AI really needs to "win"...?

Currently playing a game of C3C just for fun.
Netherlands, Huge Map , 3 Sprawling Islands and a few smaller ones.
Lots of Mountains, a moderate climate and water, water everywhere...

Got Aztecs, Russians,Ottomans, Chinese, Spain, Portugal,France and Zulu as
"competition". Aztecs have their own Island, China stuck in the cold, Russia and Ottomans shared an Island with Portugal ( KIA ), I have liberated my Island from the evils Spain and Zulu.

So basically I am playing along ( afterwork ) and try something out at a level most at
civfanatics wouldn't consider ( Warlord ).

All went normal. Until...

The AI somehow got to the idea it had to beat some of my Units.
In a very odd way.

The netherlands could rule the waves with more advanced ships but I kept back and
used the naval power mostly to settle ( transports ) or to kick some barbarians.
But yes, there are dozens of military grade ships if anyone asks for a sea battle.

Had some privateers to shadow the AI units so I could see where the AI was moving . Sure there was conflict too, I mean privateers are not really transports...
So I lost some privateers when AI players got Galleons. OK fine.
What I did not expect was, these silly Galleons would attack Frigates....and win???

1 time = maybe. 2x or more shouldn't happen.
Veteran ships ( 4 wounds ) would take some damage against other veterans and the basic A and D value should count too. Am i wrong?

Seems the AI really really wanted to get past the Frigates and invaded a little Island without any ressources. The 3 Knights ( regular ) died horribly and lonely.

Still thinking about why, the convenient solution was to march into the AI cities and take the harbors from them.
No coast city, no harbor = no m0re weird results at sea0.
Now, lots of conflict made room for barbarians.
Barbarians camps spawn warriors. ( 1850 AD...).
Barbarian warrior ( regular ) threatens workers who build a "combat-railroad" to the front lines. Got Elite Cavalry ready to intercept this threat.
How could I have known my poor Cavalry was getting slaughtered without dealing out damage to the savage axe wielder on a grass tile ?

Is it possible that the AI gets desperate? What was going on there?

You talk about Privateers and then about Frigates... You mean the galleons attacked the Privateers?
Galleons have an attack of one, and Privateers have a defense of 1, so it's plausible that the AI would attack. They have no reason not to, since it doesn't start a war.

The rest... I don't know what the question is.

The result for Galleons vs privateers is understandable. As for warrior vs cavalry, well cavalry are weak on defense and even with d=3,it might just happen that a warrior defeats it (we know that spearman with d=2 can defeat tanks so this case is still far more likely. And then we're all helpless before the RnG gods....

Theov said:

You talk about Privateers and then about Frigates... You mean the galleons attacked the Privateers?
Galleons have an attack of one, and Privateers have a defense of 1, so it's plausible that the AI would attack. They have no reason not to, since it doesn't start a war.

The rest... I don't know what the question is.

Click to expand...

Sorry if I wasn't making a point.

The Galleons attacked the Frigates 1:1 and sunk them. Both started at full health and equal experience. Galleons barely took 1 Damage...

The Cavalry ( A 6 ) attacked a non fortified warrior ( D 1 ). Cavalry was elite, warrior was a regular.
Again, both started at full health. Warrior won. Terrain boni? None....

In the first case, Frigates could have intercepted a ship sent to invade an island and both got sunk "easily". Felt like a point'n'click removal...
Second case, warriors shouldn't win such a fight without boni or extreme luck IMHO.

My point was:

is it possible that a AI of a game starts to generate weird results because things in general aren't looking good for the AI controlled players?

Always thought a piece of software wouldn't mind if it wins or looses...

It is purely random. If you want your average result to be near to what would be expected, than you need to increase the numbers a lot. If you have 10000 battles, than the frequence of results is close to what can be calculated.

http://forums.civfanatics.com/showthread.php?t=60617

There's more to it, there has to be... no-one would rationally invent a combat game with margins so ludicrously wide.

Buttercup said:

There's more to it, there has to be... no-one would rationally invent a combat game with margins so ludicrously wide.

Click to expand...

I am one of those that suspects that the Random Number Generator for Combat is, at times, severely skewed in the AI favor. There is a considerable amount of anecdotal evidence to show that there are times when the AI wins at a very disproportionate rate. Unfortunately, without access to the source code, that is not possible to prove one way or the other.

I've seen too many outcomes of battles in games from the CiV series to believe its always random.

The examples of my current game are surely not "evidence" on a level of "facts".


But, we all know the Game sometimes has "unbeatable" units, like:

- Player besieges a city, orders his Units to "storm the walls" at game turn 256 and
one AI controlled unit seems to beat anything thrown at it.

- as I like the Save and Load options the Games offer, I may just go back to the beginning of the turn and try the same assault again. And again if the result doesn't change. Believe me, the most unexpected outcomes aren't randomly, they happen
again and again.

Should it surprise me if the outcome of the battle is different when the assault is delayed to game turn 257 ?

Use the same attackers and defenders to "replay". The Order of units activated and sent forth isn't important in this case.

The consistent thing is , waiting a turn alters the outcome of battle everytime a AI controlled unit was extremly successful.
The same battle in turn 256 = or turn 257 =
AI got lucky dice for a turn ....

I have run into this AI behaviour at least 7-9 times in C3C.

Fhadhq said:

- as I like the Save and Load options the Games offer, I may just go back to the beginning of the turn and try the same assault again. And again if the result doesn't change.

Click to expand...

The result only remains the same if you set it to remain the same when starting the game. If you uncheck the option to keep the random seed saved, than the results will always vary because the seed is going to be generated whenever the savegame is loaded.

Not retaining the random seed will change the results if you save and then restart the game, but that does not solve the question is the Combat RNG skewed at times towards the AI, for which there is considerable anecdotal evidence for.

A recent example from my latest game suggests it has nothing to do with human versus computer players:

I was hosing down an island of a civ that had declared on me (me with MA Armies, the AI with Musketmen at best and no Horse Units) and I'd made an alliance with some other civs just to prevent the island civ from gaining allies and making the whole process 10 times more complicated.

Anyway, I was approaching the last city and the AI ally had landed two Cavalry next to the last city. As is the case with AIs it had a HUGE flotilla of boats which bombarded the city to a group of four red-lined defenders (what a shame it can't realise it just needed at least 4 Cavalry and maybe a few less boats, but hey-ho).

The AIs first Cavalry attack resulted in the following:

Veteran Cavalry, full hit points, died (didn't even retreat) attacking an already red-lined Veteran Musketman defender. Meanwhile I'd been taking out Musketmen with my Cavalry (left-over troops) just for the chance of a great leader, expecting a few to die, but surprisingly without loss.

However...

From all my attacks I did get one really odd result:

Veteran Modern Armour red-lined and forced to retreat attacking a non-fortified Pikeman on Plains.

So I don't think it's really a human versus computer issue, I think it's something to do with circumstances or secret stats that's coded into the entire game, AI and human alike.

Buttercup said:

A recent example from my latest game suggests it has nothing to do with human versus computer players:

I was hosing down an island of a civ that had declared on me (me with MA Armies, the AI with Musketmen at best and no Horse Units) and I'd made an alliance with some other civs just to prevent the island civ from gaining allies and making the whole process 10 times more complicated.

Anyway, I was approaching the last city and the AI ally had landed two Cavalry next to the last city. As is the case with AIs it had a HUGE flotilla of boats which bombarded the city to a group of four red-lined defenders (what a shame it can't realise it just needed at least 4 Cavalry and maybe a few less boats, but hey-ho).

The AIs first Cavalry attack resulted in the following:

Veteran Cavalry, full hit points, died (didn't even retreat) attacking an already red-lined Veteran Musketman defender. Meanwhile I'd been taking out Musketmen with my Cavalry (left-over troops) just for the chance of a great leader, expecting a few to die, but surprisingly without loss.

However...

From all my attacks I did get one really odd result:

Veteran Modern Armour red-lined and forced to retreat attacking a non-fortified Pikeman on Plains.

So I don't think it's really a human versus computer issue, I think it's something to do with circumstances or secret stats that's coded into the entire game, AI and human alike.

Click to expand...

As I do not make computer Allies, I have not seen that. However, the AI that lost the Cavalry was ALLIED to you, in effect, making it a Human verses AI combat.

I do suspect that the Combat RNG has some odd coding in it which produces the odd results.