Wow my computer is lucky!

[Dubai Vol]

Forgive a newbie, I did a search....

So I attacked at 94.5% odds and lost. OK, I've rolled my share of 20s playing D&D. So I hit him again, and this time the odds are better: 98.7%, and would you believe it, I lost again! I'm no mathematician, but I feel lucky to have witnessed that thousand-to-one shot! (1500+ to 1 if my simple math skils are up to it.)

So OK it happens, but somehow it seems to happen a lot more often than it should. I've got to where I clench my cheeks at anythng less than 99.9% because when the computer tells me I have a 66% chance I lose two of three times, and a 75% chance seems to go wrong 3 out of four times. I'm one of those skeptics who think computer-generated random numbers aren't random at all (rem the "preserve random seed" option?)

Is this a bug or am I just a paranoid loser?

 

[Arlborn]

Who knows?

Ths is an old topic. Some swears the random numbers are wrong, som swears they are completely right and you are just unlucky...Who knows, oh who knows?

 

[morchuflex]

Selective memory.

 

[Severus]

Actually there was a thread about losing at high 90% odds a long time ago where someone who understands computer coding and random number generators and the like came to the conclusion you're pretty likely to lose again at very high odds if you've just lost the last battle at very high odds, i think he said something about how the computer clumps the numbers together or something but then again i can't really remember.

Of course it's a case of selective memory, you're always going to remember the battles you lose at high odds just like you'll remember the battles your catapult won against a longbow at around 10%. I'd expect to lose a unit attacking at 75% odds fairly regularly too, after it all it should lose once in every four battles, and theres a chance it might lose more.

Just don't let it bother you, it happens to everyone.

 

[FatSean]

Selective memory may play a part, but the best random number generation can have trends as well.

 

[flamingzaroc121]

if you were attacking a unit with a first strike, the combat thing doesnt take that into account so that always messes me up

 

[Bushface]

While I'm convinced that the RNG output has a tendency to be streaky, with human memory being what it is you are much more likely to remember a bad streak (when you lose a series of battles although the odds were strongly in your favour, which you don't expect to happen) than a good one. Indeed, you are unlikely to notice a good streak at all, because who would risk a series of attacks at, say, 1% odds, just to see if you had actually hit a good streak ? Theoretically, your chance of winning at 1% odds is the same as your chance of losing at 99% odds: both have been reported in this forum, but losing at 99% far more frequently.

 

[beansbeans]

Actually there was a thread about losing at high 90% odds a long time ago where someone who understands computer coding and random number generators and the like came to the conclusion you're pretty likely to lose again at very high odds if you've just lost the last battle at very high odds, i think he said something about how the computer clumps the numbers together or something but then again i can't really remember.

I agree with this theory, but I'm not sure about the "clumping" part. I'm not an expert game programmer but I have been programming for a couple years.

 

[Leeksoup]

I have once had a galley beat up a battleship.

It was awesome.

 

[morchuflex]

I have once had a galley beat up a battleship.

It was awesome.

That was routine in Civ1, but unheard of since then, unless the BB had only 1 HP left...

 

[Chieftess]

Well, mathematically, there was a 1.3% chance that you would lose.

 

[Psyringe]

No evidence of any AI bias in combat has ever been found.

This should tell something since it's not difficult to test. I ran a test one year ago, you can find it in this post. The test is concerned with battles against the AI with 50% winning chance on Noble, but it's not difficult to test other hypotheses with the same method. I'm not sure whether the save still works though, it was done with pre-patched vanilla Civ.

 

[Stylesrj]

Hmm a galley beating a battleship. Reminds me of the time in Civ 1 many years ago when an artillery attacked my settler and lost (I was still in the Bronze Age) or when an Ironclad attacked a militia in a city and lost. Those were the days.... -sigh- Oh well Civ 1 sucks

 

[Joebasalt13]

Well I don't complain as I attack with any odds from .07%-99.9% so doesn't bother me.

 

[kuldiin]

I had a 95% lose, then a 99% lose, then a 1% (I threw in anger) win lol.

Although it died in the next attack

 

[ferenginar]

I think it would be much better if the percentage chance of victory were removed, you just can't help looking at it then complaining when you feel the combat has gone against you.

I hardly ever attack unless my percentage is over 70%, except from those suicide cats.

 

[largedarryl]

I remember playing Civ4 vanilla before any patches came out. They didn't have the percentage then, but I recall lots of people complaining because they didn't know the chances of winning. So there was a percentage added in one of the early patches.

 

[Psyringe]

Yes. Before the percentage display was added, there were lots of people complaining about "unfair" combat when in fact they just hadn't understood the power of, say, city garrison promotions.

 

[Exel]

Still, I think the exact percentage display could be replaced with a more abstract display of combat odds, like "BAD / FAIR / EVEN / GOOD / EXCELLENT". Maybe it's just me, but the exact percentages make the battles feel more like the number crunching they are rather than real combat between units.

 

[Psyringe]

For me, the number crunching would actually start when the percentages *weren't* given. Because when I just see something like "good" odds, then I'd try to calculate them better because I want to make an informed decision whether or not to attack. So I'd start doing the math in my head, calculating strengths and bonuses etc. A game that simply displays the percentage saves me from a *lot* of number crunching.