|
|
Post by Ringel on Mar 17, 2012 10:19:42 GMT -5
Does it randomly pick a rare between 1 and 42 or is the algorithm more complex? Because the stats are suggesting something more complex.
|
|
|
|
Post by Tuism on Mar 17, 2012 11:16:43 GMT -5
Oh wait, so it's supposed to be truly random, ok. But yeah it's kinda hard to say "this random isn't random enough", cos that means having to doctor it in some way... Like say... Randomising 222 cards (or however many it is in the set, I dunno), and making sure when cards are sold they're packed from that set, and resetting the set over and over again... To ensure the same number of any single card in the world at the same time. But then, is it truly random?  I really don't care all that much about how random the random is |
|
|
|
Post by Ringel on Mar 17, 2012 11:20:06 GMT -5
No, you don't have to reset anything to be truly random. A truly random generation will tend toward an equal distribution the larger the number of rares you get, but it doesn't ever have to be exactly equal.
To be truly random, when you open up a pack, you have exactly 1 in 42 chance of drawing any particular rare. Just like if you roll a die you have an exactly 1 in 6 chance of rolling any particular number.
The numbers show that this isn't the case. The dice are weighted.
|
|
|
|
Post by Ringel on Mar 17, 2012 11:35:27 GMT -5
No, your sample is too small to give any indication of such a thing. To have a good sample you would need to have at least 1% of the population. Taking only gamecenter numbers here, with 100,000 players, you would need a sample of at least 1,000 players. I'll check with Thierry if I can run some basic stats on the server, but nothing show me there is a bug there. No, My sample is not that small-- You do not need at least 1000 players. You are mistaken about the "1%". There is no connection between the size of the population and the size of a good sample, otherwise medical research would be impossible. Yes, it is possible my sample is flawed, but the evidence suggests otherwise. Since the cards are supposedly random, then a sample of cards would be a random sample. A more perfect sample than most research ever conducted! It could be flawed in that some players could have made arithmetic mistakes, or falsely reported their numbers, but take a look at the code! I teach stats for a living (actually math, some stats). I really do know what I'm talking about. |
|
|
|
Post by Tuism on Mar 17, 2012 11:36:39 GMT -5
Listen to the devs, there's no point to them lying to us  at all. Would this question even exist if you got all the cards you wanted instead of all the cards you didn't want? Anyway I still don't care enough about how random the random is to really get into this. I rest my case |
|
|
|
Post by Ringel on Mar 17, 2012 11:39:43 GMT -5
Listen to the devs, there's no point to them lying to us at all. Would this question even exist if you got all the cards you wanted instead of all the cards you didn't want? Anyway I still don't care enough about how random the random is to really get into this. I rest my case I don't think the devs are lying. I think they either made a mistake in the code, or there has been a miscommunication somewhere. The numbers are pretty solid evidence on my side. |
|
|
|
Post by Ringel on Mar 17, 2012 11:41:05 GMT -5
Listen to the devs, there's no point to them lying to us at all. Would this question even exist if you got all the cards you wanted instead of all the cards you didn't want? This has nothing to do with the cards I got. It has everything to do with me being a stats person and seeing Raphael post some surprising numbers. I would have sworn it was random until I saw the evidence otherwise. What evidence do you have? I've got numbers. |
|
|
|
Post by Tuism on Mar 17, 2012 11:52:56 GMT -5
I have no evidence, nor did I claim to, I will just allow the devs to draw their own numbers and show us.
|
|
|
|
Post by Ringel on Mar 17, 2012 11:53:48 GMT -5
I posted in the bugs thread, with the hopes they will investigate their own code.
It isn't even necessary for all rares to be equal. Maybe they want blue templar rares and black assassin rares to be rarer. That would be fine. But if they want them to all be equal, then there is a bug in the code.
To make a nonscientific guess, they randomly pick color, and then pick a rare, only they coded it so (with templar) blue is picked half as often as it should be. The code is duplicated for Assassin cards, with the colors changed, so the bug was duplicated as well.
|
|
|
|
Post by Tuism on Mar 17, 2012 12:15:26 GMT -5
Actually, reading over what I said, I do apologize for coming across as offensive or defensive, that's not the intention at all. Sorry If their code is broken, they should fix it. To me though, a) I agree that none of us has a big enough sample to say there's a skew (have you seen the rarity survey I did? I think 8 players' cards It looks pretty even from there), and b) I really don't care how random the random is, I have 1 Juno, the auction house will fix that (hopefully) |
|
|
|
Post by Ringel on Mar 17, 2012 12:30:10 GMT -5
It is okay, I did see the rarity survey you did, but your numbers aren't inconsistent with my claims either. I agree that things don't have to be random.
Why not count your rares by color and post the data here?
|
|
|
|
Post by Tuism on Mar 17, 2012 12:59:59 GMT -5
Ok ok, indulge you Gold: 99 Black: 49 Red: 105 Purple: 105 Blue: 70 Green: 110 Black and blue are the expected anomalies, but remember I have no idea how many Templar or assssins I've bought, and I don't feel like counting again The thing is the bias is different to an assassin pack vs Templar pack, and each of them doesn't only have black or only have blue - so the more assassin pack you buy the less blue you'll have and the more Templar pack you buy end less black you'll have, while the other colours remain unaffected, you'll get them regardless of Templar or assassin. That makes sense to me, and make my numbers look pretty correct. Except of course I'm just one sample. |
|
|
|
Post by Ringel on Mar 17, 2012 13:01:56 GMT -5
You have to separate by Assassin vs Templar or the numbers aren't any good!
Please keep an open mind here. I didn't start thinking things aren't random, and I'm not eyeballing the numbers. The stats tests I use are well established.
|
|
|
|
Post by Tuism on Mar 17, 2012 13:17:08 GMT -5
Man I'm too nice a guy Templar/assassin: Gold: 58/41 Black: 22/27 Red: 35/66 Purple: 57/48 Blue: 50/18 Green: 57/53 Have fun |
|
|
|
Post by Ringel on Mar 17, 2012 13:26:47 GMT -5
Thanks! I'm going to guess you miscounted those reds there-- you might want to double check. You might have miscounted other things too, because your totals are different between the posts. -- Actually only two totals don't match, red an blue. Notice that your Assassin blacks are rarer than they should be, as are your Templar blues. |
|
