Am I missing something or are combat odds not accurate?

[elementoffear]

I just lost 5! 80%+ combats in a row when my stack of elephants attacked a city (after bombarding).

I am not a mathematician but 80% times 5 is 0.032% I think, not really supposed to happen. It could happen theoretically I guess, but its not the first time I lost very favorable odds (although not this favorable).

Is it me , or are there some hidden modifiers I should have known about?

EOF

 

[Guess]

All you can do is to blame the Random Number Generator that controls the battle outcomes.

 

[Killroyan]

When I am warring I rage a lot and mostly I am almost sure I will win a battle at 95+%. This random number generator has screwed me over so much. Loosing 3 90% fights in a row and such. I wish I had that much luck in the casino

 

[TheMeInTeam]

The rng does funny things, especially when you fight 100's of battles per game.

My civ IV darius let's play saw me lose at odds like 95% 90% 80% etc all in a row, ultimately losing a city defended by an archer to a warrior . That's rage quit mode in SP, but in the LP I toughed it out .

 

[Sian]

as mentioned by others ... if playing long enough you're bound to get into situations you'd normally deem unlikely, simply because you played long enough

its like flipping a coin a million times noting down wheather its head or tails, getting a row of 15 heads ... unlikely as hell, but with a long enough string it'll come sooner or later ... winning 95% of the time means losing 5% of the time

 

[TheMeInTeam]

as mentioned by others ... if playing long enough you're bound to get into situations you'd normally deem unlikely, simply because you played long enough

its like flipping a coin a million times noting down wheather its head or tails, getting a row of 15 heads ... unlikely as hell, but with a long enough string it'll come sooner or later ... winning 95% of the time means losing 5% of the time

Now what I DON'T like is that the unlikely outcomes are rewarded heavily, despite 0 skill differentiation between winning and losing them.

 

[fathertuck]

The rng does funny things, especially when you fight 100's of battles per game.

My civ IV darius let's play saw me lose at odds like 95% 90% 80% etc all in a row, ultimately losing a city defended by an archer to a warrior . That's rage quit mode in SP, but in the LP I toughed it out .

Sorry, but I couldn't help laughing at your game when that happened. Talk about bad luck.

But on to the topic. Gamers tend to forget about lucky events and remembers only unlucky events that take place. If you win 4 battles at 95% - 90% - 85% - 80% odds you laught about it and go on with the game, but when you have bad luck and lose 4 battles at the same odds you curse the game and blame the rng for it. I think it has more to do with human psychology than buggy rng's.

 

[elementoffear]

Guys, I am not talking about perception or frustration.

The fact is that 80% X 5 losses is mathematically almost impossible, to make it statistically viable I would probably have to play millions of combats which I didn't in all my life, but I see these types of series many times.

My original thought is that there are some hidden modifiers so when you see 90% its not actually 90%... thats all. In fact, couldn't one write a mod that just logs all these odds and outcomes and then we see if the odds are right? maybe someone already has?

 

[Sian]

Now what I DON'T like is that the unlikely outcomes are rewarded heavily, despite 0 skill differentiation between winning and losing them.

just to be certain ... thinking of the experience gain from winning luckily?

 

[ParadigmShifter]

Guys, I am not talking about perception or frustration.

The fact is that 80% X 5 losses is mathematically almost impossible, to make it statistically viable I would probably have to play millions of combats which I didn't in all my life, but I see these types of series many times.

My original thought is that there are some hidden modifiers so when you see 90% its not actually 90%... thats all. In fact, couldn't one write a mod that just logs all these odds and outcomes and then we see if the odds are right? maybe someone already has?

p(5 losses in a row with 80% odds) = 0.2^5 = 0.00032 = 0.032%. Pretty unlikely, but by no means "impossible". It would be weird if these things never occurred with random variables.

I think ruff_hi did tests for his advanced combat odds mod, and the verdict was: the odds are accurate.

 

[tinstaafl]

Guys, I am not talking about perception or frustration.

The fact is that 80% X 5 losses is mathematically almost impossible, to make it statistically viable I would probably have to play millions of combats which I didn't in all my life, but I see these types of series many times.

My original thought is that there are some hidden modifiers so when you see 90% its not actually 90%... thats all. In fact, couldn't one write a mod that just logs all these odds and outcomes and then we see if the odds are right? maybe someone already has?

Not necessarily millions of battles. you said yourself that the chance for that series to happen is .032% , which is correct. Now, consider millions are playing this game. It stands to reason that some people WILL get screwed by the RNG, and some will be screwed over by the RNG more than others, and some will in fact get screwed over exactly when it hurts the most. Statistically speaking somebody has to be that person.
That's really the problem with randomness in games like this.

In any case, as far as i know the combat odds calculations have been both tested via worldbuilder tests and the code checked and appear mostly correct. I believe i read there was in fact once a problem with first strikes not being accounted for correctly, i don't know if that's still in.

 

[ParadigmShifter]

First strike calculations got fixed, can't remember when.

The odds do lie for "free" wins against barbs I think... I think when you have a free win in the bank the probability of winning each round of combat is fudged to 90% (not certain about that).

 

[NihilZero]

As players we never really get to see the flip side of this, which is winning several successive against the odds battles, for the simple reason that we aren't stupid enough to try attacking against such odds. So we will get to experience many times the dreaded defeats at 99% odds, but never will we taste the victories at 1% odds - unless defending, of course. And I do remember having some wildly improbable defensive wins from time to time.

Having said that, the RNG does seem unusually streaky in Civ 4, but I don't think we can ever prove that since statistically it works okay across a big enough sample of tests.

 

[ben-jammin]

If 90%+ GUARANTEED you a victory, would that make the game more or less interesting?


The only time its really frustrating is when you're in a situation where you relied completely on winning that battle with no backup, for me that's usually when I decide to attack the only enemy unit at high odds with my only unit defending a city. But that's like gambling your city on 10% odds of losing it for a return that isnt worth that bet, so its a poor strategy.

When you lose a number of high probability battles in a row....well, that's 'flavour' i guess.

 

[ParadigmShifter]

It was PieceOfMind that wrote the Advanced Combat Odds mod, not ruff_hi, so I must correct myself

 

[CaF]

Guys, I am not talking about perception or frustration.

The fact is that 80% X 5 losses is mathematically almost impossible...

The thing is that if 0,0032 is "impossible" for you, you have to readjust your definition of impossible. You have to add a truck load of zeros behind the decimal point to even reach a point close to impossible.

It is perception or frustration. That is pretty standard. I guess humans are not made to deal with stochastic outcomes. I would even go as far as to say that random events can make you sick.

My 2 cents on that one. Combat odds seem accurate and true, but I never ran tests.

 

[VoiceOfUnreason]

p(5 losses in a row with 80% odds) = 0.2^5 = 0.00032 = 0.032%. Pretty unlikely, but by no means "impossible". It would be weird if these things never occurred with random variables.

And as tinstaafl hints at, this isn't really the right math. It correctly describes the odds that your next 5 attacks will all fail.

But if you think about a typical game, you'll have a lot more than 5 attacks at 80% odds. The odds that you lose 5 attacks in a row at any point during the game are generally quite a bit higher.


The key point here is that the combat engine, by design, uses independent coin flips to resolve combat. In other words, the combat engine doesn't remember that it screwed you recently.

I believe that the engine in Civ V uses a different design, which exhibits "fairer" short term outcomes.

 

[ParadigmShifter]

There was some attempt to answer the question in my "Let's discuss mathematics" thread over in Sci-Tech, but we were thinking about coin tosses instead, and we didn't really come up with a definitive answer (for the distribution of the longest streak, or the distribution of probabilities for all streak lengths).

It's easy to write a computer program to simulate it though, I can't be bothered to do that

Any idea what RNG generator civ V uses? Have they published an SDK yet?

 

[shpeka]

Having said that, the RNG does seem unusually streaky in Civ 4...

I have this feeling too. Or is my imagination burdened by the old Colonization RNG bug.

I also have the subtle feeling that retreat odds for mounted units are not calculated correctly.

 

[ParadigmShifter]

It uses a linear congruential generator (LCG) random function (civ IV that is), which are known to be fairly streaky.

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