dice odds?

[Cardinalsrule]

maybe someone who's got some experience in statistics can answer these questions?

What are the odds of losing all your troops when you are attacking 6 against 1?

5 against 1?

4 against 1?

 

[Shepherd]

... it would be closer to what people EXPECT from random.

I get nervous when we start talking about taking the randomness out of the random number generator so that it will feel more random, but I hear where you're going with this Robinette. Users want the dice to be both random and predictable, but you can't have both. I can predict that I will win a 6-1 attack, and when I don't it doesn't mean that the dice aren't random, it just means that the dice aren't predictable.

And you're right, we've instituted a play style called Fixed Force for those who want predictable, not random. A 3-1 attack which the attacker should win will always be won in Fixed. But in my opinion, making the game entirely predictable takes the fun out of it. It's not for everyone.

I've always thought we should just have a chimp on a web cam rolling dice 24/7.

 

[Cardinalsrule]

I EXPECT to see streaks, both good and bad, but I don't care what anyone says, when I lose THREE 6 on 1 attacks in a 2-week period, I believe the dice program has developed a bug. The odds of that happening randomly are just too high.

 

[WidowMakers]

 

[Cardinalsrule]

lol. I think that's the one in the majcom server.

 

[MajorMajor]

nice comic strip!

I certainly wouldn't be in favour of any sort of "forced randomness". couple of things to consider:

if your "random number generator monster" has any preference towards any number or kind of number (a specific number, or higher/lower numbers, or odds, or evens) - then you're strongly favouring the defender. for example: if the odds of rolling a 3 are 20% instead of 16.67%, then there's a higher chance that both the attacker and the defender will roll a 3, in which case the defender wins more than he should.

all sorts of things might lead to such a bias. for example: multiplications tend to generate more evens than odds, because odds x odds are always odds, evens x evens are always evens, and odds x evens are always evens as well. not that I think the randomizer works like that. it's just to show how easy it is for any kind of formula to generate more of certain numbers than others.

if you take the entire history of numbers generated by the randomizer and the amount of rolls each number got isn't very, very close to 16.67%, then right there you have proof that the randomizer is off (in this case, favoring the defender). a whole bunch of things could cause it. like I said, without the formula we're just speculating. if it *is* in fact close to 16.67%, then there's still a chance that some sort of temporal component in the formula generates more of a certain number any given time, and that's evened out throughout the day.

I had actually thought of the text file containing a really long string of actual random numers, 1 through 6. any sort of bias generated by the formula would tend to be evened out by the fact that it would pick a line and then read a number from that line that has absolutely no correlation to the formula that generated it. interesting to hear another web site is using that (though I'd never heard of them).

@ cardinals: the odds of losing three 6x1s are 1 in just over 1 million.

 

[WidowMakers]

But those three- 6v1 loses happened over a two week period.
In that two week period lots of dice were rolled here at MC.
So the 1 in a million chance is not that big of a deal if millions of dice are rolled.

I guess my question is how would anyone here recommend that we make our random numbers.
I cannot promise anything will change since I do not run or know how the numbers are made now.

i am just curious.

If each player had a record of every roll they got while attacking, would you be happy then?

BUT....what if after 10,000 rolls here on MC you had 16.7% for each dice option (1-6).
Then one day you still rolled 6-1 and lost 3 times, would you still accuse the MC site of non-random dice?

 

[Bluebonnet]

I just find it funny everyone complains but no one takes the time to prove anything. If you think the system is flawed, take out a piece of paper and write down all your rolls for 4 or 5 games. Then run a statistical analysis or have a friend do it for you. It is simple math. If the numbers agree to what is posted as should be random, guess what it is.

Otherwise does as I do and accept it.

BTW I am paid to do some low level stats analysis as part of my job and it is not worth my time. Random can be odd, but ut is still random.

Nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine nine

 

[Bluebonnet]

Okay, I figured out a simple test for those who want to prove the sites dice are "fixed"

You don't even have to use statistics and the odds of 3vs2 or anything. Besides, that is all stuff proved centureies ago. All anyone has to do is prove if the numbers being generated do not average 16.67% for any given number. I figure it will take close to 1,000 roles. Make a simple 5 column chart for each role of the dice. 3 columns for attacking, 2 for defending. Then start recording. Calculate the average each number shows up in the different columns.

After setting up a random number generator in excel, here are the results for 1-4 in a single column after 1,000 roles. If they don't start nearing 16.7%, they are fixed!!!!!!!!

Roles #1 #2 #3 # 4
200 10.5% 23.5% 17.5% 15.5%
400 12.5% 19.0% 15.0% 17.8%
800 14.9% 17.3% 15.9% 17.5%
1000 15.4% 17.3% 15.7% 17.3%

 

[Atlas-shrugged]

First, to Atlas. Over at Conk (we may as well name it, or perhaps you're talking about a different site) if you do a forum search of "dice" you get 29,940 results. A thread that was begun on March 15 is already five pages long, and there are other threads started in the past few weeks as well. At that site you can call up your dice history, yet there are still thousands of posts by folks unhappy with their luck.

I don't know what Conk's current system is, but i believe they used to pull their dice rolls from a pre-generated list (of many thousands of results) rather than generating random numbers each time an attack is made. Whenever I got bad dice, a voice inside me instantly blamed the site because that list of dice must be weighted against the attacker. I decided that the list was biased against 3-1 rolls - which I remember losing a lot - but possibly favorable toward 2-1 rolls.

Coming over here, I feel the exact same way: the odds say I should win my 3-1 rolls, but often I don't. There's either something wrong with the server, or there's something wrong with my expectations. So how often do I lose 3-1 rolls? Is it close to 1/3, which is what it should be? Probably, but I don't tend to remember the 3-1s that I win; I expected to win those, so they didn't phase me.

As far as rolls happening here that wouldn't happen in real life, well, yes they could happen in real life. A run of bad luck - and good luck - can strike in any game or dice or cards. It's rare, but it happens. But when you start playing games on-line, in which the normal length of a game is compressed and you can have dozens of games going at once, you increase the odds of a good or bad streak because you're playing so many more games, and thus rolling the "dice" much more often. I bet if you sat down with friends and played 15 games of Risk over the next week, you'd see a few examples of extreme luck.

It is my understanding that the numbers generated by the server here are each generated independently. The server doesn't decide first who is going to win or lose and then assign a number, it generates a random number between one and six for each of the troops in play - up to three independent numbers for the attacker, up to two for the defender. The exact same line of script is being run for the attacking troops' numbers as for the defending troops' numbers. Let me repeat that: the system for generating the attacking dice is the same as that generating the defending dice. If anyone is suggesting that the "dice" systematically favor the defender, what they're saying is that either (1) we have programmed the server to favor the defender, or that (2) the server has a mind of its own and has decided it wants to help out the defender. Since the second option is ludicrous, that leaves only the first: that we want the attacker to lose. But why would we do this? Wouldn't it actually make more sense for us to tweak the server so that the attacker has an unfair advantage, since that's what our users see and we want to make our users feel successful and keep on playing?

No I am talking about LG but I now have changed my mind , Ignore the stuff I said(as most have already), been playing at both and have realized my mistake.I still love MC.Now That I won afew:banghead: