Which number do you roll the most?

[degaston]

Yeah, well the program has seriously taken my thread to heart..wow in 25 rolls in defense an average of 1.4..now what's going on

Confirms to me that my 1 v 1 days are over. I now only play team games in the future

Here's why:

[Agent 86]

It continues in attack, my last 25 rolls. At this pace I will have the worst dice percentage of all CC players

Assault -10% 3.26 41 27 31 33 18 29

[Agent 86]

I'll look to see if there's a technical reason that 1 is arbitrarily lower than the rest, but I performed a Chi-Square goodness of fit test for each of the 5 players rolls and that number of rolls isn't statistically significant. The p-value was essentially 0 each time...

For non stats people: Each set is, indivdually, within the margin of error to not cause alarm within the realm of randomness, but since the dice are read from a finite sheet randomly, perhaps there are a fewer number of ones than the rest.

BMO

Hey don't just fix the number for my account..make it universal please

[DoomYoshi]

Yeah, well the program has seriously taken my thread to heart..wow in 25 rolls in defense an average of 1.4..now what's going on

Confirms to me that my 1 v 1 days are over. I now only play team games in the future

Here's why:

I always pictured Lack working on a 8080 with CP/M...

[Agent 86]

Yeah well Lack is gone..so who has is now programing? I'm still rolling more 1's..nobody else is

[frankiebee]

Has this been solved? I still think there is proof in this thread that the dice are not as random as we thought.

[BoganGod]

Has this been solved? I still think there is proof in this thread that the dice are not as random as we thought.

Mate, just re read that excellent klobber blog spot you put me onto. Discusses at some length in logical detail, the difference between random, and predictable. Well worth a second read

[grifftron]

I roll 5's... my opponents roll 1's ... join my games and see.

[Metsfanmax]

Here are the dice stats for the five players with the most completed games:



It looks like 1's are showing up about .26% less often than they should. It's not a huge difference, but it definitely looks like a pattern. From what I've seen, it's the same for every player with a large sample size, so there's probably no "favoritism" going on, but I think that this would make the battle outcomes slightly different from the expected win/loss percentages.

I didn't do the math, but logically it's as if the players are using 5-sided dice .26% of the time. Having fewer numbers to chose from will slightly increase the number of ties, which would favor the defender.

I think this should be fixed ASAP.

Something to note here. The dice algorithm changed substantially in 2010, so if you analyze the stats for players who have been here for many years, they could be biased as a result of a poor randomness prior to 2010. The same logic also applies to the way random.org is generating their numbers (it could be different now from how it previously was). In order to prove in a compelling manner that something is currently awry, you need to compile these results only for people who have joined relatively recently (and you'll need a decently large sample size).

[ManBungalow]

Something to note here. The dice algorithm changed substantially in 2010, so if you analyze the stats for players who have been here for many years, they could be biased as a result of a poor randomness prior to 2010. The same logic also applies to the way random.org is generating their numbers (it could be different now from how it previously was). In order to prove in a compelling manner that something is currently awry, you need to compile these results only for people who have joined relatively recently (and you'll need a decently large sample size).

Actually, no:
I think I'm right in saying that the system started actually started recording dice stats in 2011.

May 2011 announcement: viewtopic.php?f=634&t=145904#p3183301
Lack started storing results to use for dice stats a few weeks/months before the announcement was made...but still 2011.

[waltero]

2 for me

[Agent 86]

I'm still feeling the atmospheric noise from Random.org Seems the dice are still not rolling ones except for me, hey I'm premium and pay for the better dice, I wait in patience. Soon I hope everyone else will roll the 1's with me

Still having fun trying to win against the dice gods.

86

[frankiebee]

Has this been solved? I still think there is proof in this thread that the dice are not as random as we thought.

Mate, just re read that excellent klobber blog spot you put me onto. Discusses at some length in logical detail, the difference between random, and predictable. Well worth a second read

I know it by heart, this is probabbly what klobber would have said: the dice is made to be unpredictable, which it is. But with 100% designed unpredictable dice, you would still expect to see roughly 1 in 6 players to have rolled mostly 1's. 1 in 50 would still be believable, but seriously, can you link me 1 profile of a player that played over 500 games who has thrown mostly 1's? Because if there iss'nt one player at all... there is something wrong is the adjustments of the dice.

[Metsfanmax]

Has this been solved? I still think there is proof in this thread that the dice are not as random as we thought.

Mate, just re read that excellent klobber blog spot you put me onto. Discusses at some length in logical detail, the difference between random, and predictable. Well worth a second read

I know it by heart, this is probabbly what klobber would have said: the dice is made to be unpredictable, which it is. But with 100% designed unpredictable dice, you would still expect to see roughly 1 in 6 players to have rolled mostly 1's. 1 in 50 would still be believable, but seriously, can you link me 1 profile of a player that played over 500 games who has thrown mostly 1's? Because if there iss'nt one player at all... there is something wrong is the adjustments of the dice.

I don't think we can look into this quite yet because there appear to be some issues with the dice records. I see several players who have played many games whose dice stats pages have only a very small number of stats recorded, e.g.

http://www.conquerclub.com/player.php?m ... _id=133381
http://www.conquerclub.com/player.php?m ... r_id=43426

[ManBungalow]

I don't think we can look into this quite yet because there appear to be some issues with the dice records. I see several players who have played many games whose dice stats pages have only a very small number of stats recorded, e.g.

http://www.conquerclub.com/player.php?m ... _id=133381
http://www.conquerclub.com/player.php?m ... r_id=43426

Please refer to my response to your post on the previous page:

viewtopic.php?f=6&t=194323&start=45#p4300468

[frankiebee]

Has this been solved? I still think there is proof in this thread that the dice are not as random as we thought.

Mate, just re read that excellent klobber blog spot you put me onto. Discusses at some length in logical detail, the difference between random, and predictable. Well worth a second read

I know it by heart, this is probabbly what klobber would have said: the dice is made to be unpredictable, which it is. But with 100% designed unpredictable dice, you would still expect to see roughly 1 in 6 players to have rolled mostly 1's. 1 in 50 would still be believable, but seriously, can you link me 1 profile of a player that played over 500 games who has thrown mostly 1's? Because if there iss'nt one player at all... there is something wrong is the adjustments of the dice.

I don't think we can look into this quite yet because there appear to be some issues with the dice records. I see several players who have played many games whose dice stats pages have only a very small number of stats recorded, e.g.

http://www.conquerclub.com/player.php?m ... _id=133381
http://www.conquerclub.com/player.php?m ... r_id=43426

Those two players have been inactive for a looooooooong time. They played most of their games when the dice stats weren't around, and now played a couple of games where their dice were recorded.

[degaston]

Has this been solved? I still think there is proof in this thread that the dice are not as random as we thought.

Mate, just re read that excellent klobber blog spot you put me onto. Discusses at some length in logical detail, the difference between random, and predictable. Well worth a second read

Could you share the blog post you're referring to? I don't know if the dice are predictable, but when everyone is rolling fewer 1's than they should, they are certainly not random.

[Metsfanmax]

Has this been solved? I still think there is proof in this thread that the dice are not as random as we thought.

Mate, just re read that excellent klobber blog spot you put me onto. Discusses at some length in logical detail, the difference between random, and predictable. Well worth a second read

I know it by heart, this is probabbly what klobber would have said: the dice is made to be unpredictable, which it is. But with 100% designed unpredictable dice, you would still expect to see roughly 1 in 6 players to have rolled mostly 1's. 1 in 50 would still be believable, but seriously, can you link me 1 profile of a player that played over 500 games who has thrown mostly 1's? Because if there iss'nt one player at all... there is something wrong is the adjustments of the dice.

I don't think we can look into this quite yet because there appear to be some issues with the dice records. I see several players who have played many games whose dice stats pages have only a very small number of stats recorded, e.g.

http://www.conquerclub.com/player.php?m ... _id=133381
http://www.conquerclub.com/player.php?m ... r_id=43426

Those two players have been inactive for a looooooooong time. They played most of their games when the dice stats weren't around, and now played a couple of games where their dice were recorded.

Hm. Interesting. I assumed that because they were on the scoreboard they would have been actively playing games recently, but I suppose I didn't confirm.

Anyway, as I said, we can't really confirm this as a real problem unless we have stats for people who have only been playing recently (last three months, or maybe six; we can't really know whether the problem currently exists if we include all rolls since dice stats were originally recorded). Also, we'd need to see a significant number of people more than two standard deviations away from the mean.

In other words, be aware of selection bias. If you select only players with a lot of games played, you might be polluting your sample with people who did many of their rolls during a different time on CC. Another example of selection bias: one possibility that springs to mind is that one factor in why the players at the top of the scoreboard are at the top may be because they are the ones who get fewer ones, due to random chance. This may not be a huge effect, but it can influence the results. I looked at people with low scores and lots of games played, though, and didn't see anyone with 1's the most. In fact, I still haven't found anyone that satisfies that. There's definitely something interesting here.

[BoganGod]

+1 on meta's last post.

[degaston]

...
Hm. Interesting. I assumed that because they were on the scoreboard they would have been actively playing games recently, but I suppose I didn't confirm.

Anyway, as I said, we can't really confirm this as a real problem unless we have stats for people who have only been playing recently (last three months, or maybe six; we can't really know whether the problem currently exists if we include all rolls since dice stats were originally recorded). Also, we'd need to see a significant number of people more than two standard deviations away from the mean.

In other words, be aware of selection bias. If you select only players with a lot of games played, you might be polluting your sample with people who did many of their rolls during a different time on CC. Another example of selection bias: one possibility that springs to mind is that one factor in why the players at the top of the scoreboard are at the top may be because they are the ones who get fewer ones, due to random chance. This may not be a huge effect, but it can influence the results. I looked at people with low scores and lots of games played, though, and didn't see anyone with 1's the most. In fact, I still haven't found anyone that satisfies that. There's definitely something interesting here.

I collected the dice stats for 10 players who have played 100 games or less, and joined after 2012. (I started at 100 games and worked my way down.) Here's how their rolls compared to the expected 16.66% for each number:


I'm no statistician, and feel free to correct me if any of this is wrong, but from doing a little research, I found that a chi-squared test is used to tell if sample data is consistent with a particular theoretical distribution. A value of less than .05 indicates that there's less than a 5% chance that the sample data would occur "naturally". So here's the spreadsheet:


That .0000035 in the 1's column shows that it is almost impossible to get these results from truly random dice. (The 2's and 4's results are also pretty unlikely)

[ManBungalow]

Now that is a pretty graph.

[Agent 86]

The truth is that Random.org is false and just taking every websites money that uses them. Atmospheric noise is a big joke

How many users have left because of this on this website alone?

[Metsfanmax]

I'm no statistician, and feel free to correct me if any of this is wrong, but from doing a little research, I found that a chi-squared test is used to tell if sample data is consistent with a particular theoretical distribution. A value of less than .05 indicates that there's less than a 5% chance that the sample data would occur "naturally". So here's the spreadsheet:
...
That .0000035 in the 1's column shows that it is almost impossible to get these results from truly random dice. (The 2's and 4's results are also pretty unlikely)

Please be a little more specific about the calculation you did. The null hypothesis is that the dice are fair, so the expected frequency of ones for each user is 1/6 of their total number of rolls. You turn the observed number of ones/twos/threes/etc. into an observed frequency by dividing by the number of rolls for each user (e.g. the upper left cell would be 0.163). Once you calculate the chi-squared value (the sum of the squared deviations from that expected frequency, divided by the expected frequency), you then turn it into a p-value, given N = 10 degrees of freedom. Is that number in your final row, the p-value for each number?

[degaston]

Please be a little more specific about the calculation you did. The null hypothesis is that the dice are fair, so the expected frequency of ones for each user is 1/6 of their total number of rolls. You turn the observed number of ones/twos/threes/etc. into an observed frequency by dividing by the number of rolls for each user (e.g. the upper left cell would be 0.163). Once you calculate the chi-squared value (the sum of the squared deviations from that expected frequency, divided by the expected frequency), you then turn it into a p-value, given N = 10 degrees of freedom. Is that number in your final row, the p-value for each number?

The last column is the expected number of rolls for each die face (1/6 of the total rolls). I used the Excel CHISQ.TEST function on each column of actual rolls with the last column as the expected range. I believe the output of the function is the p-value, and would assume that is uses the range size to determine the degrees of freedom.

I think it would be incorrect to perform the calculation as you described, because that would imply that the percentage that each value appeared was the actual sample size, and from the wiki page in your first link:

Assumptions
* Sample size (whole table) – A sample with a sufficiently large size is assumed. If a chi squared test is conducted on a sample with a smaller size, then the chi squared test will yield an inaccurate inference. The researcher, by using chi squared test on small samples, might end up committing a Type II error.

Whatever a "Type II" error is?

[degaston]

The truth is that Random.org is false and just taking every websites money that uses them. Atmospheric noise is a big joke

How many users have left because of this on this website alone?

As I said before, I don't think the dice are off enough to really make much of a difference in the outcome of a game - though it's possible that one changed outcome that breaks or fails to break a bonus could have a butterfly effect.

I don't even know for sure if this is to the attacker's, or defender's advantage. The defender would get an advantage from there being more ties, but the attacker rolls more dice, so they would get this advantage 33% more often than the defender.

I would think that there should be plenty of random user events occurring on the site that could be used to seed a pseudo-random number generator, and avoid having to pay anyone for random numbers. It's not the NSA!