Here I'm attacking the barbarian city Hun (defended by three longbows) with four knights, three war elephants and a warrior wich recently popped out from a goody hut. I'm told the odds are 65.9 %. Not the best, but I want Hun and attacks with the stack. Then my warrior makes a kamikaze attack! Luckily I'm able to reload a save and check the odds for the poor fellow: < 0.1 %.

Next try, with only knights and elephants: A war elephant attacks! And is killed. It's odds are 16.0 %.

Third try, with only knights: Two down, the third survives, the fourth dies. So finally I capture the city with the loss of three noble knights.

Fourth try, just to check, I start all over with the stack: The warrior is killed, then the first elephant, then a knight survives, another knight is killed, before the two last knights capture Hun. Loss: Warrior, elephant, knight.

Only explanation:

The game knows the odds and that I'm gonna loose three units. It cheats in my favour, sacrificing two weak units before my first win.

But then: Why attack with a knight the fourth time, instead of one of the remaining elephants?

(And yes: I made a fifth try, sacrificing an elephant instead of a knight. It worked. Loss: Warrior, elephant, elephant.)

Strange.

I think I'm gonna make a last attempt, getting some catapults first, just to weaken the longbows. Cheating? Yes, but the game cheated first.

The RNG is much criticized. It seems to me -- seems -- that the wins and losses are going to happen no matter what unit you use. In this case you would lose the first three battles, no matter which unit you send in. Only if you have a unit greatly superior to theirs. If you had sent riflemen, the first three would have lost to the longbows. Infantry, maybe. You could change the odds by saccrificing a cat and doing collateral damage to the other units.

Reloading the game won't give you different results unless you create a custom game with the box for new random seed checked. In your case, the random numbers used with your first 3 attacks were likely strongly against you winning even with over 50% odds.

Reloading the game won't give you different results unless you create a custom game with the box for new random seed checked. In your case, the random numbers used with your first 3 attacks were likely strongly against you winning even with over 50% odds.

The reloading was just for testing out the fundamental aspects of the game. I know the random seed's the same in each try.

The point is:

The game knows I'm gonna loose the first, second and fourth battle. In the first two it sacrifices minor units. I, the human player, would have tried a stronger unit than the warrior and the elephant. The game knew I would loose and sent in the weakest.

That's cheating in my favour. I don't like it. The game mechanics should not cheat in anyone's favour. Me or the AI civs. Especially me. That's not my kind of fun.

Agree 10CHARS

And I suddenly realized that when I'm attacking a city with a stack including siege weapons and the game gives me the retreat odds, not the attacking odds of the strongest unit, thats because the siege weapon will survive the attack.

I really got pissed at this fact in the older games and before the expansion. The "coincidental" odds were absolute crap. I think BTS made it way better, i mean it is ok that a guy with 98% odds die once every 100th turn. But god it annoyed me in the older games. You could attack with 5 tanks on a city guarded by ten warriors and lose 2 tanks just because of the large amount of units.

I mean it is ok that a guy with 98% odds die once every 100th turn.

Twice every 100 turn, you mean? :lol:

Twice every 100 turn, you mean? :lol:


no i mean once, and if it happened twice i would pour excrements all over my computer and fill the screen with severed heads, light it on fire and piss on it. :rolleyes:

Odds are odds and chances are chances, and the RNG does (correctly) sometimes produce a series of results well against the odds. For example, I have lost a fight due to the AI unit making a string of successive hits, the chance of this happening being less than one in a million (the AI unit's chance of winning a combat round being about 20%).
Although I don't have access to the RNG used in Civ4, I can use the one in QuickBasic and ran a test of a million single-precision numbers between 0 and 1, setting 0.01 as a trigger level and counting strings of successive results of that figure and below. The longest such string was of 16 numbers, which would equate to suffering 16 successive hits at 1% chance of losing, and there were quite a lot of strings of 10 or more. Running the same program again with a different seed, this time looking for strings at 0.99 and above, gave a longest string of 14, occurring twice.

EDIT : Ignore the coloured text, which is rubbish. I have rewritten my program, hopefully correctly this time, and run a test of 10 million random numbers, testing both high and low streaks with the same seed. I got exactly the same number of high and low results in total, triggering at 0.95 and 0.05: longest streak of "lows" was 8 long, and of "highs" 7. Plenty of shorter ones, though.

This is not an "oh unlucky me, I lost a battle against good odds"-thread. (Nice one here. (http://forums.civfanatics.com/showthread.php?t=256562))

This is about the game mechanics. It's obvious that the game uses the knowledge of the random seed to twist a battle in the attackers favour. If you let the game think for you, you will come out better than if you uses common sense and make the choices yourself.

For most of us common sense will be softening the defenders with siege weapons before attacking with low or medium odds. When the computer knows an maceman with 30 % odds will survive, it attacks with the maceman. If it knows the next unit will survive with 56 % odds, it attacks. If the third one will be killed with 98 % odds, the computer sends inn a trebuchet.

And because the AI never uses common sense, the AI gets an advantage.

That's how it looks to me, after recent experience.

This is not an "oh unlucky me, I lost a battle against good odds"-thread. (Nice one here. (http://forums.civfanatics.com/showthread.php?t=256562))
For most of us common sense will be softening the defenders with siege weapons before attacking with low or medium odds. When the computer knows an maceman with 30 % odds will survive, it attacks with the maceman. If it knows the next unit will survive with 56 % odds, it attacks. If the third one will be killed with 98 % odds, the computer sends inn a trebuchet.

I don't think the the AI use this to there advantage. I don't even think they ar aware if the ods. They just attack or not, how many times have not the AI charged old junk units to a city without a single win. If they know they whold loose, why whold they attack in the first place?

I don't think the the AI use this to there advantage. I don't even think they ar aware if the ods. They just attack or not, how many times have not the AI charged old junk units to a city without a single win. If they know they whold loose, why whold they attack in the first place?

1. Yes, the AI attacks. Sometimes madly. If it's at your doorstep, it seems to never retreat. Stupid attacks, very often. Sacrifices good units for nothing.

2. When it attacks, it seems to use the random seed in its favour. Stupid attacks, if you use common sense. Not so stupid if you look at the results.

3. When I attacked with my stack (as described in the first posting in this thread) the computer made some choices (on my behalf) that would be very stupid i you used common sense, but very wise if you knew the random seed.

I can't agree about not stupid, they have attacked me without a singel win. lets say 6 units attack and all loose, how can this be an advantage? This is not civ 3 where the ai know things they shold not know. ex. where all strategic resources are before knowing the techs that reveals them. they are more like us this time, but more stupid of cous.

Every unit combat combination is clearly calculated before the battle as soon as units move in range of each other - obviously this is so reloads dont affet the outcome of the battle. The AI is aware of the odds before it attacks. Through many reloads I have altered the outcome of what the AI does with it's units by adding extra into a citiy for example.
I'm happy with this, what gets me is that the battles outcome heavily determined by difficulty level and whether the AI is attacking you or you are attacking - irrespective of the percentages. What the AI can get away with when attacking your city, walls, castle etc is just mind blowing. I've had to rething my entire defensive strategy in order to cope with it.

Every unit combat combination is clearly calculated before the battle as soon as units move in range of each other - obviously this is so reloads dont affet the outcome of the battle. The AI is aware of the odds before it attacks. Through many reloads I have altered the outcome of what the AI does with it's units by adding extra into a citiy for example.
I'm happy with this, what gets me is that the battles outcome heavily determined by difficulty level and whether the AI is attacking you or you are attacking - irrespective of the percentages. What the AI can get away with when attacking your city, walls, castle etc is just mind blowing. I've had to rething my entire defensive strategy in order to cope with it.

I'm happy with everything that makes the AI's behaviour less stupid, even some degrees of "cheating". I prefere a strong opponent. But it's a problem when the game mechanics "helps" me the same way. It changes my gameplan. My offensive gameplan. I don't want the "help" of a computer with knowledge of the random seed.

To send a warrior into battle against a heavily defended city is stupid, even if you know (by knowing the random seed) that it's a "wise" move.

I always thought that the AI choice of attack order was random. Defense, of course, is matched to the attacker. I just took an Egyptian city and garrisoned it with 6 MI and 12 paratroopers, thinking surely that that would be enough. They were the surviving attacking troops, all veterans.
Egypt was a vassal of Sumeria, and Sumerian brought a stack against the citiy next turn. 27 Cavalry, 15 Riflemen and 17 Grenadiers. I reviewed the combat log after the battle, and as expected, the odds at first were 0.0. They sent 6 cavalry, half of whom retreated. Then 5 Grenadiers, then 10 Cavalry, 1 Grenadier, 6 Cav, then alternating rifleman and Grenadiers. Their troops were vets with various promotions. It looked like they sent the strongest in first. The MIs typically took one hit per attack for 10%. The computer started defending with paratroopers when the MIs got down to about 70%.
The city stood, BTW. I only hope that that is most of his strength.
I have two columns of heavy tanks coming in from the west. When i attack with a stack I choose my units. I don't like the computers choices.

The selection of Attack order goes this way

1. If any of the attacking stack gets "Good enough" odds, attack with the one with best Odds

2. If NONE of the Attacking Stack has "Good enough" odds, then you attack with the unit that will do the "best" ie most efficient damage. (the warrior didn't do the most damage, but the Warrior was cheap)


# 2 is there for suicide Catapults, but as you can see it also works for suicide Warriors


The AI did NOT know you would lose 3 units, it just knew those first ODDS (not the RNG) were really bad, so weak units were sent in to do damage, befor the "real" battle could begin. (the Third unit had a better RNG than the Fourth unit... but the AI didn't know that)

I'm not convinced by your example. When a game begins, if the RNG is not set to re-seed on reload, the odds for all the future battles are pre-determined from the point of view of the RNG. The only variable at this point is what units you build, what promotions you give them, and who you fight against (the AI battles will use the RNG too). This isn't cheating at all, although an observation by me concluded that if the RNG was truly random then it might be less fun than one that occasionally tilted the odds in your favor. I've lost enough battles when I was 98% to win that I think I've defied those odds so far.