Challange: Test combat odds yourself

[coopervegas]

Funny the thread closed as soon as I posted I was going to start testing the odds and post the results. Funny...

Challenge, put your money where your mouth is... find out for yourself.

Dual map, 2 AI oppenents, on prince level.

Keep track of every combat has odds between 80% and 90%. Write down the odds, and if you won or lost it. Do this 250 times... shouldnt take long. I am going to do this 1000 times.

If you actually do this EXACT experiment, I GUARANTEE you will see that you are not winning as much as you should. If you choose to not do this experiment, understand you are just going by intuition or the word of others about the validity of the RNG.

My postulation is that the combat cheats on the combat odds when the human is ahead, to level the playing field. I have observed this over years of playing.

 

[Ghpstage]

Your data gathering method is flawed, getting evidence of your results would require hundreds of screenshots..... but I suspect you aren't interested in evidence.

A better way would be to build big stacks in worldbuilder and smash them into each other, as Windsor suggested in the other thread. Then a save made prior to testing can be provided, and a couple of useful screenshots to back your data up. If random seed on reload is on you could use just a hundred troops, save and reload to get extra result sets.

I did this with 150 Axes attacking Axes on Deity so all 50% odds. Hadn't thought of using random seed on reload at the time, so not a huge sample but I reckon its enough to disprove the claims of AI odds cheating you made in the other thread.

Spoiler :

EDIT
Set up a second test game, this time 100 Rifles vs 100+ Macemen at Deity. Odds are 96.8% and new random seed on reload is on so with this you can get as big a sample as you like.

Civ 4 has been analysed mathematically by a lot of different people, and the games code has been dived and modified for years. If there were significant issues with the RNG they would almost certainly have been proven long ago.

 

[Ataxerxes]

I think my good luck with Combat balances out your bad luck.

But no, I think combat odds work. GPStage gives a good explanation.

I play tournament bridge. You routinely hear people complaining there that computer hand records are skewed, which they aren't. Human memory is selective. I probably have average luck but think it's good because I usually win. Remember, we usually attack at 80+%Combat Odds so we think we'll win almost all the time.

 

[Kid R]

Always happy to participate in the interests of science

Rifle vs Mace is 96.8% so they should win just over 3% after an infinite number of fights. They got a little lucky here.

Funnily enough the 5 they won were all in the first 50, but then I got a run of 49 straight. So there you go that's randomness for ya, it can look a bit funny at first appearance.



Edit: ran 500 fights (all one at a time, no stack attack) while I had the game open. Result 484/16, which is getting back almost exactly to the displayed odds.

Bonus question: how many great generals did that give?

Spoiler :

10!

 

[TheMulattoMaker]

Fun... <snip> ...playing.

I've left what to me should be the important part.

You definetely seem to be very serious about this. To each his own. And I know there are a fair amount of people on this forum who are into every tiny detail of the game. That's fine with me- everybody paid their own money for the game, everybody should enjoy it the way they see fit. But seriously- it's a game. If you can't have fun playing it, why bother to play? If the RNG gods are cheating and it's ruining your fun, cheat back. Build yourself a mod with better numbers, or WB two macemen for every time you train an axe.

Far be it from me to criticize the way anyone else spends their leisure time. Furries, and Trekkers who speak fluent Klingon, and fanboys who dress up as Boba Fett are not really my thing- but hey, knock yourself out. It just seems to me you're not having any fun, so what's the point?

Not trying to troll or flame. Just... perplexed.

EDIT: Besides, if I could get RL to give me nearly as pleasant results as the RNG, I'd be a lot happier.

 

[theskibum]

I've left what to me should be the important part.

Well, you definetely seem to be very serious about this. To each his own. And I know there are a fair amount of people on this forum who are into every tiny detail of the game. That's fine with me- everybody paid their own money for the game, everybody should enjoy it the way they see fit. But seriously- it's a game. If you can't have fun playing it, don't bother to play. If the RNG gods are cheating and it's ruining your fun, cheat back. Build yourself a mod with better numbers, or WB two macemen for every time you train an axe.

Far be it from me to criticize the way anyone else spends their leisure time. Furries, and Trekkers who speak fluent Klingon, and fanboys who dress up as Boba Fett are not really my thing- but hey, knock yourself out. It just seems to me you're not having any fun, so what's the point?

Not trying to troll or flame. Just... perplexed.

If you, sir, are as good at making mulattos as you are writing posts, there must be some damn fine mulattos about!!!

 

[thecowwarrior]

Funny the thread closed as soon as I posted I was going to start testing the odds and post the results. Funny...

Challenge, put your money where your mouth is... find out for yourself.

Dual map, 2 AI oppenents, on prince level.

Keep track of every combat has odds between 80% and 90%. Write down the odds, and if you won or lost it. Do this 250 times... shouldnt take long. I am going to do this 1000 times.

If you actually do this EXACT experiment, I GUARANTEE you will see that you are not winning as much as you should. If you choose to not do this experiment, understand you are just going by intuition or the word of others about the validity of the RNG.

My postulation is that the combat cheats on the combat odds when the human is ahead, to level the playing field. I have observed this over years of playing.

coopervegas, I too have certainly felt the effects of seeming to constantly lose battles with high odds. In a game I just played, I lost to 96%+ odds over and over again. Doesn't seem right. Half dead Longbowmen shouldn't beat full health City Raider 2 Riflemen. What really eats me though, are the battles where I have 99.9% or so odds, and although I win the battle, my unit is left crippled and barely alive. If I have such overpowering combat ability, why does my unit barely survive? That kind of nonsense is the what has, so often, killed military advances by me. The only solution is to of course just spam the hell out of the best unit available, which makes the game lame. I guess I've just learned to deal with it.

 

[MarigoldRan]

Drill and first strikes always screws the human player. At least that's what it feels like.

If they've got first strike, you'll always take more damage than what you would expect from combat odds.

If you've got first strike, expect all your skirmishers to die.

More seriously: in wars, human players need CONSISTENCY. Losing 3-4 units to take each city isn't a big deal as long as you know you'll be losing 3-4 units per city (because then you'll plan your wars so that you can constantly replace your losses). This is what makes first strike so annoying because it screws the consistency (some cities you lose 2 units, and other cities you might lose like eight- effectively forcing you to keep your stacks together).

 

[ruff_hi]

You can actually do better ...

1. install BUG and turn on the logger
2. WB in the fight you want to watch (axe v axe, tank v spear, etc)
3. Save the WB file, edit with notepad and duplicate the unit line so that you get 10,000 v 10,000 (say)
4. Open scenario
5. Stack attack
6. Review the BUG logger that will give you fight by fight results

Then you have 10,000 observations and you can run all sorts of statistical tests

I play tournament bridge. You routinely hear people complaining there that computer hand records are skewed, which they aren't.

I've heard this too. The answer I have heard is that they are skewed when compared to hand dealt deals, but that is because the deck hasn't been sufficiently randomized (ie shuffled) prior to dealing. So, hand dealt deals are typically underskewed (if there is such a word ).

 

[nfw]

Now do the same test using cavalry or something else with withdraw chance, but with the computer attacking.

 

[ruff_hi]

Dang - that other thread is closed - wanted to post a few things.

First of all ...

... however, note that this is only 1 battle and actually means nothing.

One EASY EASY test to do is to only pay attention to odds that are ~95%.

Odds of losing a 95% battle is 1 in 20, 5%
Odds of losing two 95% battles in a row is 1 in 400 0.25%

So.. only keep track of battles that are 95%+. When you lose one, note it... 5% happens quite a bit... if you lose the very next battle of 95%+.. you hit your 1 in 400 chance... thats pretty low but not imposible by any means. See how many times you hit that 1 in 400 chance... its gonna happen a lot than 1 in 400.

Your approach here is floored.

You start by saying that 2 losses in a row when the combat odds are 95% win is 1/400. Yep - full marks. The complete possible outcomes are:

win / win: 361 / 400
loss / win: 19 / 400
win / loss: 19 / 400
loss / loss: 1 / 400

Next you say to record when you have a 95% loss and watch the next battle. If a loss in the 2nd battle occurs more frequently than 1/400, then you are arguing that the combat odds are wrong.

The problem with that approach is that you are not watching 380 of the 400 possible combinations. You are only watching 20 of the 400 (the ones where you lost the first battle) and you are surprised when your losses turn out more like 5% than 0.25%.

 

[Bob1475]

Coopervegas - I have played hundreds of hours and find no evidence of anything wrong. Odds are just that - 96% is NOT 100%. You can lose four in a row and then win 97 in a row. It is RANDOM. I note you continue to raise the issue but will not post your results only claims. Let's see your test results, others have posted, what about you, not stories but data?

 

[Tortsch]

I don't think that the RNG is broken, too.

But just to set up the experiments right:
In his first post Coopervegas is postulating that the RNG cheats, when the human player is ahead in score.

So the human player should be significantly ahead in score during the World Builder stack fights, in order to disprove Coopervegas' assumption.

 

[Lemon Merchant]

Coopervegas - I have played hundreds of hours and find no evidence of anything wrong. Odds are just that - 96% is NOT 100%. You can lose four in a row and then win 97 in a row. It is RANDOM. I note you continue to raise the issue but will not post your results only claims. Let's see your test results, others have posted, what about you, not stories but data?

I'm not sure that the OP's intent was anything other than meaningless complaining. His previous thread was closed because he was insulting everyone. I'm not sure that there is any constructive point to this at all.

To add something constructive here: The RNG is pseudo random. That's the way that computers do it. It's an approximation of randomness. I can't be bothered to work out all of the statistics and the math, but the basic fault here is not so much the RNG, but the human mind. We're conditioned to remember adversity because it helps us learn to deal with our environment. Animals do the same thing. It helps us to learn to avoid unpleasant experiences, or to learn not to repeat a course of action where something dangerous might have happened.

 

[KLarsen]

The only time I've suspected the combat odds to be broken was when playing the Legends of Revolution mod, where I lost most of my battles at about 95% odds, tank vs. tank (and yes, I did lose about 3/4 of those battles, but did not take screenshots or anything to prove it). However, I'm pretty sure this was due to some bug in LoR, as unmodded Civ 4 follows the odds reasonably well.

But then again, some people are just extremely unlucky, perhaps the OP is the kind of person who is repeatedly struck by lightning when venturing outside?

 

[Farm Boy]

Ruff hi - while that could just be a 1.3% chance win, scout wins vs barbarians are usually a product of low difficulty settings. I do not remember the specifics, but on the lower 2 or 3 settings the human player receives a could of "free wins" regardless of odds vs barbarians.

 

[Lemon Merchant]

Ruff hi - while that could just be a 1.3% chance win, scout wins vs barbarians are usually a product of low difficulty settings. I do not remember the specifics, but on the lower 2 or 3 settings the human player receives a could of "free wins" regardless of odds vs barbarians.

Yup. IIRC, on Warlord, it's 2 free wins, and Noble is one? (I think) And that's why there is an XP cap on barb wins, too.

 

[elmurcis]

I did small check just for myself... 100 my spearman vs 100 enemy (not barbarians) Horse Archers, odds was 76.3% win, after 100 battles I got 83 wins and 17 loses, so combat simulation from my view is correct... will check today again with some other % and units.
And I have experienced these hard situations when have 90% chance and lose 2 units in row and think all is bad... than later win again 2 battles with <30% chance and even don't see that later...

And btw - games are made for fun, also CIV4..... go and play something else if don't like how it works...

 

[ruff_hi]

Ruff hi - while that could just be a 1.3% chance win, scout wins vs barbarians are usually a product of low difficulty settings. I do not remember the specifics, but on the lower 2 or 3 settings the human player receives a could of "free wins" regardless of odds vs barbarians.

True - but this was well into the game and I had used my 'free' wins.

To add something constructive here: The RNG is pseudo random. That's the way that computers do it. It's an approximation of randomness.

It gets worse - consider the two sequences of dice rolls ...

3 2 4 1 5 6 4 3 4 4 2 1 2 3 5 4 5 3
3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3

What is the probability of getting the first sequence and what is the probability of getting the second sequence.

Spoiler :

A: The same .. 1/6 ^ 18

Also, some people say 'random' when they mean 'uniform ... consider the story of tiler laying 95 white tiles and 5 black tiles. He was told to lay the 5 black tiles randomly ... when 2 of them turned up very close together, the person paying for the job was upset and said 'why are these two so close together' ... the answer: you said to lay them randomly and I did.

 

[Bigv32]

Lol it has been a long long time, but I too use to feel that the odds being displayed were incorrect.

However, I used world builder/new seed option and tested it over and over again, and here is what I finally came up with.

The odds are accurate (within an acceptable error of +- 1.5-2%) per game.

What I mean by that is that during an average game (for me on an earth map as Rome) I ended up having between 1000-1500 combat situations. Note: I got these numbers by adding up the total units killed and lost.

Now I used world builder to have 2000 axe on axe battles and in each group of 2000 battles I got between +-2% at most.

The reason I have the error per game could probably be better stated error per seed.

Anyway, however, when I added all the results together, I got a very negligible error. Therefore in the long run, the odds are correct.

The only problem with any massive test involving thousands of tests does not really help a person in the short term. What I mean by this is you can flip a coin ten times but you will more likely than not not get a 50 50 heads to tails.

My personaly opinion of the odds is that overall they are accurate, but I am personally only ever going to remember the battles that I lose when the odds say I should not.

The only beef I still have with the odds system is when it rounds up to 100% and I still lose lol.