all came before the run of bad luck you got the use of all those lvl 2 TFs for more time then if they had been evenly distributed thus garnering approx 2000 extra FPs.
[glass half full]Seems lucky to me.[/glass half full]
You're right, a 14 week run of misses is a reason to ask what is going on.
So you came to the forum and asked about it. Reasonable action on your part.
Your response to the answers you got isn't as reasonable. It ain't a matter of folk being a clique ganging up on you, just folk who understand probability better then you telling you what you don't want to hear 'cause it conflicts with what you want and what your 'common sense' is telling you.
Your problem here isn't a 14 week long run of bad luck. Your problem was caused by you selecting that 14 week run out of context of and ignoring your total experience of collecting 7 TFUs in 32 weeks.
Selecting part of your data to analyze is a no-no in the probability game. Lead you to the bad conclusions that prompted this thread.
Understandable that you would make the mistake of selecting the data you want given the following exchange:
I did a double take and laughed hard and long.
Typo? Did you mean to say probably?
'Cause you nailed it dead on with what you did type.
Everything you said in this thread and from September screams you don't understand probability.
No problem there as long as you don't bet money.
Here's what you do if you want to gain some understanding of probability. Buy your insurance (doesn't matter what kind of insurance) agent a cup of coffee (dinner night be more appropriate, a bottle of their favorite spirits moreso) and ask them to explain how probability influences insurance companies.
Their livelihood depends on that interaction. To wit, individual interactions don't matter, just the collective results.
Unless you have a specific question, I'm done here. No point in discussing probability with someone who doesn't grasp the fundamentals.
Johnny B. Goode
Well-Known Member
A little math. So the TF upgrades have been available for 32 weeks. The first 18 weeks you won 7. That's way over the stated chance percentage. Yet you didn't complain. Now after 32 weeks you still only have those 7, but that still leaves you at pretty much the stated chance percentage. I am thinking that the gist of your complaint is that the game paid out the stated chance percentage too early for you and you would rather have those 7 wins spaced out over the 32 weeks more evenly. Because the math doesn't support any other conclusion.
Emberguard
Well-Known Member
There's been countless times outside of FoE across different platforms where I've gotten 10 failed flips in a row at a 50/50% chance. Terrace Farm upgrade has even less chance of showing up at only 25%. 14 straight weeks isn't that unusual no matter how much we may sympathize with your lack of TF upgrades
A win every third win is 33%, higher then the 25% chance.
25% of 30 weeks (change was implemented July 20th) would mean receiving:
(30 / 100) x 25) = 7.5 upgrades
So if you've already received 7 upgrades, that's right on point for the amount of time lapsed
Emberguard
Well-Known Member
If you want a fail safe method of getting a reward for a extended period of time with fails, you'd also need a fail safe method to prevent you from winning extra. Which would either mean rigging results or straight up dumping the % altogether and just giving a rotating set prize
Not true. There's been players who receive multiple Daily Specials in a row just a there's been players who have bad luck streaks. In statistical probability these are called outliers
Flavius Belisarius
Active Member
Wow!
Didn't know there were TF Upgrades (because I never went to LVL4) until today when I got a Face of the Ancients Upgrade.
Even old dogs can learn new tricks!
Didn't know there were TF Upgrades (because I never went to LVL4) until today when I got a Face of the Ancients Upgrade.
Even old dogs can learn new tricks!
Flavius Belisarius
Active Member
I found this and realized... I don't care.
If you do, great; but is it really worth fretting of randomness?
Computer generated "Random Numbers" are used in many applications. Indeed, there is a whole set of numerical "Monte Carlo" techniques based on them. This page describes how (most) random number generators work and most importantly it lets you design and test your own random number generator. To do so just click on the applet nearby.
What is a random number generator?
Most random number generators generate a sequence of integers by the following recurrence:
x 0 = given, x n+1 = P 1 x n + P 2 (mod N) n = 0,1,2,... (*)
The notation mod N means that the expression on the right of the equation is divided by N, and then replaced with the remainder.
To understand the mechanics consider the following simple Example. (Choose Example on the applet to study this example further.)
x 0 =79, N = 100, P 1 = 263, and P 2 = 71
Then
x 1 = 79*263 + 71 (mod 100) = 20848 (mod 100) = 48,
x 1 = 48*263 + 71 (mod 100) = 12695 (mod 100) = 95,
x 1 = 95*263 + 71 (mod 100) = 25056 (mod 100) = 56,
x 1 = 56*263 + 71 (mod 100) = 14799 (mod 100) = 99,
Subsequent numbers are: 8, 75, 96, 68, 36, 39, 28, 35, 76, 59, 88, 15, 16, 79, 48. The sequence then repeats . (This indicates a weakness of our example generator: If the random numbers are between 0 and 99 then one would like every number between 0 and 99 to be a possible member of the sequence.
The parameters P 1, P 2, and N determine the characteristics of the random number generator, and the choice of x 0 (the seed ) determines the particular sequence of random numbers that is generated. If the generator is run with the same values of the parameters, and the same seed, it will generate a sequence that's identical to the previous one. In that sense the numbers generated certainly are not random. They are therefore sometimes referred to as pseudo random numbers.
Transformation of the original sequence.
Of course one may want random numbers not as integers in a given range, but for example as uniformly distributed real numbers in a certain interval, or perhaps as real numbers of (almost) arbitrary size, but clustered around the origin. Distributions of that sort can be obtained by suitably transforming the original random numbers. For example, to transform a sequence defined as above into an evenly distributed set of real numbers in the interval from 0 to 1 simply divide each of the original numbers by N. In the remainder of this page, though, we just consider the sequence defined by (*) itself.
What makes a good random number generator?
That's a good question! Several answers are possible, for example:
The applet on this page takes a different tag. It plots a certain number of points
(x i ,x i+k )
for certain values of k = 1,2,3,... . Intuitively, for a random sequence, one should obtain a set of points distributed "evenly" , "randomly" or "uniformly " over a square. It is not easy to make these concepts precise, but it is sometimes glaringly apparent when a set of points is not distributed in this way. Plotting 100 points with k=1 for our example generator above generates the picture nearby. (It's is shown here at half its original size.) In the figure the first coordinate measures horizontal distance from the left margin of the red box, and the second coordinate measures vertical distance downwards from the upper margin of the red box. Thus some of the points in this box have coordinates (79,48), (48,95), (95,56), etc. The values of the coordinates are scaled to fill the entire red box (which in this case measure 200 by 200 pixels). It's clear that there are only 20, points, and, since 100 were drawn, five lie on top of each other for each of the black dots. Moreover, the dots appear to lie along six slanted lines. As pointed out above, for a " good" random number generator there should be 100 points, and the distribution should be " random".
The second figure nearby shows a distribution of 1,000 points obtained with a widely used and well tested and analyzed random number generator using
P 1 = 16807, P 2 = 0, and N= 2 31 -1 = 2147483647.
This generator is described in the reference by Park and Miller given below.
One reason for the seemingly peculiar choice of N is that that particular number is the largest integer than can be represented on a Unix machine or in Java.
To illustrate the abilities of this applet consider the following Figure which shows three sequences of 100,000 points each, using the same generator, for k=1 (red), k2 (green), and k=3 (blue). Reassuringly, no systematic patterns are readily apparent.
Operating the Random Applet
Clicking on the box at the beginning of this page will cause a control window to pop up that looks much as illustrated in the image nearby. A drawing window that will contain the plots also appears.
There are mostly text input windows and buttons marked ">", "<<" etc.. To change an entry of a text window highlight the text and write a new number. You can also change parts of the number. It is important to terminate every change by pressing the Return or Enter key. The green buttons increment or decrement the text windows they surround. "< " and ">" decrement or increment by 1, the other buttons by larger amounts, e.g., by going to the closest prime number or the closest power of 2. Points of the form
(x i ,x i+k)
can be drawn for up to three different values of k, as determined in the three lines labeled Cycle 1, 2, and 3.
Following is a description of some less obvious features:
If you do, great; but is it really worth fretting of randomness?
Computer generated "Random Numbers" are used in many applications. Indeed, there is a whole set of numerical "Monte Carlo" techniques based on them. This page describes how (most) random number generators work and most importantly it lets you design and test your own random number generator. To do so just click on the applet nearby.
What is a random number generator?
Most random number generators generate a sequence of integers by the following recurrence:
x 0 = given, x n+1 = P 1 x n + P 2 (mod N) n = 0,1,2,... (*)
The notation mod N means that the expression on the right of the equation is divided by N, and then replaced with the remainder.
To understand the mechanics consider the following simple Example. (Choose Example on the applet to study this example further.)
x 0 =79, N = 100, P 1 = 263, and P 2 = 71
Then
x 1 = 79*263 + 71 (mod 100) = 20848 (mod 100) = 48,
x 1 = 48*263 + 71 (mod 100) = 12695 (mod 100) = 95,
x 1 = 95*263 + 71 (mod 100) = 25056 (mod 100) = 56,
x 1 = 56*263 + 71 (mod 100) = 14799 (mod 100) = 99,
Subsequent numbers are: 8, 75, 96, 68, 36, 39, 28, 35, 76, 59, 88, 15, 16, 79, 48. The sequence then repeats . (This indicates a weakness of our example generator: If the random numbers are between 0 and 99 then one would like every number between 0 and 99 to be a possible member of the sequence.
The parameters P 1, P 2, and N determine the characteristics of the random number generator, and the choice of x 0 (the seed ) determines the particular sequence of random numbers that is generated. If the generator is run with the same values of the parameters, and the same seed, it will generate a sequence that's identical to the previous one. In that sense the numbers generated certainly are not random. They are therefore sometimes referred to as pseudo random numbers.
Transformation of the original sequence.
Of course one may want random numbers not as integers in a given range, but for example as uniformly distributed real numbers in a certain interval, or perhaps as real numbers of (almost) arbitrary size, but clustered around the origin. Distributions of that sort can be obtained by suitably transforming the original random numbers. For example, to transform a sequence defined as above into an evenly distributed set of real numbers in the interval from 0 to 1 simply divide each of the original numbers by N. In the remainder of this page, though, we just consider the sequence defined by (*) itself.
What makes a good random number generator?
That's a good question! Several answers are possible, for example:
- The sequence generated by (*) isn't random at all, so there is no good random number generator of that form.
- A sequence (*) is good if it passes several well established statistical tests.
- Or, it's good if it gives good results in particular applications (where of course the meaning of "good results" is heavily dependent upon the context).
(x i ,x i+k )
for certain values of k = 1,2,3,... . Intuitively, for a random sequence, one should obtain a set of points distributed "evenly" , "randomly" or "uniformly " over a square. It is not easy to make these concepts precise, but it is sometimes glaringly apparent when a set of points is not distributed in this way. Plotting 100 points with k=1 for our example generator above generates the picture nearby. (It's is shown here at half its original size.) In the figure the first coordinate measures horizontal distance from the left margin of the red box, and the second coordinate measures vertical distance downwards from the upper margin of the red box. Thus some of the points in this box have coordinates (79,48), (48,95), (95,56), etc. The values of the coordinates are scaled to fill the entire red box (which in this case measure 200 by 200 pixels). It's clear that there are only 20, points, and, since 100 were drawn, five lie on top of each other for each of the black dots. Moreover, the dots appear to lie along six slanted lines. As pointed out above, for a " good" random number generator there should be 100 points, and the distribution should be " random".
P 1 = 16807, P 2 = 0, and N= 2 31 -1 = 2147483647.
This generator is described in the reference by Park and Miller given below.
One reason for the seemingly peculiar choice of N is that that particular number is the largest integer than can be represented on a Unix machine or in Java.
To illustrate the abilities of this applet consider the following Figure which shows three sequences of 100,000 points each, using the same generator, for k=1 (red), k2 (green), and k=3 (blue). Reassuringly, no systematic patterns are readily apparent.
Operating the Random Applet
Clicking on the box at the beginning of this page will cause a control window to pop up that looks much as illustrated in the image nearby. A drawing window that will contain the plots also appears.
There are mostly text input windows and buttons marked ">", "<<" etc.. To change an entry of a text window highlight the text and write a new number. You can also change parts of the number. It is important to terminate every change by pressing the Return or Enter key. The green buttons increment or decrement the text windows they surround. "< " and ">" decrement or increment by 1, the other buttons by larger amounts, e.g., by going to the closest prime number or the closest power of 2. Points of the form
(x i ,x i+k)
can be drawn for up to three different values of k, as determined in the three lines labeled Cycle 1, 2, and 3.
Following is a description of some less obvious features:
- DONT/DRAW: Normally the picture in the drawing window gets redrawn every time a change is made in the control window. You can suppress the drawing (e.g., if you want to make many changes) by pressing the yellow button in the top row. Its label "DONT" indicates whether drawing currently takes place. Click on "DONT" to suppress it or on "DRAW " to activate it.
- Quit. As you might expect, you can dispose of the control wind and the drawing window by clicking on the Quit button. You can obtain the same effect by clicking on the applet at the top of this page, or by typing "q", "Q", "x", or "X" in either the control window or the drawing window
- The Menuin the top row lets you set the applet to analyze specific random number generators. More may be added in the future, but right now four are available:
- The default generator that does not work very well and lets you discover strange patterns/
- The high quality generator proposed by Park and Miller in the reference quoted above.
- A random number generator I use in my Microscope Software. Interestingly that generator does quite well in the tests performed by this applet, but it shows non-random effects in the Microscope package. (The reason I have not removed it is that doing so would effect all the examples in the manual.)
- The Example generator used for illustration at the beginning of this web page.
- The 16 bit version of the infamous RANDU discussed at length in Park and Miller (referenced below).
- A random number generator published in 1978 by P. Grogono. Display it with k=14 or k=38.
Johnny B. Goode
Well-Known Member
Maybe Inno should just pay a team of people to pull numbers out of a hat 24/7. I'm sure people would stop complaining then. 
Or not.
Or not.
Agent327
Well-Known Member
Maybe Inno should just pay a team of people to pull numbers out of a hat 24/7. I'm sure people would stop complaining then.
Or not.
They would tell the hat is rigged and those people that pull numbers biased.
RazorbackPirate
Well-Known Member
Just got my 8th TF upgrade from GE. That puts me at 8 in 32 weeks. Right at 25%. Seems all is well with the RNG.
Emberguard
Well-Known Member
*30 weeks including current week
RazorbackPirate
Well-Known Member
Thanks for the clarification. Let me rephrase.
Just got my 8th TF upgrade from GE. That puts me at 8 in 30 weeks. Just ahead of 25%. Seems all is well with the RNG.
P C C
Active Member
It would be possible to code a permutation or 'sampling without replacement' approach to randomness, for example setting up so every 8 rewards on this encounter gives 4x goods, 2x TF upgrades, 2x other reward in random order (or every 12 giving 6, 3, 3 ...). But the coding would be much more complicated, and they'd have to keep all of that in memory. And you're focused on one specific GE encounter you care about. But if they were to take that approach, why would it be just for this one random event? How about the other GE encounters, or random 4x in Settlements, or even random damage in battle? That would be a giant undertaking and a massive waste of time, energy, memory.
As for the probability, I've also got 8 TF upgrades so far right around the expected.
As for the probability, I've also got 8 TF upgrades so far right around the expected.
Glockgemini
Member
How many useless Gate, Face and TS upgrades do I need to trade for a TF or SOK upgrade? Inno passes this crap out like candy at Halloween but the useful stuff is like finding a needle in a haystack.
RazorbackPirate
Well-Known Member
Funny how it works that way.