Combat.... say WHAT!?

[Brutus66]

After an hour and a half of play in which I lost way too many battles where I had odds of 90% or better, I too am suspicious as to what's going on. Sometimes it sure seems like the computer is throwing loaded dice.
I am assuming combat results are decided by a random number generator, but I sometimes have real doubts that it's being properly randomized.
Is this how it works? http://www.virtualp.us/Dilbert-Oct_25_001.jpg

[Solver]

The computer doesn't cheat as far as combat odds go. It doesn't receive a bonus like that. Of course, it goes without saying that, when you fight enough battles, you'll lose some at 95% - and win some at 3%.

From a programming perspective, though, the random numbers are, of course, pseudo-random. They do appear random but are never truly so... which is a limitation of computers per se. But for all intents and purposes in a game of Civ4, the random numbers are indeed random.

[Kuciwalker]

People, if you really think the odds are fudged, show me where. You have the source code.

[Brutus66]

I'm not a programmer. I just know what I see.
I try very hard to remain objective, but when I lose several battles with more that 90% odds in my favor in a short period of time, I get to wondering.
I have become fatalistic now. If the odds look good, I do the battle, but I am no longer surprised when I lose no matter how favorable the odds were. I have decided not to attach a lot of significance to the numbers anymore.

[Solver]

Well, if you're not a programmer, then what you see is your subjective impression. You humans are faulted that way, you're biased. We computers see things the way they are

Anyway, with a proper strategy, if you lose at high odds, it won't really matter. Losing one or two units where you shouldn't, odds-wise, shouldn't be enough to stop your assault.

[Jaybe]

If the odds look good, I do the battle, but I am no longer surprised when I lose no matter how favorable the odds were. I have decided not to attach a lot of significance to the numbers anymore.

That's a GOOD attitude towards Civ battles, and should not be labeled "fatalistic".

A leader in the field would not have any percentages at all, though he might have other things instead, some of them misleading. Because we DO have them (and they are basically ALL we have), we tend to give them oracle status.

[Brutus66]

Originally posted by Solver
Anyway, with a proper strategy, if you lose at high odds, it won't really matter. Losing one or two units where you shouldn't, odds-wise, shouldn't be enough to stop your assault.

No, it certainly doesn't slow my offensive to any significant degree.
In this particular game (noble level), I have six military instructors in London, which along with building improvements and civics give each tank that rolls off that city's assembly line some some pretty darn respectable promotions.
It just ticks me off to lose them while trying to winkle mediocre riflemen out of cities.

[Solver]

Hey, what general in history hasn't been "ticked off" at having to lose men, some of them experienced veterans, as his offensive advances?

[zeace]

Ok folks, I'm going to try and use some of the statistics I learned in college and apply it to civ battles:

In this first table I monitor a unit that is fighting battles with a 95% chance of winning. The last column shows the probability of this unit surviving past that battle. There's two components to it: Chance of dying in that battle (5%) and the chance of being killed any time previous to this.

95% Battle Odds
Battle Chance of Dying in battle Chance of Surving past battle
1 5.00% 95.00%
2 4.75% 90.25% (5%* 95%, 95% - that)
3 4.51% 85.74%
4 4.29% 81.45%
5 4.07% 77.38%
6 3.87% 73.51%
7 3.68% 69.83%
8 3.49% 66.34%
9 3.32% 63.02%
10 3.15% 59.87%
11 2.99% 56.88%
12 2.84% 54.04%
13 2.70% 51.33%
14 2.57% 48.77%
15 2.44% 46.33%
16 2.32% 44.01%
17 2.20% 41.81%
18 2.09% 39.72%
19 1.99% 37.74%
20 1.89% 35.85%
21 1.79% 34.06%
22 1.70% 32.35%
23 1.62% 30.74%
24 1.54% 29.20%
25 1.46% 27.74%
26 1.39% 26.35%
27 1.32% 25.03%
28 1.25% 23.78%
29 1.19% 22.59%
30 1.13% 21.46%
31 1.07% 20.39%
32 1.02% 19.37%
33 0.97% 18.40%
34 0.92% 17.48%
35 0.87% 16.61%
36 0.83% 15.78%
37 0.79% 14.99%
38 0.75% 14.24%
39 0.71% 13.53%
40 0.68% 12.85%
41 0.64% 12.21%
42 0.61% 11.60%
43 0.58% 11.02%
44 0.55% 10.47%
45 0.52% 9.94%
46 0.50% 9.45%
47 0.47% 8.97%
48 0.45% 8.53%
49 0.43% 8.10%
50 0.40% 7.69%
51 0.38% 7.31%
52 0.37% 6.94%
53 0.35% 6.60%
54 0.33% 6.27%
55 0.31% 5.95%
56 0.30% 5.66%
57 0.28% 5.37%
58 0.27% 5.10%
59 0.26% 4.85%
60 0.24% 4.61%

Notice that the unit only has around a 50% chance of making it past his 13th battle, even though he only has a 1 in 20 chance of dying. Also notice that he has almost a 5% chance of lasting 60 battles. So even though half of the time this unit dies before his 13th battle, the chance of making it to 60 or more battles ensures the average will be 1 in 20.


If we look at 90%:

90% Battle Odds
Battle Chance of Dying in battle Chance of Surving past battle
1 10.00% 90.00%
2 9.00% 81.00%
3 8.10% 72.90%
4 7.29% 65.61%
5 6.56% 59.05%
6 5.90% 53.14%
7 5.31% 47.83%
8 4.78% 43.05%
9 4.30% 38.74%
10 3.87% 34.87%
11 3.49% 31.38%
12 3.14% 28.24%
13 2.82% 25.42%
14 2.54% 22.88%
15 2.29% 20.59%
16 2.06% 18.53%
17 1.85% 16.68%
18 1.67% 15.01%
19 1.50% 13.51%
20 1.35% 12.16%

This unit beats the odds just by surviving 7 battles. Of course this unit has a 1 in 20 chance of seeing nearly 30 battles.


So any unit (i.e. a Warlord) that you are using in low 90% battles is not going to last very long. What if we only attack at 98%:

98% Battle Odds
Battle Chance of Dying in battle Chance of Surving past battle
1 2.00% 98.00%
2 1.96% 96.04%
3 1.92% 94.12%
4 1.88% 92.24%
5 1.84% 90.39%
10 1.67% 81.71%
15 1.51% 73.86%
20 1.36% 66.76%
25 1.23% 60.35%
30 1.11% 54.55%
35 1.01% 49.31%
40 0.91% 44.57%
45 0.82% 40.29%
50 0.74% 36.42%
55 0.67% 32.92%
60 0.61% 29.76%


This unit will last less than 35 battles more often than not. On the other hand it has a reasonable chance of surviving for a very long time. Still, one time out of 10 it will see less than 6 battles. So even when fighting "sure thing" battles like this it is not incredibly unlikely that you'll lose this unit very quickly.

---

The special nature of the warlord unit makes us all too aware when he meets an untimely demise, but the truth is that even if we only attack at 98% odds this unit is not going to last forever, and will probably not last longer than a war or two. There's always the off-chance that your early axeman warlords will win hundreds of 98% battles and will last the entire game, and this offsets the likelihood of an early departure.

So stop whining about the odds. I agree that losing your warlord on his first battle at 99% sucks, but the computers not cheating...

[Solver]

Uhm. The very first thing about statistics is that odds should add up to 100%. The odds in your tables stop doing that very early on.

[Jaybe]

Solver, the first column is the chance of the unit expiring in THAT battle -- note the "(5%* 95%, 95% - that)" at battle 2 @95%. It works out.

[Solver]

Ah, I misunderstood it then.

[Kuciwalker]

zeace

[wodan11]

Brutus,

One thing to realize is that early versions (without the patches, Warlords, and BetterAI mod) do not properly calculate battles. Especially first strike.

So, without additional data, I'd say chances are good that you're playing with unmodified CIV, which means the odds calculator in the game isn't giving you good information....

Wodan

[Six Thousand Year Old Man]

Has anyone tested the odds at the bottom of the list?

I mean, if one wins more than expected with 1% odds, then the situation balances out somewhat, and this phenomenon stops being "the game is stacked in favour of the AI" and starts being "the odds aren't calculated right". Which is annoying, but less annoying than thinking that the game is stacked against you.

I know that my suicide Catapults win a lot more of those 1%-5% battles than I expect them to.