Conquer Club

A solution to the "dice problem"

Talk about all things related to Conquer Club

Moderator: Community Team

Forum rules
Please read the community guidelines before posting.

A solution to the "dice problem"

Postby Tennisie on Wed Aug 11, 2010 11:53 am

I've read hundreds of complaints about "rigged" dice and "streaks" of bad dice. If you have any doubt about the randomness of the dice, check out random.org. I'm familiar with random number generation through my profession and clearly, random.org has done a good job of ensuring the numbers obtained from its site are random and smoothly distributed. Of course, that doesn't eliminate the possibility of dice manipulation by conquerclub.com's software, but that's a subject for a different thread.

Assuming there's no manipulation by conquerclub.com, since the dice are truly random, one would expect streaks to occur naturally. The most extreme example I know of occurred to me when I lost 87 to 0 during a single auto-attack. The chances of that happening involve 20 digits of unlikelihood, BUT IT HAPPENED. And this leads to the biggest source of complaints on conquerclub.com: streaks of bad dice.

The purpose of this thread is not to complain, but to suggest ways to mitigate these streaks. The first attempt to mitigate this problem was made when the City Mogul map was created. This map facilitates many attackers against many defenders which produces a large number of dice rolls and reduces the effect of streaks. It works, but limiting this benefit to one map is too restrictive. A better solution is one that allows all maps to experience the same benefit. One method to do this is to have multiple rolls of dice for each single attack attempt. For example, if you attack (not auto-attack) with a single roll of dice, the site could automatically roll X times and the most frequent winner would win the single attack attempt. This would reduce the effect of streaks, and the larger X is, the more the streaks would be reduced. This would be an easy solution to implement if every roll was one attacker die versus one defender die: the software would simply roll one vs one X times and the most frequent winner would win the attack. However, since attacks are usually three attackers vs two defenders, the math becomes much more complicated.

OK, mathematicians and brainstormers, it's your turn...
Last edited by Tennisie on Wed Aug 11, 2010 6:09 pm, edited 1 time in total.
You can't escape Cosmic Cat į•™(⇀‸↼‶)į•—

dwilhelmi wrote:Well dangit-drat-dagnabbit
User avatar
Major Tennisie
 
Posts: 120
Joined: Sat Oct 13, 2007 10:50 pm
Location: U.S.

Re: A solution to the "dice problem"

Postby donkeymile on Wed Aug 11, 2010 12:31 pm

I love the idea - mostly because I am so pissed off about the ridiculous nature of the streaks. It has caused me to decide to walk away from this site for at least a while because its so tiresome dealing with month long 'bad dice' streaks, and ranking dropping out from beneath you, and usually having nothing to do with gameplay.

Just sick of it and that's it - done with my short rant. Sick of the bullshit. The 'good' or at least 'fair' dice streaks NEVER last that long consistently the way the bad ones do .....
User avatar
Captain donkeymile
 
Posts: 228
Joined: Fri Jan 23, 2009 8:10 pm

Re: A solution to the "dice problem"

Postby Bones2484 on Wed Aug 11, 2010 12:39 pm

Tennisie wrote:The most extreme example I know of occurred to me when I lost 87 to 0 during a single auto-attack. The chances of that happening involve 20 digits of unlikelihood, BUT IT HAPPENED.


Pics or it didn't happen.

As for the rest of your "suggestion", can you imagine the strain on this? It would take forever to auto assault in certain situations.

edit: And honestly? This suggestion would make attacking LESS random as everything would default closer to expected results.
User avatar
Major Bones2484
 
Posts: 2307
Joined: Mon Sep 17, 2007 11:24 am
Location: Los Angeles, CA (G1)

Re: A solution to the "dice problem"

Postby lord voldemort on Wed Aug 11, 2010 1:37 pm

i didnt think there was a dice problem.
you get good dice and you get bad dice....people just get angry when their dice fail...they rarely remember when they go well
Image
User avatar
Lieutenant lord voldemort
 
Posts: 9596
Joined: Sat Oct 20, 2007 4:39 am
Location: Launceston, Australia

Re: A solution to the "dice problem"

Postby natty dread on Wed Aug 11, 2010 2:14 pm

It's a game. A game that involves luck. Get over it or play chess. Yawn.
Image
User avatar
Sergeant 1st Class natty dread
 
Posts: 12877
Joined: Fri Feb 08, 2008 8:58 pm
Location: just plain fucked

Re: A solution to the "dice problem"

Postby ParadiceCity9 on Wed Aug 11, 2010 3:03 pm

natty_dread wrote:It's a game. A game that involves luck. Get over it or play chess. Yawn.


Precisely put, good sir.
Corporal 1st Class ParadiceCity9
 
Posts: 4239
Joined: Thu Feb 15, 2007 4:10 pm

Re: A solution to the "dice problem"

Postby jaimito101 on Wed Aug 11, 2010 3:33 pm

also part of the charm is getting to see which dice are thrown. How would your plan present the result and keep the sexiness the visible 3vs 2 die bring along with it?
Colonel jaimito101
 
Posts: 176
Joined: Tue Jun 30, 2009 1:36 pm

Re: A solution to the "dice problem"

Postby Tennisie on Wed Aug 11, 2010 6:18 pm

jaimito101 wrote:also part of the charm is getting to see which dice are thrown. How would your plan present the result and keep the sexiness the visible 3vs 2 die bring along with it?

Thus the request for mathematicians. It would require an algorithm, probably fairly simple, to generate the dice results we currently see. The algorithm should take three inputs: the number of rolls for each attack (X), the number of attacking dice (A) and the number of defending dice (D).
User avatar
Major Tennisie
 
Posts: 120
Joined: Sat Oct 13, 2007 10:50 pm
Location: U.S.

Re: A solution to the "dice problem"

Postby Metsfanmax on Wed Aug 11, 2010 6:30 pm

If we accept that the current system is truly random, or indistinguishable from it, then your system wouldn't change the likelihood of streaks.
User avatar
Sergeant 1st Class Metsfanmax
 
Posts: 6722
Joined: Wed Apr 11, 2007 11:01 pm

Re: A solution to the "dice problem"

Postby Tennisie on Wed Aug 11, 2010 9:25 pm

Metsfanmax wrote:If we accept that the current system is truly random, or indistinguishable from it, then your system wouldn't change the likelihood of streaks.

Yes it would, because your luck evens out the more you roll. Sometimes it can take many rolls to even out, but eventually it regresses to the average. Its just a matter of finding a large enough value for X but one that doesn't overburden the conquerclub server. This is similar to the techniques used to "smooth" the data from a true random number generator: combine enough samples together so that a localized streak is swamped by the probabilities.
User avatar
Major Tennisie
 
Posts: 120
Joined: Sat Oct 13, 2007 10:50 pm
Location: U.S.

Re: A solution to the "dice problem"

Postby Bones2484 on Thu Aug 12, 2010 12:05 am

Tennisie wrote:
Metsfanmax wrote:If we accept that the current system is truly random, or indistinguishable from it, then your system wouldn't change the likelihood of streaks.

Yes it would, because your luck evens out the more you roll. Sometimes it can take many rolls to even out, but eventually it regresses to the average. Its just a matter of finding a large enough value for X but one that doesn't overburden the conquerclub server. This is similar to the techniques used to "smooth" the data from a true random number generator: combine enough samples together so that a localized streak is swamped by the probabilities.


No. No. No.

Doing this would just give a huge advantage to the attacker. If the dice were "averaged" out, the rolls would get closer and closer to the expected results of 3 beating a 2 every time.

Take your statement here:

Tennisie wrote:the site could automatically roll X times and the most frequent winner would win the single attack attempt


Given enough rolls in your averaging, the "most frequent winner" would be the attacker as it carries the largest odds of occurrence (37% for 2-0, 34% for 1-1, and 29% for 0-2). The results would not follow the 37%-34%-29% pattern anymore. Instead, you'd see a huge rise in the 37% and a dramatic fall in the 29%.

Think of it this way. If you rolled a 3v2 one hundred times in order to see who the "most frequent winner was" to determine who wins the roll, you would expect the results to be close to the 37%-34%-29% pattern nearly every time. Meaning the result of a 3v2 roll in your suggestion would be a 2-0 victory... nearly every time.
User avatar
Major Bones2484
 
Posts: 2307
Joined: Mon Sep 17, 2007 11:24 am
Location: Los Angeles, CA (G1)

Re: A solution to the "dice problem"

Postby jaimito101 on Thu Aug 12, 2010 2:02 am

Bones2484 wrote:
Think of it this way. If you rolled a 3v2 one hundred times in order to see who the "most frequent winner was" to determine who wins the roll, you would expect the results to be close to the 37%-34%-29% pattern nearly every time. Meaning the result of a 3v2 roll in your suggestion would be a 2-0 victory... nearly every time.


hadden't thought about it that way, but i think bones is following the correct line of thought here. By trying to improve the randomness, you actually would skew it tremendously in the attackers favour in this fashion!
Colonel jaimito101
 
Posts: 176
Joined: Tue Jun 30, 2009 1:36 pm

Re: A solution to the "dice problem"

Postby Metsfanmax on Thu Aug 12, 2010 5:49 am

Tennisie wrote:
Metsfanmax wrote:If we accept that the current system is truly random, or indistinguishable from it, then your system wouldn't change the likelihood of streaks.

Yes it would, because your luck evens out the more you roll. Sometimes it can take many rolls to even out, but eventually it regresses to the average. Its just a matter of finding a large enough value for X but one that doesn't overburden the conquerclub server. This is similar to the techniques used to "smooth" the data from a true random number generator: combine enough samples together so that a localized streak is swamped by the probabilities.


The problem with your smoothing analogy is that it assumes individual results. But on CC, the individual results of a roll aren't what matters; rather, it's the combined effect of the 3 vs 2 dice. There is no "dice problem;" what people seemingly take exception to is when the luck of the rolls actually does result in a random streak. But it is actually the interplay of the 3 vs 2 that determines the winner, and not any individual die, so you can't smooth it out. In smoothing data from a TRNG, say you get 666 on your first roll of three die; then to reduce the streakiness, all you have to do is roll again enough times so that one of the numbers is not a six. This is a particularly easy thing to do. But on CC, you could get 231 and still win the round if your opponent got 21. You can't make the same argument on CC, because it is the comparison of rolls that determines the winner.
User avatar
Sergeant 1st Class Metsfanmax
 
Posts: 6722
Joined: Wed Apr 11, 2007 11:01 pm

Re: A solution to the "dice problem"

Postby Tennisie on Thu Aug 12, 2010 6:06 am

Bones2484 wrote:
Tennisie wrote:
Metsfanmax wrote:If we accept that the current system is truly random, or indistinguishable from it, then your system wouldn't change the likelihood of streaks.

Yes it would, because your luck evens out the more you roll. Sometimes it can take many rolls to even out, but eventually it regresses to the average. Its just a matter of finding a large enough value for X but one that doesn't overburden the conquerclub server. This is similar to the techniques used to "smooth" the data from a true random number generator: combine enough samples together so that a localized streak is swamped by the probabilities.


No. No. No.

Doing this would just give a huge advantage to the attacker. If the dice were "averaged" out, the rolls would get closer and closer to the expected results of 3 beating a 2 every time.

Take your statement here:

Tennisie wrote:the site could automatically roll X times and the most frequent winner would win the single attack attempt


Given enough rolls in your averaging, the "most frequent winner" would be the attacker as it carries the largest odds of occurrence (37% for 2-0, 34% for 1-1, and 29% for 0-2). The results would not follow the 37%-34%-29% pattern anymore. Instead, you'd see a huge rise in the 37% and a dramatic fall in the 29%.

Think of it this way. If you rolled a 3v2 one hundred times in order to see who the "most frequent winner was" to determine who wins the roll, you would expect the results to be close to the 37%-34%-29% pattern nearly every time. Meaning the result of a 3v2 roll in your suggestion would be a 2-0 victory... nearly every time.


Very perceptive, Bones, you are exactly right. Which is why we'd need an algorithm. The example I gave was a simple illustration using 1 on 1, but that is never the actual case. I don't know what the algorithm should be to ensure the results hew to the standard pattern, so I'm soliciting ideas.
User avatar
Major Tennisie
 
Posts: 120
Joined: Sat Oct 13, 2007 10:50 pm
Location: U.S.

Re: A solution to the "dice problem"

Postby Metsfanmax on Thu Aug 12, 2010 6:28 am

Tennisie wrote:
Bones2484 wrote:
Tennisie wrote:
Metsfanmax wrote:If we accept that the current system is truly random, or indistinguishable from it, then your system wouldn't change the likelihood of streaks.

Yes it would, because your luck evens out the more you roll. Sometimes it can take many rolls to even out, but eventually it regresses to the average. Its just a matter of finding a large enough value for X but one that doesn't overburden the conquerclub server. This is similar to the techniques used to "smooth" the data from a true random number generator: combine enough samples together so that a localized streak is swamped by the probabilities.


No. No. No.

Doing this would just give a huge advantage to the attacker. If the dice were "averaged" out, the rolls would get closer and closer to the expected results of 3 beating a 2 every time.

Take your statement here:

Tennisie wrote:the site could automatically roll X times and the most frequent winner would win the single attack attempt


Given enough rolls in your averaging, the "most frequent winner" would be the attacker as it carries the largest odds of occurrence (37% for 2-0, 34% for 1-1, and 29% for 0-2). The results would not follow the 37%-34%-29% pattern anymore. Instead, you'd see a huge rise in the 37% and a dramatic fall in the 29%.

Think of it this way. If you rolled a 3v2 one hundred times in order to see who the "most frequent winner was" to determine who wins the roll, you would expect the results to be close to the 37%-34%-29% pattern nearly every time. Meaning the result of a 3v2 roll in your suggestion would be a 2-0 victory... nearly every time.


Very perceptive, Bones, you are exactly right. Which is why we'd need an algorithm. The example I gave was a simple illustration using 1 on 1, but that is never the actual case. I don't know what the algorithm should be to ensure the results hew to the standard pattern, so I'm soliciting ideas.


What you're asking for is inherently illogical. You either want 3v2 to win 2-0 37% of the time, or you don't. If you generate an algorithm that rolls the die 10 times in a row but still somehow manages to adhere to the current distribution pattern, you've done nothing to change the likelihood of streaks.
User avatar
Sergeant 1st Class Metsfanmax
 
Posts: 6722
Joined: Wed Apr 11, 2007 11:01 pm

Re: A solution to the "dice problem"

Postby niMic on Thu Aug 12, 2010 6:29 am

Tennisie wrote:The most extreme example I know of occurred to me when I lost 87 to 0 during a single auto-attack.


I'm having a very hard time believing that.
Image
Highest score: 3772
Highest rank: 15
User avatar
General niMic
 
Posts: 1021
Joined: Tue Mar 14, 2006 2:02 pm

Re: A solution to the "dice problem"

Postby natty dread on Thu Aug 12, 2010 6:33 am

Dice, they keep rolling
butthurt people keep trolling
normal CC day.
Image
User avatar
Sergeant 1st Class natty dread
 
Posts: 12877
Joined: Fri Feb 08, 2008 8:58 pm
Location: just plain fucked

Re: A solution to the "dice problem"

Postby Tennisie on Thu Aug 12, 2010 4:14 pm

Metsfanmax wrote:
What you're asking for is inherently illogical. You either want 3v2 to win 2-0 37% of the time, or you don't. If you generate an algorithm that rolls the die 10 times in a row but still somehow manages to adhere to the current distribution pattern, you've done nothing to change the likelihood of streaks.


I just realized what the problem is: the algorithm would need to remember the previous attack result to ensure the current result hews to the standard distribution, so it may be practical only for an auto-assault, not a single assault.
User avatar
Major Tennisie
 
Posts: 120
Joined: Sat Oct 13, 2007 10:50 pm
Location: U.S.

Re: A solution to the "dice problem"

Postby JBlombier on Tue Aug 17, 2010 4:32 pm

I have no idea about maths or randomness. But having the system remember the previous attack, whatever the system does with that information, doesn't sound very random to me. Shouldn't one attack be totally apart from the other?
Image
User avatar
Major JBlombier
 
Posts: 1435
Joined: Mon Jun 04, 2007 5:47 am
Location: Gouda

Re: A solution to the "dice problem"

Postby natty dread on Tue Aug 17, 2010 6:02 pm

Tennisie wrote:I just realized what the problem is: the algorithm would need to remember the previous attack result to ensure the current result hews to the standard distribution, so it may be practical only for an auto-assault, not a single assault.


There's so many things wrong with this I don't know where to start...

Ok, let us assume it would be technically feasible to create a system that adjusts the dice rolls based on your previous rolls to ensure the dice rolls follow the "standard distribution".

So how much deviation would be allowed? ie. in how many rolls would the median have to be achieved? Every 10, every 100 rolls?

Because currently, you do get the standard distribution, if you just roll enough dice. But apparently you want it to be achieved in less dice rolls. If every single dice roll must follow this standard deviation then the dice rolls would alternate between 3 and 4. Not very exciting eh? Ok, let's say every 100 rolls or so would need to average out. This means that all you have to do is record the last 50 rolls, and you can predict your next 50 rolls. Cheaters would have a blast. It would be extremely easy to write a script that keeps track of your dice rolls and predicts your next dice rolls.

I don't know about you but I prefer my dice random.

But then, there's always those people who just can't accept that their dice are random - they make a far too convenient scapegoat, something to blame when you notice you aren't the best of the best despite what your mom always told you... ;)
Image
User avatar
Sergeant 1st Class natty dread
 
Posts: 12877
Joined: Fri Feb 08, 2008 8:58 pm
Location: just plain fucked

Re: A solution to the "dice problem"

Postby Serbia on Tue Aug 17, 2010 6:45 pm

One solution would be DON'T BREAK THE RULES, and then none of this would happen.

Also, everyone who starts a thread about intensity cubes, whether positive or negative in nature, should receive an automatic 3 day forum ban.

Bones, great analysis. =D> And I agree completely with you as well also. The dice ARE currently random, and much like plagiarism, it should just be left alone, because if we continue to mess with it, we will actually be REMOVING the randomization from the equation, nullifying it completely.

So who's with me? DICE ARE RANDOM NOW, SO DON'T TOUCH MY DICE!
CONFUSED? YOU'LL KNOW WHEN YOU'RE RIPE
saxitoxin wrote:Serbia is a RUDE DUDE
may not be a PRUDE, but he's gotta 'TUDE
might not be LEWD, but he's gonna get BOOED
RUDE
User avatar
Captain Serbia
 
Posts: 12267
Joined: Sun Jan 14, 2007 10:10 pm
Location: Detroit

Re: A solution to the "dice problem"

Postby stuart133 on Wed Aug 18, 2010 8:03 am

As a player of RL wargames (The scary ones with 1000 counters :P Dweeb alert :D) I really can't see why people think the dice aren't random, and why streaks need to be "controlled". When rolling with real dice, the same things happen, that is the point of the dice, to add randomness, and prevent the same things from happening every game.

I am sure that anyone who has played a lot of RL wargames, or any other game involving a lot of dice will agree with me. Leave the damn dice as they are .... Though if there was some way I could roll them with my own hands ... I do miss that. :P
User avatar
Private stuart133
 
Posts: 40
Joined: Fri Jun 27, 2008 4:02 pm
Location: London

Re: A solution to the "dice problem"

Postby Robinette on Wed Aug 18, 2010 10:20 pm

If all you really want to do is tighten up the results, then just drop 1 number from the dice.
So it would be 1-5...

although... this would give a slightly greater advantage to the attacker than we have now
Image
User avatar
Brigadier Robinette
 
Posts: 2944
Joined: Sat Apr 29, 2006 1:32 pm
Location: Northern California

Re: A solution to the "dice problem"

Postby AndyDufresne on Sun Aug 22, 2010 1:06 pm

I think we should make dice with only 1 number on them. Unfortunately then the defender will always win, and we'll have to change the name of Conquer Club to the Defense Division.


--Andy
User avatar
Corporal 1st Class AndyDufresne
 
Posts: 24935
Joined: Fri Mar 03, 2006 8:22 pm
Location: A Banana Palm in Zihuatanejo

Re: A solution to the "dice problem"

Postby maasman on Sun Aug 22, 2010 1:53 pm

Robinette wrote:If all you really want to do is tighten up the results, then just drop 1 number from the dice.
So it would be 1-5...

although... this would give a slightly greater advantage to the attacker than we have now

I actually think it would be the defender, because this will increase ties and the defender wins ties.
Lets say the dice only have 2 possibilities, 1 and 2. This means the possible combinations are 1-1, 1-2, 2-1, 2-2, with the defender on the left. This means that out of 4 possible combinations, the defender wins 3 of them. This is simpler than 5, but as you add numbers the attackers slowly gain an advantage and vice-versa.
Image
User avatar
Major maasman
 
Posts: 543
Joined: Wed Apr 11, 2007 6:45 pm
Location: Goose Creek, USA

Next

Return to Conquer Club Discussion

Who is online

Users browsing this forum: No registered users