Great Leader from elite unit within Army?

"No, why shouldn't it be able to be taken? You just will need some attackers... The rest lays in the hands of the... ahhh.. RNG."

Please explain this Randomn Number Generator to me. There seems to be more to it than simply rolling the dice for each side each round and adding the relevant attack/defense strength. Does it link rolls that should be independent? i.e. if that spearmen luckily wins the first round against your modern armor, he is strangly likely to win the second and third rounds?

"Remember, the army *might* heal during the inter-turn phase.."

Explain this *might* too please, I don't have Conquests yet.
 
I'm not a programmer or anything but as far as I understand it a computer can not create a true random number like rolling a dice can because of this the computer is stuck with simulating the creation of a random number and some RNGs do this better then others. I could be horribly wrong but I heard someone say this once.
 
My programmer friend says you're right. Fair enough, but a game as extensive and well designed as Civ3 it seems to me should be able to do a better job of it. Losing two or three undamaged tanks that are attacking ancient units in one war seems like a pretty big distortion, and I've seen this happen not infrequently.
 
Originally posted by a4phantom
Losing two or three undamaged tanks that are attacking ancient units in one war seems like a pretty big distortion, and I've seen this happen not infrequently.

Weird... as it seems to happen 'not infrequently' to you, care to share respective save game files next time?
:groucho:

It's because the frequency of such files being uploaded showing such streaks is zero AFAIK.:(

That aside, I could not be bothered to lose even 5 tanks or so vs a rival with ancient units. What's such a tactical loss worth here? A good laugh.
Rival will be toast in no time anyways. And that's what counts.:)
 
The AI civs don't seem to upgrade or decomission a lot. Don't you ever eliminate the enemy's riflemen and cavalry or even tanks and infantry only to find that there are still a horde of old, often elite spearmen, pikemen, etc waiting for you to retire them? And no it's no huge loss for me to lose even a bunch of tanks to wierd RGN, mostly because so far I've only played on easy levels (I will be trying out Regent this weekend) but it does delay the taking of cities or leave damaged offensive units open to counterattack. When an elite unit is lost to bizarre numbers (say a slightly wounded elite modern armor losing four consequtive rounds to a musketman who is not receiving many bonuses) it decreases chances of getting a GL to rush wonders (I'm still playing PTW).

If I get another game where several modern units are quickly lost to ancient units under bizarre circumstances I will figure out how to upload the save for you.
 
Originally posted by a4phantom
"No, why shouldn't it be able to be taken? You just will need some attackers... The rest lays in the hands of the... ahhh.. RNG."

Please explain this Randomn Number Generator to me. There seems to be more to it than simply rolling the dice for each side each round and adding the relevant attack/defense strength. Does it link rolls that should be independent? i.e. if that spearmen luckily wins the first round against your modern armor, he is strangly likely to win the second and third rounds?

"Remember, the army *might* heal during the inter-turn phase.."

Explain this *might* too please, I don't have Conquests yet.

The RNG in principle is a small computer (sub-)program. As especially Windows-users know, computers are predictable in their behaviour :p
In general that means, if you press button A, you will cause reaction B. Always. That is, what computers are constructed for, and at the current moment, anything different from that would imply a severe hardware problem. Period.
But, this of course is not at all, what we want for games' purposes.
Now the RNG enters the stage. It just doesn't do anything else than calculating rows of numbers (something between 0 and 0.99999999999999999....), according to a certain inherent algorithm. To avoid to have the same numbers all the time, the RNG is fed by a socalled "seed" (you may have seen this on the first page, when creating a new game in C3C 1.15). That seed is the base for the calculation to start. Now, different seeds then mean different rows of numbers. Typically, a RNG is fed multiple times by various seeds during the course of a game. Often, the time between two user inputs (mouse movement, pressing a key) are used for that, since we humans are unperfect and for that, unpredictable. So, even if you try to click on a certain key in constant frequency, there will be very small differences. But the computer is quick enough to measure those and for that gets different seeds, then.
Now, taking all of that into account, it is obvious that the determining factor for the "quality" of a RNG is the algorithm, which is used for it. Different algorithms may mean different rows of numbers while using the same seed.
Nevertheless, as soon as you know the seed and the algorithm, and the number of calculations already made, you would be able to predict the next number. For that, a computer's RNG in fact in just a SRNG = semi-random number generator. It just seems to be random, what number comes up next since you don't know the seed, the algorithm and the number of calculations already made.

Next thing. As I understand it, in Civ (PTW, C3C) the RNG doesn't roll dice for both, the attacker and the defender, but it "rolls" just on "dice", that is, it creates a number in the above mentioned interval between 0 and 0.999... To allocate the outcome to one of the two sides, it does the following:
Assumed, you have an attacker with A=4 and a defender with D=2. Then the probability for a win(attacker) would be 4/(4+2)=4/6=0.6666666
That means, you may say that every result of the RNG below 0.666666 would count as a win for the attacker (or, you just reverse it, and would say everything above (1-0.66666666) would be the attacker's win)

Hmm.. that's in principle, what I recall from my university about RNGs.....

About the "might heal". As far as I have experienced it, YOUR units will need a complete turn without action to heal completely in a city with barracks. Sometimes I get the impression, that for the AI units the inter-turn is sufficient to be healed.
 
About the upgrading of units:
That of course depends very much on the amount of gold available, and the availability of barracks and ressources (and the location of the specific unit).

At my current game, I can watch the Arabs moving SODs back to their home territory, and some units come back again. Home they send spearmen and bowmen, to show up with riflemen.

So, the AI *DOES* upgrade. That is another good reason for selling techs to AI's, since by that you limit it's chance to upgrade, thus making it weaker compared to your state-of-the-art military... ;)
 
Originally posted by Grille


Weird... as it seems to happen 'not infrequently' to you, care to share respective save game files next time?
:groucho:

It's because the frequency of such files being uploaded showing such streaks is zero AFAIK.:(
[...]

The problem about this is that you rarely know in advance about a certain 'strange' outcome of a battle. Now, you will say, what about the auto-save?
Correct, but what about sending you an autosave and a list of 200 units to be moved in a certain order, a certain direction, a certain range etc, to allow you to see the results for unit#200+x?

Not to mention all the decisions you've made at the end of certain cities' production, some negotiations and what else...

The problem is as if you would ask somebody to repeat his results at the Roulette.
 
Originally posted by Commander Bello


The problem about this is that you rarely know in advance about a certain 'strange' outcome of a battle. Now, you will say, what about the auto-save?
Correct, but what about sending you an autosave and a list of 200 units to be moved in a certain order, a certain direction, a certain range etc, to allow you to see the results for unit#200+x?

Not to mention all the decisions you've made at the end of certain cities' production, some negotiations and what else...

The problem is as if you would ask somebody to repeat his results at the Roulette.

Oh, although it may be a bit tedious, I would not care about moving too many units. For one, if 'preserve random seed' is checked (option must be checked anyways, otherwise it's pointless), AFAIK the movement order/attack sequence should not matter. That is, within a 'simple' set of action happening during that turn. Plus, the production prompts etc would have been done when you load.

So if you move (or attack with...) like 187 units and the 188th unit is a tank which loses against a spear, you could reload from auto-save and you *should* be able to pick that same tank right after turn start as first unit to be moved.
AFAIK each unit will get the exact same result regardless of movement/attacking order. Some RNG-dependent occurancies may not work, though. (I'm guessing about e.g. GL-creation here; just imagine you'd get 2 GLs in one turn, but after reloading, you don't immediatelly use the first one which would block off the second one - so I think this could somehow 'mess up' the exact same pattern of numbers being drawn for each unit after reloading.)

Ok, at least I know from some testing that picking up the units in a different movement order with a simple background (no GL usage etc) will normally not mess up the battle outcome after reloading.


Anyways, it's just that you can read like 100+ posts (FYI, there's actually a similar thread in GD right now) about the 'quite frequently happening' tank vs spear loss, but yet no author showed a save file, whether there would have like 200+ units to be moved or not.
OTOH, if something really rare & interesting happens in a game, you can 'frequently' (;)) find a save-file (or a screenie) here on CFC for a proof. And this is why I get the feeling that the statement 'frequently, a tank loses to an ancient unit' is misleading. Apart from that, I cannot detect RNG-conspiracies in my own games. Granted, I have lost battles which looked liked no-brainers on the paper - but I have also had great luck in another instance.
Now there is of course a luck-factor present - question is, if you could (should) ever remove it. This does not only refer to battle outcomes, but as well to the strating location, AI neighbours and their behaviour etc.
I think that the luck factor is out-evened in the long run.

Here's a not-so-serious rule:
If you plan to attack a town defended by 1 warrior with 1 cavalry only, you *expect* a win, but you'll lose.:mad:
If you plan to attack a metro defended by 2 infantry with 50 cavalry, you expect severe casulties, but lose no single unit and take the metro. You get happy.:D

Conclusion: don't rely on one comfortable probability to win one single battle.:)
 
Aye, there is no doubt that if you're serious about taking a city, you bring as many units as you can to the battle...

But I'm still not convinced there hasn't been some change in the RNG. Anyone notice how barbarians now lose *constantly*? If anyone has a right to complain about bad luck, it's those poor guys! Not that I really feel bad for them, mind you...
 
@Grille:

I am not quite sure if you're right with your assumption that the order in which a certain unit is managed doesn't change the outcome of it's battles (of course under the assumption that the "preserve" function is on).
Since a RNG does nothing else than to create rows of numbers, your statement would mean that for each unit the system should keep track on an "internal" order number which would be regardless of it's "external" order (<= handling by the human player). I agree although to the expectation, that if it would fight the battle #24 in the original turn and battle #24 (with no other access to the RNG in between) after reloading, it should give the same result. For that, my explanation in my previous post wasn't very good, as I've to admit.
But, are you able to memorize all your battles in a turn *AFTER* you suddenly think you notice something strange? I'm pretty sure, you can't.
And that is the problem with those moments. You just know about them *AFTERWARDS*.
Again, it is somewhat similar to Roulette. You still remember, that your 28th bet was on black, and you lost. But, if you would have to change colors after a win, than you would have to recall all other 27 bets before, to get the same result for bet #28 (given, that the sequence for the colors would not change)

But this is not the main problem, I think. The main problem is, that we all live dependant on our experiences. And our experience constantly tells us, that certain things are allowed to be expected, since they have proven in the past to happen with a certain regularity. To notice this, just forms our experience.
Nevertheless, our experience tells us as well, that there may happen exceptions from time to time.
That is, as somebody explained in previous thread about the proposal of the new combat system, why a game like Civ doesn't have a combat system like Chess.
In real life there may be luck, ingenuity or whatever, to make a certain event result in an unexpected outcome.
In single player games this just doesn't happen (or, in case of the ingenuity it would solely benefit the human player).
So, to "simulate" those unpredictabilies, they make use of the RNG. But, as has been pointed out often enough, this doesn't work either, since now we have those "streaks", which are inherent to RNGs. To compensate for this, the hitpoints have been invented (and, of course, the A/D values).
Now, we have three factors which determine the result of a battle. RNG, hitpoints, and relation of A/D values.
It is my personal point of view, that the hitpoints in general are too low for the aspired effect of a good simulation of what experience from real life (or the transfer of our "real life" experience to combats as in Civ - as I am pretty sure no living being ever has experienced a death-match between a tank battalion and some experienced spearmen) tells us what we will have to expect in a Civ battle.
But even that is not the point, of course.

At the bottom-line, I hold my point that "streaks" seem to happen more often in C3C than they did in PTW. What may be the reason for it, I have no idea about.
It might be a modification of the RNG. That could have happened just by compiling the C3C source code with another compiler than the one for PTW, so it would have happened without intent and without even changing a single line regarding battle result calculation.
Or, they have changed a little bit. I mean, obviously they had been planning to change the battle system, anyway.
So, this suspicion is not that far away.

But, final clarity may perhaps only be obtained if both RNGs would be decompiled and scientific tests would be performed.
 
Top Bottom