German UA and difficulty level

Creepy Old Man

Warlord
Joined
Oct 10, 2010
Messages
295
A few days ago, I was trying an immortal level game with Germany. I had a weak start, and after clearing six barbarian camps without gaining a single unit, I decided to restart. On the second game I got two unit from eleven camps through the game.

Last night, I tried a zero-city-challenge game, on settler level. After converting all eight of the first eight camps I cleared, I counted through the whole game ... and converted 39 of 46 camps.

The German UA is supposed to convert 50% of defeated barbarians in camps. On Immortal level, 50% = 2/17. On settler level, 50% = 39/46.

Has anyone else noticed this sort of bias? Are any of the other UAs dependent upon difficulty level?
 
It's purely 50% meaning you just happened to got screwed over a lot by the RNG. I don't think there's a bias.
 
Humans are notorious for having no grasp of probability theory- as Sonereal said, it's a 50% average conversion rate and you were just very unlucky. In fact, you converted about 65% of the camps, which is completely normal, and I highly doubt the developers would lie to us about something like this.
 
Actually, that's more like 85% and with 46 rolls it definitely sounds a little fishy. The probability of hitting 39 or more out of 46 rolls is 0.0000009. The probability of hitting 2 or less out of 17 is 0.001. The probability of this happening in back-to-back games is pretty small. It doesn't mean it can't happen, but it's definitely unlikely. It's not like it would be the only bug in the game, not everything in this game works like the devs intended.
 
What you have here is what's called empirical evidence, which is evidence based on observation.

Flipping a coin and getting heads 100 times in a row is not impossible, it's just very, very improbable.


Threads like this are a good example of why I think statistics should be a mandatory course in high school.
 
Threads like this are a good example of why I think statistics should be a mandatory course in high school.

They call it Discreet Mathematics in my school (and I'm taking it). But really, people should take a intro course into Psychology since they talk about the Illusionary Colleration (sic) pretty early on.

It isn't a bug. If he was playing on Deity, hit 20 camps and got something each time, there would be a 0.00000095367431640625 chance of that happening but still a chance.
 
Dude, he wasn't saying this is evidence of a bug. He is asking if this bias continues with other players so as to see if the developers investigation of the possibility of this being a bug is warranted.

If others would try and see if this happens instead of attacking his knowledge of statistics maybe we can get a good conclusion here.
 
Dude, he wasn't saying this is evidence of a bug. He is asking if this bias continues with other players so as to see if the developers investigation of the possibility of this being a bug is warranted.

If others would try and see if this happens instead of attacking his knowledge of statistics maybe we can get a good conclusion here.

The problem is that the only people who would know if its a bias would be the devs and any modder that looked at the UA. If there was a difficulty bias, it would've been found and noticed by now.

He didn't post evidence of there being a bias or bug. He just compared two single games that give no indication of a bias. The UA is "50%" of the time not "every other time".

The conclusion is right there. The best thing would be to repeat the test several times and posting the results.

No one was attacking his knowledge of statistics.
 
A few days ago, I was trying an immortal level game with Germany. I had a weak start, and after clearing six barbarian camps without gaining a single unit, I decided to restart. On the second game I got two unit from eleven camps through the game.

Last night, I tried a zero-city-challenge game, on settler level. After converting all eight of the first eight camps I cleared, I counted through the whole game ... and converted 39 of 46 camps.

The German UA is supposed to convert 50% of defeated barbarians in camps. On Immortal level, 50% = 2/17. On settler level, 50% = 39/46.

Has anyone else noticed this sort of bias? Are any of the other UAs dependent upon difficulty level?

I didn't keep count in my two German games. As a side note, I think the German "convert" achievement carries over across playthroughs. But I kept such a lousy count, I can't even be sure of that.
 
Actually, that's more like 85% and with 46 rolls it definitely sounds a little fishy. The probability of hitting 39 or more out of 46 rolls is 0.0000009. The probability of hitting 2 or less out of 17 is 0.001. The probability of this happening in back-to-back games is pretty small. It doesn't mean it can't happen, but it's definitely unlikely. It's not like it would be the only bug in the game, not everything in this game works like the devs intended.

Not so. Your probability calculations are accurate, but as the physicist Richard Feynman said (in "The Meaning of it All"), "it doesn't make sense to calculate after the event. You see, you found the peculiarity, and so you selected the peculiar case."

If I were to play a game as the Germans, I'm sure I'd find something odd. Either the number converted would be below 50%, or it would be above 50%, or it would follow some pattern like alternating back and forth between converting and not converting, or it would go convert, not convert, convert twice, not convert twice, convert thrice, not convert thrice, etc., or something else. Look at Creepy Old Man's examples: (s)he played twice and found SOMETHING odd both times, but not the same thing. Which suggests (based on the available data) that the odds are pretty good that every time he plays the Germans he'll be able to (after the fact) come up with something that was odd about the "pattern." But that's telling us something about him, and nothing about the programming of the Germans.

When you make the prediction ahead of time (and specifically--not something vague like "significantly more/less than 50%" where you don't quantify what counts as "significant" ahead of time) and it happens reliably, then you'll have something. When you calculate "the odds" afterwards, you're using mathematics incorrectly: the odds of something that has already occurred is always 100%.
 
It's very easy to check. Just play the biggest map possible with only one ai opponent as Germany on Deity, and repeat a couple of times, with the only aim to take barb camps. If the results are something like 25-45 out of 70, then its a straight 50/50 ratio as advertised, end of discussion. However, if you only get say 15 or so out of 70, then its most likely a sliding scale, with 50% at noble and then + for harder lvls and - for lower levels. To be completely thorough, a few settler lvl run thoughs should be made too.

The original info the OP gave, is good cause to investigate, and highly suggestive of a sliding scale, while not being definitive proof.
 
Humans are notorious for having no grasp of probability theory- as Sonereal said, it's a 50% average conversion rate and you were just very unlucky. In fact, you converted about 65% of the camps, which is completely normal, and I highly doubt the developers would lie to us about something like this.


Are you not a human? :undecide:


It would be interesting to test this. But one game alone can't prove anything.
 
From what I can tell from my games using the Germans is that it doesn't matter difficulty you juet get lucky or unlucky.

I've played games where I have gone 0/5 and then others where I go 5/5. I have gotten both of those results on prince difficulty. 5 is the biggest gap I have gottne inbetween camps without getting a unit. I don't usually count total just how many till I get 1.

I will do several trials on different difficulties but I expect simular results.
 
Not so. Your probability calculations are accurate, but as the physicist Richard Feynman said (in "The Meaning of it All"), "it doesn't make sense to calculate after the event. You see, you found the peculiarity, and so you selected the peculiar case."

If I were to play a game as the Germans, I'm sure I'd find something odd. Either the number converted would be below 50%, or it would be above 50%, or it would follow some pattern like alternating back and forth between converting and not converting, or it would go convert, not convert, convert twice, not convert twice, convert thrice, not convert thrice, etc., or something else. Look at Creepy Old Man's examples: (s)he played twice and found SOMETHING odd both times, but not the same thing. Which suggests (based on the available data) that the odds are pretty good that every time he plays the Germans he'll be able to (after the fact) come up with something that was odd about the "pattern." But that's telling us something about him, and nothing about the programming of the Germans.

And so according to you guys no correlation can ever be proven. You are over thinking this. Sure it could be a freak occurence but 39/46 and 2/17 is far enough past any reasonable statistical significance that it should at least warrant some suspicion that what the civilopedia/devs tells us is incorrect. I'll play a game as the germans on settler now and see what I find (pangea, 3 AIs, standard).
 
And so according to you guys no correlation can ever be proven. You are over thinking this. Sure it could be a freak occurence but 39/46 and 2/17 is far enough past any reasonable statistical significance that it should at least warrant some suspicion that what the civilopedia/devs tells us is incorrect. I'll play a game as the germans on settler now and see what I find (pangea, 3 AIs, standard).

No it isn't. It's a freak occurrence. Start up ten more games on both difficulties and grab the EXACT same number of camps each time and count conversion. Then there would be statistical significance. The RNG allows for EXTREMELY improbable events to happen every once in a while, like a spearman killing a tank.
 
From what I can tell from my games using the Germans is that it doesn't matter difficulty you juet get lucky or unlucky.

I've played games where I have gone 0/5 and then others where I go 5/5. I have gotten both of those results on prince difficulty. 5 is the biggest gap I have gottne inbetween camps without getting a unit. I don't usually count total just how many till I get 1.

I will do several trials on different difficulties but I expect simular results.

But this would agree with the thesis. On prince you would expect the probability of converting a camp to be 50%. On settler it would be higher and on deity lower. Also, there's a big difference between 0/5 and 39/46.
 
No it isn't. It's a freak occurrence. Start up ten more games on both difficulties and grab the EXACT same number of camps each time and count conversion. Then there would be statistical significance. The RNG allows for EXTREMELY improbable events to happen every once in a while, like a spearman killing a tank.

You can't know that's it's a freak occurrence any more than I can know that it's not. Your dismissal of the possibility that it's not a freak occurrence is just as irrational as claiming with certainty that it isn't. Also if the probability is really unaffected by difficulty or anything else you don't need to play several games, let alone pop the exact same number of huts. That said, so far in my game I'm not seeing anything like the OP did.
 
You can't know that's it's a freak occurrence any more than I can know that it's not. Your dismissal of the possibility that it's not a freak occurrence is just as irrational as claiming with certainty that it isn't. Also if the probability is really unaffected by difficulty or anything else you don't need to play several games, let alone pop the exact same number of huts. That said, so far in my game I'm not seeing anything like the OP did.

I can dismiss it as a freak occurrence because there's a small chance of it happening regardless. He only had two games to show as an example. I could easily point out how in my first Settler game I managed to only convert a few camps after a fun game of "smash the camp".

If probability is unaffected by difficulty (which it is), you're right we don't need to play a butch of games. Especially since that we'll never get a straight 50/50 thing going on. I said pop the same number of huts because 17 is really different from 46.
 
The problem is that the only people who would know if its a bias would be the devs and any modder that looked at the UA. If there was a difficulty bias, it would've been found and noticed by now.

Modders can't know for sure since the source code isn't released yet.

And if you heard Sid Meiers talk at GDC2010 on how they manipulated the RNG in CivRev I can't see how anyone can say for sure they didn't do anything in Civ5.
 
And if you heard Sid Meiers talk at GDC2010 on how they manipulated the RNG in CivRev I can't see how anyone can say for sure they didn't do anything in Civ5.

I'm not saying they didn't do anything in Civ5. I'm just wondering what's the point in having a info screen that lies to you and says you got a 50/50 chance when your chances is more 60/40.
 
Back
Top Bottom