Why do I have such bad luck?

[ElMikkino]

My luck just has never been good in Civ 4. It always seems like I lose more battles than I win at around the 50%-70% range, especially when I am doing anything that needs to be done, or the consequences will be pretty bad. Lets take my current game, for instance: I've lost around 7 out of 9 or so battles at 60%-70% when I am attacking (haven't bothered to check the ones where I am defending), and I even lost 2 battles where I had 80.1% chance to win, and 89.1% chance to win, respectively. And this bad luck isn't usually counterbalanced by good luck at the other end of the scale, it seems. In fact, I really can only remember two times where I won below 50%, one at around 46.5%, or something like that, and one where I had around a 36.6% chance to win (which actually led me to capture a city, though). This bad luck is also starting to impact me psychologically, making me not want to play Civ as much, which is unfortunate, since I don't have many other games I can play, since my comp's fan is broken (Civ isn't as hard on it as Rome or Medieval II Total War, and the lag is less noticeable.). I'm starting to even think its a bug in my game, since I think I'd call myself lucky with almost any other game, and real life. I wish I just actually won most battles around the 50%-70% range...

 

[jeffreyac]

It happens, sometimes....

I was playing my last game - lost a critical unit (a legendary unit, playing with the Legends of Revolutions mod) on a 95% chance on attack, then lost 4 out of 5 at 75% odds, then a little later lost 6 straight 40% odds (I know, you're thinking 'Well, at 40%, that's not so bad... the catch is the first 5 failed to do any damage at all, leaving the defenders in perfect health... )

Of ocurse, then there was the game before, when I defeated Rome in an early warrior rush when my last warrior beat his last defender on only a 9.5% win chance, to secure my own continent early in the game... So just try and remember that it goes both ways!

 

[Ataxerxes]

A lot of times it feels like you just get nailed The RNG gods can play evil tricks on us. However, we often remember the bad and take the good for granted - at least I notice that in myself.

That said, the percentages the game gives you aren't always totally accurate. I don't believe things like first strikes are taken into account when the game gives you win/loss percentages.

 

[AveiMil]

You don't have bad luck, it's called variance.

 

[Jerrymander]

Or, invariance, since he lost nearly every 70% battle.

 

[AveiMil]

You wouldn't last very long in Poker if you were upset at loosing 5 70% draws in a row.

 

[Scoottr]

Feels like I lose more often when my chances are in the 70's than if they are in the 60's. More so if in the 90's than if in the 80's....just feels that way.

 

[sampo]

Siege is the answer for bad luck!

 

[ruff_hi]

You wouldn't last very long in Poker if you were upset at loosing 5 70% draws in a row.

no one lasts long in poker if you lose 5 70% draws in a row.

 

[AveiMil]

Actually when playing cash games good poker players should not be troubled by loosing 10 70% draws in a row without it having drastic negative effects on your bankroll. If it does you're investing too much relative to your bankroll in each hand. This has a few not-so-obvious parallels to Civ4 actually.

 

[Earthling]

Luck will even out over time, but one of the problems psychologically and game-balance wise is that civ IV combat is VERY dependent on tipping points. The first combat out of 10 or 15 combats between two stacks can determine whether like a following 13 of those battles are won or lost. Losing a top defender/attacker at poor odds then means the rest of your army has a huge blow against it.

Part of the problem is the lack of control of course and not having overall stack-based battles. Since I don't think the latter would work, I favor fixes like giving Seige units more predictable bombard abilities, giving cavalry units abilities to flank (attack weakest unit in stack if no anti-horse defenders left) and stuff like that.

Suicide catas and just giving CR promotions does get tedious, and it does get really annoying when the CG3 Longbowman combat goes wrong - when on defense you lose that defender at like 2% odds, or 8 of your catapults attacking don't even damage the enemy on defense at all or something.

Most annoying of all is barbs though, and it might be cool to have a mod where barb wins/losses are guaranteed based on combat percentage - never lose to a barb over 50%. But it would require adjusting barb behavior otherwise or else the purpose of barbs would be totally lost (I would spawn more "stacks/invasion groups of barbs" rather than individual wanderers)

 

[ElMikkino]

Ok, my luck reached an all time low for a bit in one of my games. I lost a battle I had a 99.5% chance of winning . I also lost a few more battles around the 75%-90.1% range, but this play session's luck went back to the odds going the way they were supposed to after a bit. But then the game crashed when I tried to open up someone's Diplomacy .

 

[RD-BH]

I have played too many games where the "reported" odds cross magnitude boundaries in execution. The numbers are either massaged (psychological effect) or erroneously calculated (human error). It could never be the RNG rolling up a 1000000:1 number sequence every game.

Now, if we could just get programmers to use the RNG to shuffle a stack of outcomes.
You could draw numbers until one crosses a magnitude boundary, then reshuffle the stack.
Maybe not very realistic, but more satisfying psychologically.

... or maybe the machine gods just don't like me

 

[ElMikkino]

I have played too many games where the "reported" odds cross magnitude boundaries in execution. The numbers are either massaged (psychological effect) or erroneously calculated (human error). It could never be the RNG rolling up a 1000000:1 number sequence every game.

Now, if we could just get programmers to use the RNG to shuffle a stack of outcomes.
You could draw numbers until one crosses a magnitude boundary, then reshuffle the stack.
Maybe not very realistic, but more satisfying psychologically.

... or maybe the machine gods just don't like me

No, I am very certain that that one battle was 99.5%, but maybe its not because of luck that time, rather a unit that just seemingly can't be destroyed, like this Rifleman that was guarding a barb city. I attacked him around 7 times, all with favourable odds, and around 4 in the 90-99% range, and he still did not die, even after taking damage from TWO nukes! He only died after a THIRD nuke hit him and killed him. Though, I think this happened because he was a beserker I worldbuilt in before the barbs upgraded him. I don't really know, though...

 

[RD-BH]

No, I am very certain that that one battle was 99.5%, but maybe its not because of luck that time, rather a unit that just seemingly can't be destroyed, like this Rifleman that was guarding a barb city. I attacked him around 7 times, all with favourable odds, and around 4 in the 90-99% range, and he still did not die, even after taking damage from TWO nukes! He only died after a THIRD nuke hit him and killed him. Though, I think this happened because he was a beserker I worldbuilt in before the barbs upgraded him. I don't really know, though...

A question or two or three then:

Q: When I lose a stack of troops against a lone defender with 62+% odds, and the defender does not sustain any damage (less than 8% chance according to extended odds in Buffy 3.19.003) ...
... Is there an error in the odds as reported?
... Hidden data in the calculation?
... Did the RNG roll the right number 12 times in a row?

Example:
12 vs 1 @ 62% odds of victory
62% is in range 7/12 .. 8/12 victories
Magnitude boundary is in range 2/12 .. 3/12 attacker victories
0/12 victories is below the boundary

Defender is expected to win 4/12 .. 5/12 times
From attacker PoV,
... defender victories could reach as high as 10/12
From defender PoV, (if permitted to defend against all 12 attacks)
... defender victories could fall as low as 2/12
12/12 defender victories crosses a magnitude boundary
Something is out of order ...

It isn't impossible for this to happen, but when it happens repeatedly, it starts to feel "cooked".

 

[Jerrymander]

0/12 victories is below the boundary
[...]
12/12 defender victories crosses a magnitude boundary
Something is out of order ...

That's not, conceptually, how it works. 100% defender victories never crosses a 'magnitude boundary' (whatever that is ) unless the chance for one side's victory is exactly 0%. Each battle is a separate, unique event that is not causally affected by previous battles.

With 38% chance to win a battle, the chance of a defender winning 12 battles in a row is 0.38^12 or 0.0009%. It's not statistically impossible, even thought its statistically unlikely. It's akin to winning the lottery by picking random numbers. It happens.

Unless magnitude boundary refers to the standard deviation - in which case, yes, it crosses the 'magnitude boundary' rolleyes, but that fails to prove anything because it's entirely possible that every battle will fail to fall within the 65%/97% ranges.

 

[RD-BH]

magnitude = exponent

10^1/2 .. 10^1 .. 10^2

When an observed event crosses a magnitude boundary from the predicted odds, there is usually a mathematical error in the prediction equation.

You go from "apples to apples" to "apples to oranges".
... or more to the point, from "square feet" to "cubic feet"?

 

[ParadigmShifter]

That's just wrong.

What are the odds this weeks lottery numbers were picked? Pretty small.

And magnitude is the absolute value. It has nothing to do with exponent.

x^1/2 > x^2 if 0 < x < 1

 

[RD-BH]

That's just wrong.

What are the odds this weeks lottery numbers were picked? Pretty small.

And magnitude is the absolute value. It has nothing to do with exponent.

x^1/2 > x^2 if 0 < x < 1

I realize, as an older person, mathematical naming conventions change from time to time, but an order of magnitude is still an exponent.

BtW, your example of lottery numbers doesn't cross a magnitude boundary.

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

 

[ParadigmShifter]

Ah OK you are referring to order of magnitude as log10 of a number, I see.

I'm not familiar with that as applied to statistical events though (and I did 2 years of stats during my maths degree). It is several standard deviations away from the mean however.

I'd be surprised if you lost 12 attackers against 1 defender at 62% odds though. Are you sure you didn't try attacking once 12 times and reloading inbetween (tedious though that sounds)? If you don't check "new random seed on reload" you always get the same result.

I'm not sure if CivIV uses the C++ rand() function or the Python Mersenne Twister method. rand() can be fairly clumpy, I'm not really a fan of it myself.