Civ IV combat odds are rigged.

[Kid R]

Firaxis may have put information into the games code that is not possible for modders to see or have access to.

Reminds me of the old days on 3270 ("black and green") terminals where you could actually write code which didn't show up on the screen. Basically write it in black text on the black background.

Why would they do that?

To get a kick out of watching new a guy on the team trying to find the bug

 

[Skallagrimson]

Pure coincidence and/or selective memory/bias.

A 70% battle is always 70% chance for victory (or very very close based on how good RNG is). There are no exceptions...not relative strengths, not what kind of bonuses led there, nothing...although if you get to 70% using drill vs combat you will have slightly different expected final HP outcomes, the 70% to win or lose is nevertheless accurate.

On electronic paper maybe, but that's not how it truly plays out. 80% odds is supposed to mean 1 in 5 battles you lose. I do lose some at 80% but it's nowhere near 1 in 5, more like 1 in 20 or so. If I were using selective memory I'd remember the 1 in 20 as if it were 1 in 5 as it's more painful to lose.

The converse is 50% odds. I could run 100 fights at 50% odds using the Civ RNG and I guarantee you I'd not win more than 20 of them.

That's just in general odds how they play out, but when you look at units that have bonuses results get even more skewed. Pikemen with Formation are my favorite example because they show up at usually in the mid-40s of % odds against a cavalry depending on what promotions the cav has. That numerically does factor in the pikeman's formation bonus against mounted units. WITH formation... 40% odds. I can bet you a good sum of money that the pikeman will win more than 60% of the time, even in my favor (where here again selective memory would be telling me the opposite).

The principle behind what I'm observing here is in the level of micro-probability versus macro-probability. I can go to a casino and play a game that advertises a 90% payback rate, and that's great because in my mind I can walk in with $1,000 and feel guaranteed to walk out with $900 of it. Not so. The micro-probability of winning is 90% at a given execution of the odds, but the macro-probability is that the House is guaranteed to get most of the money you walked in with, not because you have "selective memory" of winning or losing, but because probability in micro will skew over time in macro. You walk out with just enough change for a buffet meal if it's your lucky day, and that's how the big stats geeks that work for the casino, programmed it.

Combat odds in civ is micro-probability, but behavior in macro gets skewed. RNG seeding could be a part of it, or it could be anything really.

 

[r_rolo1]

The converse is 50% odds. I could run 100 fights at 50% odds using the Civ RNG and I guarantee you I'd not win more than 20 of them.

Waiting with that with popcorns in hand... It would be a first in the history of the game, so I want to see it.

 

[Ghpstage]

Micro and macro probability, wth? This is getting silly now.

I can bet you a good sum of money that the pikeman will win more than 60% of the time, even in my favor

Test was 100 Cavalry vs 100 Agg Pikes.
Admittedly I didn't give the Pikes formation as it would have been too time consuming, but the Cavalry lacks any promotions too.

Test setup

Spoiler :

Given odds were about 72% wins, and 80% survival. Bonuses against Cavs were 110% which according ot you will skew results in their favour a lot

After the fight

Spoiler :

82 survivors in all, pretty much as expected....

Stat screen

Spoiler :

70 kills in all with 12 withdrawals. Not exactly a far cry from the 72 and 8 we expected is it?

If I were using selective memory I'd remember the 1 in 20 as if it were 1 in 5 as it's more painful to lose.

Selective memory isn't restricted to odds that go against you, I myself remember one epic fluke where my Warrior took a Longbow defended city with a Castle.

 

[Antilogic]

You really should try the "new seed after reload" get your stack of six and have them attack and reload and see how you get different results. I used to do the same thing when I didn't want ANYTHING to die. Now I don't care and sacrifice whatever.

Hehehe...we used to call it Jason-probability, after a notorious abuser of it in Civ2. What he wanted to happen would happen in some universe of his game, and that is the only one he wanted to play.

but yeah ... the RNG it game uses is kinda streaky, prone to roll out odd strings of unlikely loses/wins every now and then, but macroviewing its tested and true

This is also a form of human bias--since you know the numbers are random, you constantly expect them to be different. But if you flip a coin and get 10 heads in a row, the chance the next flip is a heads is still 50/50. Streaks happen, even long ones.

 

[oldskald]

This is also a form of human bias--since you know the numbers are random, you constantly expect them to be different. But if you flip a coin and get 10 heads in a row, the chance the next flip is a heads is still 50/50. Streaks happen, even long ones.

There was an observed behaviour with the National Lottery in the UK. If a jackpot was claimed there would be an increase the following week in the number of people who picked the previous week's jackpot numbers on their tickets. Commentators mocked this behaviour, but it was actually no more stupid than picking any other random 6 numbers - there was the same chance that the previous week's numbers would repeat than any other set of 6.

 

[Kid R]

^ I think it was valid if they were mocked. Those numbers would have poor payout odds if a lot of people were betting on them. Imagine winning the jackpot and having to share it with a thousand other people

 

[oldskald]

^ I think it was valid if they were mocked. Those numbers would have poor payout odds if a lot of people were betting on them. Imagine winning the jackpot and having to share it with a thousand other people

A fair point... I should probably re-write to say "mocked more than usual for spending hard-earned money on a 14 million-to-one chance of winning". It isn't colloquially known as "idiot tax" for nothing.

 

[Kid R]

Hehe, plus that damn Lancelot machine is RIGGED I tell you! I always seem to lose on him.

A tip for lottery players: the machines can see if you part-filled out your ticket earlier in the week, and they will go for those numbers more.

Spoiler :

(...nod to the recent "wonder building cheat" thread )

 

[Sinyail]

Well I'm just checking back on this thread after having been absent from Civ IV for a couple of months and am glad it got some attention. Sorry, Paradigm_Shifter. I was a little angry that day. I believe now that the Civ IV RNG is not rigged. For many reasons. One of the reasons is sometimes when I attack the ai with about 15% chances or even 5% chances I win. Definitely is not very common, but it does happen. I just don't notice it as much. Thanks GHP_Stage and everyone else

 

[Sinyail]

Now I've learned that the whole "Civ IV combat odds are rigged!" phenomena is quite common, but, to a large extent, has been refuted. Inevitably, though, many people will be angry and claim the same when they see their horse archer get defeated by a archer when the odds were in the horse's favor by 90% hehe

 

[Sinyail]

This thread got 1600 views? Whoah seems to be on a lot of peoples minds haha

 

[Sinyail]

Much of the frustration must be simple human emotion and ego. Many of us humans really have little to no self control lol

 

[ruff_hi]

What I'm specifically talking about though is a pattern.

I tested this issue in this thread.

 

[AlphaShard]

Well if there's a ten percent chance of failure there is a chance of failure actually happening. You won't always roll those twenties and eventually you will roll ones.

 

[Windsor]

This thread got 1600 views? Whoah seems to be on a lot of peoples minds haha

The number of views and the frequency of threads such as this shows that Civ5 did make a step forward with combat odds. Getting "decisive victory", "stalemate" and so on instead of a % number is much better because of how the human brain works.

The combat odds in Civ4 may not be rigged, but as long as they give the player a feeling of being cheated it's not good.

 

[vincentz]

hehe. I remember a bug in one of the RevDCM versions (or was it BBAI cant remember) where the combat odds got bugged. It would show one thing and do another.
Not saying its like that in BtS, but not saying a bug hasnt been implemented in a mod or two

 

[Ataxerxes]

This thread got 1600 views? Whoah seems to be on a lot of peoples minds haha

It's often an interesting topic, like some of the things in off-topic. Also, the thread was dormant for months, which often leads people to look at the same thread multiple times. I do agree that it's on people's minds more than it should be. They have an emotional reaction of "I got a 1% event that was negative, the system is rigged!" Understanding probability doesn't come naturally and a lot of people never get it - witness the success of Las Vegas.

 

[Kid R]

Civ5 did make a step forward with combat odds. Getting "decisive victory", "stalemate" and so on instead of a % number is much better because of how the human brain works.

Playing a game shouldn't require a maths degree to understand what's happening, at least for something you do virtually every turn like combat. User interfaces in general should adhere to the "Principle of Least Astonishment":

http://en.wikipedia.org/wiki/Principle_of_least_astonishment

... and astonishment is a pretty good word for how many people feel about these unlikely outcomes. Taking some of the rough edges off the uncertainty like they did in 5 makes a slicker game experience and loses zero game depth. Perhaps even adds some tactical depth.

Not saying there shouldn't be a ton of low-level maths in the game that experts and number lovers can delve into of course. All that stuff's great so long as the general principles are obvious, e.g. build a library and it's generally a good for your tech, or whatever.

 

[Antilogic]

hehe. I remember a bug in one of the RevDCM versions (or was it BBAI cant remember) where the combat odds got bugged. It would show one thing and do another.
Not saying its like that in BtS, but not saying a bug hasnt been implemented in a mod or two

In the original vanilla Civ4, first strikes were not appropriately factored in, so sometimes you would have a 104% chance of winning, sometimes it would say 100% due to rounding but you would still lose.

Playing a game shouldn't require a maths degree to understand what's happening, at least for something you do virtually every turn like combat. User interfaces in general should adhere to the "Principle of Least Astonishment":

http://en.wikipedia.org/wiki/Principle_of_least_astonishment

... and astonishment is a pretty good word for how many people feel about these unlikely outcomes. Taking some of the rough edges off the uncertainty like they did in 5 makes a slicker game experience and loses zero game depth. Perhaps even adds some tactical depth.

Not saying there shouldn't be a ton of low-level maths in the game that experts and number lovers can delve into of course. All that stuff's great so long as the general principles are obvious, e.g. build a library and it's generally a good for your tech, or whatever.

I don't know too much about Civ5's combat system, but it doesn't take a degree in any sort of mathematics to know that 60% or even an 80% chance of victory is not a guaranteed win. People get frustrated when they lose battles they were favored to win, but it is a possible outcome. What I'm more concerned with is whether or not it is a fair system, as in I can choose the terrain to fight on or units to build to gain an advantage, and whether it is a transparent system, meaning I know all the factors affecting the outcome of the battle. Civ4 passes both of these tests.

The "random" numbers used in the Civ engine is an abstraction representing leadership, supplies, tactics, etc. I would love it if the Civ combat engine took these factors into account like in the AGEod games (or at least the Paradox games) instead of leaving them to be randomly determined. But that would be a very different game from Civ.