More combat model tests

[TheNiceOne]

Here's some more tests of Firaxis' suggested new combat model, now also tested with 2 and 3 numbers in addition to 4.

The numbers below show tests of how two veteran units facing each other will behave. The defender is always fortified on plains.
The columns mean the following:
rolls: Number of rolls to take the average roll from. 1 is the current combat model, and 4 is what Firaxis intended, but ditched from the beta patch.
A: Attacker's attack factor.
D: Defender's defense factor, modified by terrain and being fortyfied
new A: The attacker's effective new attack factor. I.e. A new A of 9.3 for an immortal vs a warrior is identical to having A of 9.3 with the old (current) combat system.
A win%: The win chance of a veteran attacker against a veteran defender (both starting at full strength).

rolls   A     D      new A   A win % (veterans)
(Immortal vs warrior)
  1     4     1.35     4      92.6
  2     4     1.35     9.3    99.3
  3     4     1.35    17.2    99.9
  4     4     1.35    29.8    100.0

(Immortal vs spearman)
  1     4     2.7      4      70.4
  2     4     2.7      5.6    83.9
  3     4     2.7      6.6    88.6
  4     4     2.7      7.8    92.2

(Immortal vs pikeman)
  1     4     4.05     4      49.3
  2     4     4.05     4.0    48.7
  3     4     4.05     3.9    48.5
  4     4     4.05     3.9    48.1

(Archer vs spearman)
  1     2     2.7      2      34.0
  2     2     2.7      1.5    22.1
  3     2     2.7      1.4    17.7
  4     2     2.7      1.2    13.8

(Swordman vs spearman)
  1     3     2.7      3      55.7
  2     3     2.7      3.3    61.2
  3     3     2.7      3.4    62.7
  4     3     2.7      3.6    65.0

(Swordman vs pikeman)
  1     3     4.05     3      34.0
  2     3     4.05     2.3    22.1
  3     3     4.05     2.1    17.8
  4     3     4.05     1.8    13.8

 

[socralynnek]

So, the persians would rule the world!

Nice work!

Is anyone in favor of this model?

 

[HellboundHeathcliff]

I think 2 rolls would do what they want. It would certainly make a noticable difference.

 

[anarres]

Originally posted by watorrey
I think 2 rolls would do what they want. It would certainly make a noticable difference.

It would make a difference, but not in the way this change was proposed for. The change was to reduce the percieved streakyness of the RNG, when in fact this just makes likely things more likely and unlikely things less likely.

 

[HellboundHeathcliff]

By 'percieved streakyness' i am guessing you mean, as an example, a knight attacking and losing to a longbowman and the longbowman never takes damage?

 

[Master Shake]

Percieved streakiness refers to a perception that rather tahn wins and losses being completely random, rather, tehy come in streaks. Like as though you were flipping a coin and you tended to get several heads in a row, several tails, several heads, and so on--almost never having for example an HTH or THT sequence.

-mS

 

[wilbill]

This steakiness is "perceived" because it's a predictable part of randomness. But a lot of players perceive it as being something wrong with the RNG.
@ TheNiceOne - thanks for the number crunching.

 

[anarres]

hehe. I spelt 'perceived' wrong and only wilbil spotted it.

As others have said, 'streakiness' is inherent in random number generation. The golden rule to remember is that one extreme result does nothing to affect the next result.

Randomness means you will get streaks of extreme numbers, whereas a human who writes a list of what they consider to be 'random numbers' will not have as much streakiness as a real RNG. Humans are flawed, tis all.

 

[syndicatedragon]

The perception is enhanced because we only see such a small sample of rolls at a time (i.e. units typically have only 4 or 5 hp). If you were to look at all the numbers over the course of a game, or the course of several games, I'd bet that everything looks like what people would expect. I personally think combat is fine as is. It's frustrating to lose when you think you should win, but if you won every time you thought you should win, it would take a lot of strategy out of the game IMO. Part of strategy is adapting to a new/unexpected situation...

 

[Commander Bello]

I guess, the "perception problem" is caused by an additional factor, about which I haven't read anything until now:

Random figures may build any consecutive row, thus making even the strangest results to happen. That is, what all of us are clear about.
Why does this happen then almost only to the AI? Is there a hidden cheat functionality?
No. It is just, that the most of human players don't make "silly" moves. The least of them will attack fortified musketmen on a hill with some spearmen, and so on.

That means, the human player in most times doesn't allow "his" RNG to create those strange results. On the other hand, if the human musketman wins over the attacking horseman, you won't see, if you have won with an internal 3 or with an internal 4.
So, you can't perceive the row of 10 consecutive 4's (in case you had a stack of muskets being under attack).

This just means, that there is a good portion of psychology in the whole case.

 

[SirPleb]

Originally posted by TheNiceOne
Here's some more tests of Firaxis' suggested new combat model, now also tested with 2 and 3 numbers in addition to 4.

Thanks TheNiceOne! The results for 2 and 3 rolls suggest strongly, to me, that not even an increase to 2 rolls should be made without either:
1) Making the number of rolls a parameter, with the default being 1, or
2) Re-playtest all units and rebalance as required for the new combat results

Originally posted by socralynnek
So, the persians would rule the world!

Pretty much, unless they came up against one of the few other Civs who get enhanced by the change. Most Civs, being more "average" in Ancient Times, would suffer at the hands of those strengthened Civs. This was discussed on the "Conquests Patch Notice" thread a few times. (I think I was first to mention the Persians becoming even more powerful, but only because I was up in the middle of the night when the thread got going )

Originally posted by Commander Bello
Why does this happen then almost only to the AI?

It doesn't, and your subsequent explanation thereof isn't necessary. Complaints about a player's Tank failing to defeat a Spearman, or a number of somewhat stronger units being defeated by one somewhat weaker defender, are common. Those complaints are the result of exactly this same thing.

 

[The Last Conformist]

SirPleb: I think you got Commander Bello's post backwards; I'm reading him to say precisely that people complain of their Tanks losing to AI Spearmen etc, but never of the opposite.

(I don't complain, but I do make a point of recounting stories of bizarre battle outcomes in my favour in "AI combat cheats" threads; my most recent outrage would be losing only two Medievals to take out a hill city defended by four Riflemen, without artillery support.)