I believe combat is rigged in this game and it ruins it for me

[Lul Thyme]

Originally posted by xxFlukexx
Are you sure about that Soren? Then maybe the RNG is broken?



I shouldn't have a folder that I just created for these and already up to screenshot #15. Ok maybe I lose that battle 1/10 times and maybe I don't get many shots in, but never a shot in? And this happens regularly? Sorry, but when it takes 4 axeman to kill an enemy swordsmen with my advantage always greater than 50% I get suspicious.

I'm sorry buddy but there are enought statiscal minded people here that you will not convince anybody with screenshots, which is as it should be.
You need to do statistical tests.

[Lul Thyme]

Originally posted by Urban Ranger



Suppose a destroyer vs destroyer combat takes 5 rounds (because of the 0.2 multipler). So, the side that wins needs only to win 3:2.

Suppose you stage 50 fights in the WB and loses all ten. If each combat is resolved by a die roll losing all 50 is a 1/(0.5)^50 occurance, extremely rare. However, in a multiple round resolution system the other side needs to only get 50 more wins out of 250 rounds than you to win all of them. It suddenly becames much more likely.

(I assume you meant to say lose all 50 in the first sentence)
Let me put it this way.
If a combat has 50% chance of being won, then the distribution of wins is the same regardless what the model for the actual combat is (number of rounds etc...)
In a multiple round system, the chance of winning each combat is still 50% and so the chance to lose all 50 fights is still 1/(0.5)^50 , extremely rare.
It's still the same Bernouilli trial, since each combat is still a 50/50!!!
I've seen you post in off-topic UR, I'm sure you will understand your mistake in this case by thinking just a little bit about it.
Or do a test on a computer if you don't believe me.

EDIT: one way to see why your explanation is flawed:
the probability in the second case is NOT the probabilty to lose 50 more rounds out of 250, as the order is very important here. For example if you win the first 50 and they win 200, you still won many combat while if you alternate WLLLLWLLLL... you don't.


(btw as some ppl got out of the closet about this, I also have a degree in mathematics)

[DrSpike]

Originally posted by Urban Ranger

This just proves my point that you are not a trained scientist.

Appeal to Authority isn't going to fly.

Well I'm not going to get into a debate about what constitutes science - suffice to say I know considerably more of relevance than you do.

I'm not appealing to authority. I'm not right because I have a Ph.D. However, my knowledge in this area means I have more insight than you, and as it happens, I am right, as a result. I don't understand why you enjoy so much posting strong opinions on subjects about which you know little.

More people have posted on the control I see, predominantly to agree with me. I think the clearest way to see the point is that you don't need a control in this case because the counterfactual case is known. All we have to do is (as Mergle sensibly did) is test whether the results are statistically significantly different from that counterfactual.

If the counterfactual wasn't known you'd need an experimentally based control. Since it is known all the experimental control does is introduce bias.

To see this take the coin flipping example. Say I flipped one 100 times. The most likely outcome for a fair coin is 50 heads, but there is a distribution around this most likely outcome. Lets say I get 48 heads, a not unreasonable outcome.

Now let's say I have a second coin, and I strongly suspect it is weighted to give more heads than normal. I flip my second coin 100 times too. Do I test the number of heads I get in this experiment against the distribution suggested by the control case, in this case a null hypothesis of p=48/100=0.48? No, it is clear that all using the 'control' does is introduce bias. I know the best null hypothesis to test my results using the possibly weighted coin against because it's perfectly obvious.

Ok that's enough, and I'm not going to respond to whatever nonsense UR comes up with next.

[DougM]

Originally posted by DrSpike


Well I'm not going to get into a debate about what constitutes science - suffice to say I know considerably more of relevance than you do.

I'm not appealing to authority. I'm not right because I have a Ph.D. However, my knowledge in this area means I have more insight than you, and as it happens, I am right, as a result. I don't understand why you enjoy so much posting strong opinions on subjects about which you know little.

More people have posted on the control I see, predominantly to agree with me. I think the clearest way to see the point is that you don't need a control in this case because the counterfactual case is known. All we have to do is (as Mergle sensibly did) is test whether the results are statistically significantly different from that counterfactual.

If the counterfactual wasn't known you'd need an experimentally based control. Since it is known all the experimental control does is introduce bias.

To see this take the coin flipping example. Say I flipped one 100 times. The most likely outcome for a fair coin is 50 heads, but there is a distribution around this most likely outcome. Lets say I get 48 heads, a not unreasonable outcome.

Now let's say I have a second coin, and I strongly suspect it is weighted to give more heads than normal. I flip my second coin 100 times too. Do I test the number of heads I get in this experiment against the distribution suggested by the control case, in this case a null hypothesis of p=48/100=0.48? No, it is clear that all using the 'control' does is introduce bias. I know the best null hypothesis to test my results using the possibly weighted coin against because it's perfectly obvious.

Ok that's enough, and I'm not going to respond to whatever nonsense UR comes up with next.

Almost exactly right.

It's important to note that even if you are using idealized coin (50/50) as your control, that the experiment still has a control.

In fact, it is quite important to understand that you are comparing weighted coin to idealized coin as your control and not any real coin as your control, precisely because a real coin might in fact have a bias that you are not aware of. A real coin (for example, an actual US penny) might in fact be 48% heads, and in fact it is unlikely that any real coin is ideal (with different features on the two sides making airflow over the two faces uneven) .

[Fried-Psitalon]

Bleh, nevermind.

[DrSpike]

Originally posted by DougM


Almost exactly right.

It's important to note that even if you are using idealized coin (50/50) as your control, that the experiment still has a control.

In fact, it is quite important to understand that you are comparing weighted coin to idealized coin as your control and not any real coin as your control, precisely because a real coin might in fact have a bias that you are not aware of. A real coin (for example, an actual US penny) might in fact be 48% heads, and in fact it is unlikely that any real coin is ideal (with different features on the two sides making airflow over the two faces uneven) .

Like rjm told you, even though he works or has worked in this area he doesn't know anyone that refers to the distribution that occurs under a Platonic thought experiment as a control. I don't either.

But it's clear that we agree; it's just terminology.

[LordShiva]

Originally posted by DougM
In fact, it is quite important to understand that you are comparing weighted coin to idealized coin as your control and not any real coin as your control, precisely because a real coin might in fact have a bias that you are not aware of. A real coin (for example, an actual US penny) might in fact be 48% heads, and in fact it is unlikely that any real coin is ideal (with different features on the two sides making airflow over the two faces uneven) .

Congratulations to the winner of the first ever instalment of LordShiva's Nitpicking Contest!!!

[DrSpike]

Originally posted by Fried-Psitalon
Bleh, nevermind.

I saw your post that this debate isn't about Civ.

Actually it is, because the thread is dealing with ascertaining whether combat is rigged, and this requires some knowledge of statistics.

Your post was actually the threadjack.

[DrSpike]

Originally posted by LordShiva


Congratulations to the winner of the first ever instalment of LordShiva's Nitpicking Contest!!!

[vee4473]

[xxFlukexx]

Originally posted by Lul Thyme
I'm sorry buddy but there are enought statiscal minded people here that you will not convince anybody with screenshots, which is as it should be.
You need to do statistical tests.

I did, genius. I posted that because in the beginning of the thread people were saying that stuff like the screenshot I posted above didn't happen. Please read the whole thread before posting about what I have posted.

[DougM]

Originally posted by DrSpike


Like rjm told you, even though he works or has worked in this area he doesn't know anyone that refers to the distribution that occurs under a Platonic thought experiment as a control. I don't either.

But it's clear that we agree; it's just terminology.

Hmm, that's interesting, that's the way I always see it. Different fields maybe (computer science?)

[DougM]

hooray!?

Originally posted by DrSpike
Congratulations to the winner of the first ever instalment of LordShiva's Nitpicking Contest!!!

Yay! Oh wait...



Maybe I didn't want to win that.

[Lul Thyme]

Originally posted by xxFlukexx


I did, genius. I posted that because in the beginning of the thread people were saying that stuff like the screenshot I posted above didn't happen. Please read the whole thread before posting about what I have posted.

Anything will happen given enough trials.
And you were replying to Soren who was only saying there is no AI combat bonus which your screenshot doesnt help contradict or imply....
I congratulate you for doing some tests afterwards as these are the kinds of things that help, but stand by my point that your screenshot shows nothing and that nobody claimed stuff like your screenshot does not happen.

[rjmatsleepers]

Originally posted by DougM


Almost exactly right.

It's important to note that even if you are using idealized coin (50/50) as your control, that the experiment still has a control.

In fact, it is quite important to understand that you are comparing weighted coin to idealized coin as your control and not any real coin as your control, precisely because a real coin might in fact have a bias that you are not aware of. A real coin (for example, an actual US penny) might in fact be 48% heads, and in fact it is unlikely that any real coin is ideal (with different features on the two sides making airflow over the two faces uneven) .

Dr Spike has already pointed out that some of us find this usage of the word control unusual.

And actually, there is a problem with your idea that we are comparing with an idealised coin. The binomial distribution is a mathematical construction that can be derived without any reference to coins (real or idealised) and indeed without any reference to any object that has a physical existance. It may help your understanding to think of a probability distribution as describing the results of tossing an ideal coin, but that is only one particular application of it.

I suspect that those who wish all scientific experiments to have a control subscribe to a particular philosophy of science in which a control is part of the definition of "scientific". As a result the meaning of "control" gets expanded.

Anyway, back to the OP data. The results appear to be significant, but have not yet been replicated. I have made a mental note to save my next game position that offers the opportunity for a test - just in case using world builder has an effect. If I get something suitable, I'll post it, but don't hold your breath. (And can someone remind me whether there is a penalty or bonus for barbarians - there was in civ 2, but I can't remember if there is in civ 4.)

RJM at Sleeper's