Conquer Club

I lose 15-5, so other guys +6 bonus holds. I lose 9-4. +9 bonus holds. I attack with every single attacking army I have and get 4 territories. Meanwhile other guy is rolling me 3-3 losing nothing. Which inspired me to start this thread.

The question is not, are the dice random... but how often do you find that one player seems to have particularly good dice and the other player seems to crap out every time? And I mean regardless of whether your getting the good luck or the bad luck (as I'll admit I'm often on the good side as well as the bad side), who often do you feel like the dice are one sided... maybe on a scale of "A" being perfectly even (regardless of good or bad), "B" being somewhat lopsided, and "F" being all one sided good dice... other guy/you get all bad rolls.

I'd say in 50% of my games the dice rolls even out. 40% are somewhere in the B range... mostly evens out either good or bad, and then there are about 10% of games that, like the one above, just fail ... they earn the big fat F ... there is no semblance of balanced dice. Anyone else care to share their experience...

maybe we need a new section in the forum just for dice complaints and tells of woe from recent turns in games that we are never given the game numbers too... i can't even read them, especially when they are the same wordy complaints by the same wordy people... sigh... my eyes just glazed over 2 lines in, and i woke up with drool on the keyboard of my laptop... "is it over yet? huh?! huh?! where am i?!!"... sigh...-0

my thoughts, I think dice go in streaks... not random, they even out, yes. But they start good, go bad vice versa... don't try to change my mind and say they are random cuz I don't believe it.... I believe they even out in alternating streaks....

Top Dog wrote:my thoughts, I think dice go in streaks... not random, they even out, yes. But they start good, go bad vice versa... don't try to change my mind and say they are random cuz I don't believe it.... I believe they even out in alternating streaks....

go talk to the guys at random.org... can we move this now?(*,) -0

I went back and looked at the setup for the dice. Here's what the site Faq says:

The dice are based on high quality random numbers from Random.org. The numbers are read from a large file containing columns of numbers from 1 to 6, in the format A1 A2 A3 D1 D2. When the dice are rolled, the game engine reads a line from the file and discards it. The appropriate numbers are used and the others are ignored. The file contains 500,000 lines of dice rolls and is re-loaded when all the lines are used up. As of November 2007 we consume 850,000 lines of dice rolls per day.

They are using 2,500,000 fixed random numbers. The outcome of the rolls is predetermined in every game. That makes a lot of sense, as I've sometimes found that taking a few minutes to wait out the game engine seems to change my rolls if I'm in one of the troughs to the peaks... i.e., let someone else read through the crappy rolls where the defender is predetermined to win.

If you think about it, this really is a crummy way to simulate rolling dice. In fact, what they've done is superimpose a pattern of 3-2-3-2-3-2-3-2 onto a quasi-random file that contains a random walk of measured radio disturbances. Given that it's a random walk, even if you've got just a slightly higher probability that the next digit is going to land on the same value as the previous digit, when you use fixed chains in this way, the longer chain will have a higher likelihood to be bunched, meaning win-streaks and lose-streaks. So it's exactly what I've suspected based on personal experience, and what top tog states in more qualitative terms... the dice will go in streaks.

Let me give you guys a clue as to how to do it right... take a uniformly distributed irrational number, like pi (last I checked, there are like a trillion digits calculated, and use the A-A-A-D-D framework on that. Pi may not be uniformly distributed (I'd have to check on that), but it's not a random walk. Alternatively, if the folks in charge are just too attached to random.org to revisit this issue, at a minimum, the file that is generated by random.org should be rehashed so that you read in the dice A-D-A-D-A rather than A-A-A-D-D...

But anyway, I'd still like some more posts as to how often people find the dice "even", or "even up" vs. "totally lopsided"...

Excellent Idea.

Yeah, anyone can complain about the dice, but there is some validity here.
I play a lot of speed and you see the same drop pattern in certain maps, and I've found that if you start by attacking country "X" right off the hop, ( usually Argentina, or Indonesia ) You loose. Almost evey time. I feel it's more patterns then true randomness.

I don't think the dice are random enough. In the end it's fair because we are all playing under the same circumstances but I would like to see Lack consider widening the process.

And Owen: If the consistant "there's Somthing Wrong With The Dice" posts bother you so much, then why do you feel compelled to retort With your typical nonsense and rambling drivel? These complaints are ongoing, and gdeangel has a pretty good idea on how it might be addressed. Perhaps with a few dozen members consistantly bitching, we might examine the approach that there COULD BE A GOD DAMN PROBLEM.

Johnny Rocket Fingers

In the most annoying moments (for example when I am about to get struck out) my opponents get incredible dice -_- Quite often...

gdeangel wrote:I lose 15-5, so other guys +6 bonus holds. I lose 9-4. +9 bonus holds. I attack with every single attacking army I have and get 4 territories. Meanwhile other guy is rolling me 3-3 losing nothing. Which inspired me to start this thread.

The question is not, are the dice random... but how often do you find that one player seems to have particularly good dice and the other player seems to crap out every time? And I mean regardless of whether your getting the good luck or the bad luck (as I'll admit I'm often on the good side as well as the bad side), who often do you feel like the dice are one sided... maybe on a scale of "A" being perfectly even (regardless of good or bad), "B" being somewhat lopsided, and "F" being all one sided good dice... other guy/you get all bad rolls.

I'd say in 50% of my games the dice rolls even out. 40% are somewhere in the B range... mostly evens out either good or bad, and then there are about 10% of games that, like the one above, just fail ... they earn the big fat F ... there is no semblance of balanced dice. Anyone else care to share their experience...

I think this is what you would tend to expect. 50% of the time the dice are fairly even, 40% of the time the dice noticeably favor one side or the other, and the other 10% of the time, it's hopeless for one side or the other. That seems to suggest a fairly normal distribution ("normal" in the Gaussian sense) of overall rolls.

I've played about 70 games, almost all teams, I remember one insanely good streak by the opponent (started with 16 on 2 territories and took about 11 territories), 1 insanely bad streak by us, and two insanely good runs by my team. This is in about 2100 turns.

There are games where you just can't win, but these are typically the ones where you lose about 10 3v2s in rounds 1/2 with no compensation. But there should be a similar number of games, playing the same strategy, where you win a bunch of 3v2s early and quickly build up an insurmountable advantage.

its a pleasure to actually see something constructive in a dice complaining thread. thanks to that poster!

i do think tho that this pic is something to look at:

That's a fair point Scarecrow, but a couple of things to consider:

1) You can have an equal number of each number come up on the dice, but if they come in chains, it wouldn't show up in the top panel of your dice analyzer.

2) The place it might show up is in the bottom panel, but you don't have the same display as I do apparently... my dice analyzer shows one column with my outcome %'s for a particular match-up and then the statistical ideal percentage.

3) You also don't have a ton of data in your dice analyzer... I have maybe 1.5x your data in my dice analyzer (I don't use firefox that much), and, for example, my R-R-W-W rolls have turned out a whopping 11 perentage points fewer "lose 2" rolls than the ideal. But, admittedly, there's not much data there even in my case.

4) Even if you had ideal conflict outcomes, it would not contradict the notion that the wins and loses come in streaks. I think Bean is right to say that we're looking at something that may as a whole appears to be distributed about what you'd expect... the question is whether the "lumps" or "peaks and valleys" have been distorted in some way that would not occur by randomly rolling actual dice.

Lest it just sound like I started this thread due to bad luck... I can think of one game not long ago where I literally lost a total of 10 armies attacking the whole game while taking out 50+ opponent units and at the same time he took about 6 units on the offensive. It was 1v1 on Mongol. It seems like there are "grooves" like this in the data file and once your game gets slotted into one of those grooves, the only thing you can do is maybe try to walk away and let the engine eat up those lines of dice for some other players game...

gdeangel wrote:2) The place it might show up is in the bottom panel, but you don't have the same display as I do apparently... my dice analyzer shows one column with my outcome %'s for a particular match-up and then the statistical ideal percentage.

Firstly, the ideal case is the statistical expected value i.e the same for everyone and regrdless of sample size and all such things. It is set at about 39/33/29 percent for a 3v2 battle.

Secondly, he does have that stat there. It is just shown as percentage off (in the brackets) instead of in separate columns (where did he make that change? I like it, I need it... ).

gdeangel wrote:I went back and looked at the setup for the dice. Here's what the site Faq says:

The dice are based on high quality random numbers from Random.org. The numbers are read from a large file containing columns of numbers from 1 to 6, in the format A1 A2 A3 D1 D2. When the dice are rolled, the game engine reads a line from the file and discards it. The appropriate numbers are used and the others are ignored. The file contains 500,000 lines of dice rolls and is re-loaded when all the lines are used up. As of November 2007 we consume 850,000 lines of dice rolls per day.

They are using 2,500,000 fixed random numbers. The outcome of the rolls is predetermined in every game. That makes a lot of sense, as I've sometimes found that taking a few minutes to wait out the game engine seems to change my rolls if I'm in one of the troughs to the peaks... i.e., let someone else read through the crappy rolls where the defender is predetermined to win.

If you think about it, this really is a crummy way to simulate rolling dice. In fact, what they've done is superimpose a pattern of 3-2-3-2-3-2-3-2 onto a quasi-random file that contains a random walk of measured radio disturbances. Given that it's a random walk, even if you've got just a slightly higher probability that the next digit is going to land on the same value as the previous digit, when you use fixed chains in this way, the longer chain will have a higher likelihood to be bunched, meaning win-streaks and lose-streaks. So it's exactly what I've suspected based on personal experience, and what top tog states in more qualitative terms... the dice will go in streaks.

Let me give you guys a clue as to how to do it right... take a uniformly distributed irrational number, like pi (last I checked, there are like a trillion digits calculated, and use the A-A-A-D-D framework on that. Pi may not be uniformly distributed (I'd have to check on that), but it's not a random walk. Alternatively, if the folks in charge are just too attached to random.org to revisit this issue, at a minimum, the file that is generated by random.org should be rehashed so that you read in the dice A-D-A-D-A rather than A-A-A-D-D...

But anyway, I'd still like some more posts as to how often people find the dice "even", or "even up" vs. "totally lopsided"...

How would using pi make it more random? The sequences, as the number goes on, would have the EXACT same properties as the random numbers brought in. And, also, the fact that a sequence will come up more the more numbers have come in, is in itself a fallacy. If you center on one section, the numbers will appear randomly spread out. Taken as a whole, you have sections that will repeat, but you have that in normal sections, too.

It's random numbers, and regardless of complaints, it IS random. Pi is not random. These numbers, regardless that they are pulled from a file, are randomly produced, and just saved. Pi never changes, these numbers are different each and every time.

You also have to consider odds. These numbers are produced for many, many sites. And even if you only take this one site into consideration, many streaks are going to happen for people every day. The way the dice are read, it is reading multiple turns per second, so it's not just going against you in a row. It's also reading others turns, at the same time.

Plutoman wrote:How would using pi make it more random? The sequences, as the number goes on, would have the EXACT same properties as the random numbers brought in. And, also, the fact that a sequence will come up more the more numbers have come in, is in itself a fallacy. If you center on one section, the numbers will appear randomly spread out. Taken as a whole, you have sections that will repeat, but you have that in normal sections, too.

That's not really true. If you look at how random.org generates random numbers, it is measuring some type of natural phenomenon and running it through some type of multi-variable function. However, as a general rule regarding a continous function that has an equal probability that the next value will be either higher or lower, the most likely next value is the current value. I suspect, but don't know for sure, that something like that is going on with the numbers generated by random.org. Now I know there are natural phenomenon that don't follow a continous path, and so it might be possible that they are transforming their measurements to correct for it, but that would involve something like application of principles from quantum mechanics (and even there, you have people going back to a continous "string" model to explain what's going on...) I don't know for sure, but I'm skeptical. I'd say there's a reasonably good chance that random.org could just as well fill a file using a time-series of the fractional share price of Google stock as what they are doing with radio waves....

And I think we have a different view of what CC is doing with it's random.org file. To me, it sounds like they have one file of dice lines, and once they get to the end, they start back over from the beginning of the file. That doesn't mean you'll get the same dice every game, because they read out of the file sequentially for every game that is being played... today you might play a game using lines 379-450. Tomorrow, you might play using lines 110,000-110,089. And for the game that you play once a day, your going to be all over the map. That would be just like taking a fixed irrational number. Yes there will be repeating chains in that as well, but they will repeat randomly.

It's random numbers, and regardless of complaints, it IS random. Pi is not random. These numbers, regardless that they are pulled from a file, are randomly produced, and just saved. Pi never changes, these numbers are different each and every time.

The only reason the number are differnt every time is because you are playing your game at a different point in the file, as I read it. I could be wrong here though.

You also have to consider odds. These numbers are produced for many, many sites. And even if you only take this one site into consideration, many streaks are going to happen for people every day. The way the dice are read, it is reading multiple turns per second, so it's not just going against you in a row. It's also reading others turns, at the same time.

Yes, this is a saving grace. But the more I've played, the more I question whether there is actually that type of volume on the site. 850,000 lines of dice per day is about 11 lines per second. Yes, that sounds fast, but there are probably times where that number drops significantly.

Going back to one thing you said:

And, also, the fact that a sequence will come up more the more numbers have come in, is in itself a fallacy. If you center on one section, the numbers will appear randomly spread out.

If I tell you that the next value in a series has an equally likely chance to be higher or lower than the current value, your best guess for the next number is the number itself. Based on the range of possible values you are looking at, this might be just a nominal blip on the probability distribution, but if I chop that up into series of 3 and 2 consequtive measurements, you would expect to have more repeats in the "3 series" than the "2 series than would be produced by a completely random string. There should be a way to test the data file for this... I'l have to think about it a bit....

The statistical ideal would be to have 6 out of 216 rolls be a tripple repeat. And to have 6 out of 36 be a doubt repeat. You could take a binary function of the absolute value of the difference of all A digits in every row over the number of rows, and a binary function of the absolute value of the difference of all D digits in every row over the number of rows, and compare them to the respective ideals... I think that would settle the question. So whose got the dice file...

Plutoman wrote:It's random numbers, and regardless of complaints, it IS random. Pi is not random. These numbers, regardless that they are pulled from a file, are randomly produced, and just saved. Pi never changes, these numbers are different each and every time.

No no no. The numbers aren't random, it's per definition impossible for an algorithm to make up random numbers. They are, to use a better word, highly unpredictable (using the same argument as KLOBBER, kind of makes me feel dirty *shudder* ). That's still good enough for our purposes but in an ideal world I'd prefer using a generator that makes a new number when needed instead of using a pre-determined one.

Thezzaruz wrote:

Plutoman wrote:It's random numbers, and regardless of complaints, it IS random. Pi is not random. These numbers, regardless that they are pulled from a file, are randomly produced, and just saved. Pi never changes, these numbers are different each and every time.

No no no. The numbers aren't random, it's per definition impossible for an algorithm to make up random numbers. They are, to use a better word, highly unpredictable (using the same argument as KLOBBER, kind of makes me feel dirty *shudder* ). That's still good enough for our purposes but in an ideal world I'd prefer using a generator that makes a new number when needed instead of using a pre-determined one.

Pi is not random -- believe I read somewhere that an algorithm to produce random numbers cannot be shorter than the numbers it produces. By that definition, pi clearly fails.

Atmospheric noise may or may not be random, but at least it is nonalgorithmic (at least as best we can tell). There is obviously an algorithm used in converting the atmospheric noise to usable random bits, but the algorithm itself is not used to generate the randomness.

The dice file, last I heard, is reused, so that is a deviation from randomness, even if the numbers were random the first time through.

gdeangel wrote:

Plutoman wrote:How would using pi make it more random? The sequences, as the number goes on, would have the EXACT same properties as the random numbers brought in. And, also, the fact that a sequence will come up more the more numbers have come in, is in itself a fallacy. If you center on one section, the numbers will appear randomly spread out. Taken as a whole, you have sections that will repeat, but you have that in normal sections, too.

That's not really true. If you look at how random.org generates random numbers, it is measuring some type of natural phenomenon and running it through some type of multi-variable function. However, as a general rule regarding a continous function that has an equal probability that the next value will be either higher or lower, the most likely next value is the current value. I suspect, but don't know for sure, that something like that is going on with the numbers generated by random.org. Now I know there are natural phenomenon that don't follow a continous path, and so it might be possible that they are transforming their measurements to correct for it, but that would involve something like application of principles from quantum mechanics (and even there, you have people going back to a continous "string" model to explain what's going on...) I don't know for sure, but I'm skeptical. I'd say there's a reasonably good chance that random.org could just as well fill a file using a time-series of the fractional share price of Google stock as what they are doing with radio waves....

And I think we have a different view of what CC is doing with it's random.org file. To me, it sounds like they have one file of dice lines, and once they get to the end, they start back over from the beginning of the file. That doesn't mean you'll get the same dice every game, because they read out of the file sequentially for every game that is being played... today you might play a game using lines 379-450. Tomorrow, you might play using lines 110,000-110,089. And for the game that you play once a day, your going to be all over the map. That would be just like taking a fixed irrational number. Yes there will be repeating chains in that as well, but they will repeat randomly.

It's random numbers, and regardless of complaints, it IS random. Pi is not random. These numbers, regardless that they are pulled from a file, are randomly produced, and just saved. Pi never changes, these numbers are different each and every time.

The only reason the number are differnt every time is because you are playing your game at a different point in the file, as I read it. I could be wrong here though.

You also have to consider odds. These numbers are produced for many, many sites. And even if you only take this one site into consideration, many streaks are going to happen for people every day. The way the dice are read, it is reading multiple turns per second, so it's not just going against you in a row. It's also reading others turns, at the same time.

Yes, this is a saving grace. But the more I've played, the more I question whether there is actually that type of volume on the site. 850,000 lines of dice per day is about 11 lines per second. Yes, that sounds fast, but there are probably times where that number drops significantly.

Going back to one thing you said:

And, also, the fact that a sequence will come up more the more numbers have come in, is in itself a fallacy. If you center on one section, the numbers will appear randomly spread out.

If I tell you that the next value in a series has an equally likely chance to be higher or lower than the current value, your best guess for the next number is the number itself. Based on the range of possible values you are looking at, this might be just a nominal blip on the probability distribution, but if I chop that up into series of 3 and 2 consequtive measurements, you would expect to have more repeats in the "3 series" than the "2 series than would be produced by a completely random string. There should be a way to test the data file for this... I'l have to think about it a bit....

The statistical ideal would be to have 6 out of 216 rolls be a tripple repeat. And to have 6 out of 36 be a doubt repeat. You could take a binary function of the absolute value of the difference of all A digits in every row over the number of rows, and a binary function of the absolute value of the difference of all D digits in every row over the number of rows, and compare them to the respective ideals... I think that would settle the question. So whose got the dice file...

Interesting thoughts. I seem to recall from random.org that they use a skew correction mechanism, but I guess it's not perfect -- see http://www.random.org/statistics/source-purity/

For example, assuming 2-bit data, if you get 3,3,3,2,2,2,1,1,1,0,0,0, the results will be 11 11 11 10 10 10 01 01 01 00 00 00, which post-skew correction is still 111000. (I guess this means they shouldn't use 2-bit data and a 2-bit correction mechanism.) Assuming 3-bit, and something like 7,7,7,6,6,6,5,5,5,4,4,4, you would get:

111 111 111 110 110 110 101 101 101 100 100 100 or (in 2-digit pairs)

11 11 11 11 11 10 11 01 10 10 11 01 10 11 00 10 01 00

Application of the antiskew algorithm gives you:
10110110
which seems not inconsistent with a reasonable result.

You should contact Twill. He has the dice files and can run the tests on the output data.

the dice are not random, the dice suck, the dice are crappy, high rankers have perfect dice, my opponents roll only sixes, i lost 7v1, bla bla bla. so what? the dice are the same for everybody. sometimes you lose sometimes you win. i lost a 153vs72 auto attack. it ended 3vs48. streaky dice? of course. i looked at the rolls and i saw several times where the defender rolled double sixes. the most were 6 double sixes in a row. what was even stranger is that at some point in the rolls there was the same line of dice repeated 4 times. i don't remember what exactly it was but it was kinda like this: 1-3-4 vs 1-6. the same line 4 times in a row. tell me what's the chance of that happening? do i care if it's a bug? no i don't because lack will never change how the dice work here. why? because he can simply say: "the dice are the same for everybody" and he would be right.
in a recent game i went 11 vs 23 in a desperate move. i lost just 3 and killed all. and just 5 minutes before i lost a 7v1 and a 9v1.

i learned one thing in 600+ games. dice can kill you dice can save you but in time skill will prevail.