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

[JackRudd]

If you do n tests, and p is the probability of any one test going in the human's favour, then the probability of exactly r tests going in the human's favour is given by the following formula:

(n!/r!*(n-r)!)*(p^r)*((1-p)^(n-r))

[Hauptman]

I am almost possitive the rng got broken in the patch...

Enough to try a test.

Take a fresh 1.0 instal of civ 4. enter the world builder. make a few "runways" 21 squares long with rails. take a tank with zero promotions and place 1 at the end of each runway. then fill all other squares with barbarian weak units, say warriors on 1 , archers on the next then spearmen etc etc...

Now start the game and attack down the rails (you can attack 20 times in a turn with a tank as long as it kills the unit...) and see how far the tanks can get before they die.

It requires me to uninstall the game to do a fresh install, so i havent done it yet. But i'm willing to bet that version 1.0 the tank could probably kill much farther than in 1.52. And I would believe, from what ive seen, the difference would be massive. perhaps along the lines of version 1.0 tanks killing upto 10 axemen, while 1.52 tank would die on the 10th archer, or worse.

someone with some time try this one?

[Blake]

It requires me to uninstall the game to do a fresh install, so i havent done it yet. But i'm willing to bet that version 1.0 the tank could probably kill much farther than in 1.52. And I would believe, from what ive seen, the difference would be massive. perhaps along the lines of version 1.0 tanks killing upto 10 axemen, while 1.52 tank would die on the 10th archer, or worse.

If you try this you'll be SO dissapointed.
The 1.52 tank will do MUCH better.

[TheHateMale]

Originally posted by JackRudd
If you do n tests, and p is the probability of any one test going in the human's favour, then the probability of exactly r tests going in the human's favour is given by the following formula:

(n!/r!*(n-r)!)*(p^r)*((1-p)^(n-r))

Cheers!

[Blake]

Heheheh because I'm a huge nerd (I stand over 6"5 tall) I actually bothered to run the test.
Enjoyz0rs.



Notes:

• The tanks started off the top of the screen. They then moved 10 times using "down arrow".
• I used Pikemen as the fodder, as often the tanks would beat warriors without damage. Pikemen were the only unit that could reasonably be called a Spearman that could actually kill tanks when stacked 10 deep. Tanks would mow down 10 spearmen like they aint there.
• Note they are the 28 str tanks, and aggressive pikemen.
• Yes. One of the 1.00 tanks only managed to kill 2 Pikemen before kicking the bucket .
• It's difficult to perform multiple trials because 1.00 doesn't have the random seed option. I did however do some frigate vs frigate fights in the corner to "clear out" the RNG, should some people be superstitious about such things.

[Hauptman]

I love you blake... wish you woulda tried other units too though... and ran comparisons on up to infantry, because its driving me batty the frequency now at wich i loose sure thing fights.


so keep it up, more SSs, i love it.


PS. is it only 10 kills max per turn or can they do 20?

edit pss. it is "units get 10 movement on rails" not " units use 1/10th of a move on rails" so 2 move units dont get a benefit.

[xxFlukexx]

I think I'll do a test too, it'll be the axemen test. If I make one stack of 50 axemen attack another stack of 50 axemen using stack combat, that should give me an accurate output, right? I'll post it real late tonite or early tommorrow.

I forget how to calculate how many attempts you need to do to get an accurate result but I know how to find out.

[Blake]

Railroad is 10 tile movement for all units. Tanks move 10 tiles. Infantry move 10 tiles. It's the realistic way to do it.

The RNG is working fine, the system may be "Broken as Intended" because it allows for some bizzare, unexpected and unintuitive results, but it is "Working as Designed" anyway.

I've played dozens of Monarch games and always go heavy military. You just learn there is no such thing as a sure-thing battle. For 50% odds, you bring double the number of units +1 (like if you want to kill 5 axemen with your own axemen, you'd want 11 axemen to be sure of wiping the whole stack out).

Like destroyer vs destroyer. There's a 3% chance the enemy destroyer will win with full health remaining. That's 1 in 30 fights so is actually quite probable statistically. There's a 3% chance it'll beat the next destroyer too (if it took any damage in the first fight, it's much less likely to win the second). There's only a 0.09% chance of two flawless victories in a row, which is 1 in 1000. These odds are likely enough to happen atleast once in your destroyer vs destroyer career. The odds of the destroyer beating 3 of your destroyers is something like 1 in 2000, those odds are so slim you don't need to bet on them. You do need to bet on the destroyer killing one of your destroyers.

In short, you have to prepare for losing every fight, pretty much regardless of odds, unless the odds are 2.5:1 or better, altough I have lost one (and only one) 2.75:1 fight. In any case the chance of losing a 3:1 fight is extremely low, if you do it's something to celebrate because it's just that darn unlikely, it's like winning lotto without buying a ticket.

So the rule of thumb is to always bring twice as many units as you think you should need, and then an extra unit for good luck. Usually, it's better to bring overwhelming numbers than rely on being lucky .

[Hauptman]

mix it up for my benefit, prove to me it is just that im'a unlucky sob, and that the patch didnt break nuttin'

Try attacking axes vs spear stacks, then visa-versa.
then samurai vs mecemen (for the the first strike question)

and so on-so forthe.



If a 50 stack of axes only kill 40-60% of the spears we know somethin's broke!~


edit (response to blake without a DP)

Yeh i agree that there should be a chance, but i scream "YOU HAVE GOT TO BE KIDDING ME!" much more frequently than i used to, even though 1.52 was supposed to make the stronger units even MORE likely to win combat.

it is only frustrating because i run militaries at 50% of an ai at my peak but use bombers and arty to weaken them for my real units. So an actual unit loss hurts me enough to get rather frustrating.

[Blake]

even though 1.52 was supposed to make the stronger units even MORE likely to win combat.

You're decieved. 1.52 only makes strong WOUNDED units more likely to win vs weaker units. It doesn't change the odds of full-health fights one bit.

[Urban Ranger]

A lot of players seem to have misconceptions about statistics and probability.

1. P&S is based on a theoretical, infinitely large sample space.
2. Runs (a consecutive sequence) do not indicate something is broken with a random number generator. See 1. For example, if you toss a coin and record the results, runs of heads or tails that are 4 or 5 in length are in fact quite common.
3. Just because you have a 99% chance of winning any single combat doesn't mean you won't lose. In fact, how many battles do you fight in an average Civ 4 game?

Another thing. If you wish to conduct an experiment, you must have a control. An experiment without a control is meaningless.

[Blake]

To add to what UR said, lack of runs, lack of losing 99.9% battles etc actually means the RNG is broken. As long as the RNG works well, there must be a chance of losing any battle, however close to 100% it is. So as I said, Working as Designed, Broken as Intended.

[Lansing]

I'm afraid pitting superior units against inferior units and seeing how many they can kill is not at all the kind of "rigging" I believe is going on. What I believe is happening (among other things) is that when a fight is 50/50 according to the numbers, the computer will win much more than 50% of the time. I would guess it is closer to 75%.

I also believe that when the computer has a decided advantage, such as a destroyer vs. a battleship, it is almost impossible to kill the battleship. I just ran a destroyer against a battleship 100 times and did not win once.

[Urban Ranger]

Originally posted by Lansing
I'm afraid pitting superior units against inferior units and seeing how many they can kill is not at all the kind of "rigging" I believe is going on. What I believe is happening (among other things) is that when a fight is 50/50 according to the numbers, the computer will win much more than 50% of the time. I would guess it is closer to 75%.

I also believe that when the computer has a decided advantage, such as a destroyer vs. a battleship, it is almost impossible to kill the battleship. I just ran a destroyer against a battleship 100 times and did not win once.

As I said, you need a control.

Set up the same scenario twice, once in a single player game against a computer player, another one in a multi-player game against another human. Then run them at least 200 - 300 times to see if there is some kind of statistically meaningful difference between the two. If there is, your contention becomes weakly supported. Then we need to repeat the experiment, at least 200 - 300 times, to see if somebody else can get what you get.

[rjmatsleepers]

Originally posted by JackRudd
If you do n tests, and p is the probability of any one test going in the human's favour, then the probability of exactly r tests going in the human's favour is given by the following formula:

(n!/r!*(n-r)!)*(p^r)*((1-p)^(n-r))

I think a little approximation would help. Roughly 90% of the time you would expect the number of human wins to be in the range n*p +or- 2*sqrt(n*p*(1-p))

If you get a result outside this range you should be suspicious.

RJM at Sleeper's