I Have Mathematical Proof That The Random Number Generator Is RIGGED

[Wiglaf]

Let me break this down for you, fresh from algebra II.

A 95% chance to win 5 times means a (95*5) chance to win = 475% chance to win. Therefore my stack of pikemen being destroyed by a single horse archer makes no sense and defies logic. My study is better than previous ones, which did not have a stack of the same unit with the same experience doing the same trial.

I believe whoever designed the RNG is also behind the apostolic palace. Stay tuned

[dschur]

I believe your math is flawed. If your chance to win is 95%, your chance to lose is 5%. So your chance to lose 5 times in a row is (simplifying here) .5%^5 = .00003%

Yes this is a very small number, but it's not actually zero.

[Vampgelus]

The mistake you made was to take a percentage (chances out of 100), to multiply it 5 times, and consider the result as a percentage too. The result, are chances out of 500. 475 winning fights out of 500.

In addition, you don't take the damage the enemy HA takes each fight into consideration, giving you a result that is only correct for 5 pikemen attacking each one a different healthy HA.

[jbp26]

you guys aren't getting his point.

[rah]

I only have one problem with the rng.

When attacking a city the first time. THe first unit I attack with seems to lose at good odds more than later attacks. Example. 4 units attacking at 75%, it just seems that the first one is the one that loses the most.
I've just stopped attacking with my GGs first and they seem to last longer now.

OF course in my last MP game I lost a GG on a 98.8 attack.

[TriMiro]

Look guys, there is a way to test the random number generator, just not the way above. Also I do not believe the numbers that you get in the beginning of the attack are correct in terms of showing what is the actual chance to win vs. actual chance to loose.

The numbers that you get with the right click are nothing more than the strengths of the two units plus all the applicable modifiers (+25% vs mele, +100% vs horse and so on). Those odds really only apply to the first round of combat. Whichever unit looses the first round of combat will have reduced strength for the second round and the odds for the second round are modified.

Dumb Example: Pikemen vs Horse Archer, no promotions. Strengths 6 and 6, but P has +100% modifier, thus the odds are 12 vs 6 for P, or in other words there is a 66% chance that P will win the first round vs 33% that HA will win. If we get the 1/3 odds and HA does win, then the strength of P would be reduced to say 5.4 (I am not sure about the exact number). Then 5.4 would get modified by +100% and the odds for the second round would be 10.8 vs 6 for P. In other words there is a 64% chance that P would win vs 36% chance that HA would win. Repeat until one unit dies.

To test the RNG for the meaning of the first R, you need to look at the rounds of combat. Then you need to consider the odds of winning vs. loosing and over many battles consider the deviation of actual wins/looses vs. predicted wins/looses. The rest is math/stats theory. I could tell you how to test it and/or what equations you need to use, but I will neither gather data nor crunch numbers myself. IMHO it is a useless task.

The only truly random thing in a RNG, and make that any RNG, is the seed. Everything else that follows is actually deterministic, i.e. it could always be perfectly predicted. The seed is determined on game generation and is preserved for the duration of the game to protect from save-fight-loose-reload-and-repeat until victory abuse.

Save the game before battle then fight see the outcome then reload and fight again. The outcome would be the same. That option was first added in civ III conquests, I believe.

So the RNG that is being used is designed to mimic truly random behavior and as such it cannot work perfectly. The question is not whether it is wrong, but how often and how wrong is it. I have no general complaints, if the odds are in my favor in the beginning of the fight I will usually win.

If you attack a city, it is only normal to lose the first couple of battles. Suppose you have a stack of identical attackers, for the first one, the city would pick its best defender, say an archer with two CG promotions. Odds are that you will lose that battle, however, the archer would be wounded and would no longer be the best defender available. The second defender could be an archer without CG and thus you will have much better odds of winning.

I always attack a city with weak and non-promoted units first and use the veterans only after the odds have turned in to favor me. That way I would have several very strong units available for defense and control of the newly conquered territory.

There is a random effect involved in the AP, so there is definite link between AP and RNG.

[rah]

I realize that the best defender defends first. That wasn't my point. My poing is that if I have 5 80% chance attacks, my 1 in 5 lose will usually happen on the first attack.



Please note that first strike is not calcualted in the odds.
Quite noticable when you have 4th level drill units that consisitently win even attacking at suposed 33%.
I had one stack of 8 of them, with none showing more than 30-32% chance, and all 8 won. So you must keep first strike in mind when you read the %

[Vampgelus]

What is the forumla for damage done in one round ?

I was often wondering about first strikes efficency, knowing this would help me alot.

[Dale]

If anyone knew anything about computer RNG's, you would realise his comment IS 100% correct (but not the way he thinks).

Computer RNG's are not random, but based on a complex calculation. Computer chips are not able to "pick a random number".

Their RNG is based on a seed which is fed into a formula which creates a random number table. The computer then grabs these numbers in order.

So the effect is if you ran the RNG once you may get:
3,7,5,4,8,0,4,1,8
If you ran it through multiple times, each time you would get:
3,7,5,4,8,0,4,1,8

So how do you get around that? You randomise the seed. This is based on a simple formula based off the internal chip clock. This random seed then creates a new random number table for the computer to use for its RNG.

[rah]

Good explanation.

Using SAS, you can specify a random seed (based off the clock) or set the seed (in case you're testing and want to replicate results or doing something where replication is important)

It performs an arithmatic formula on the seed number to generate the next number. Performs the same formula on the new number to get the next number. etc. etc

The string is always dependent on the starting seed number.

[snoopy369]

Originally posted by TriMiro
Look guys, there is a way to test the random number generator, just not the way above. Also I do not believe the numbers that you get in the beginning of the attack are correct in terms of showing what is the actual chance to win vs. actual chance to loose.

The numbers that you get with the right click are nothing more than the strengths of the two units plus all the applicable modifiers (+25% vs mele, +100% vs horse and so on). Those odds really only apply to the first round of combat. Whichever unit looses the first round of combat will have reduced strength for the second round and the odds for the second round are modified.

Nope. That's not true at all. The percentage chance is of actually winning/losing the entire battle, not just the first skirmish. Look at the combat log for details there. The % chance to win takes into account hitpoints, firepower such as it is in civ4, and such... IIRC the only thing it doesn't deal well with is first strikes. It's a complex calculation (look at the code) not a simple percentage

[TriMiro]

Civ IV preserves the seed, i.e. if you save-fight-loose-reload-fight, you are guaranteed to loose again. Civ III has "preserve random seed" option, Civ IV has it in-build and I do not think it could be removed unless moded.

The formula for damage per hit is unknown to me. If someone knows the formula please share it.

I have a number of questions on how first strikes work, but I am not sure if this is the right tread. Since the first several turns of combat determine the winner, first strikes could provide huge advantage. I have seen drill III caravels kill pirates in 2 of 2 and drill II frigates kill SoL three of three. That, however, is poor statistics, i.e. the sample is too small.

[snoopy369]

Civ4 has it as an option as well, actually.

Read the combat log, it will show you most of the details, iirc including the damage per hit.

FS are resolved as such:

(# of FS of attacker) - (# of FS of defender) = FS#

If FS# is 0: Fight as normal
If FS# is positive: Attacker gets FS# of attacks first, resolved as normal except if defender wins, nothing happens (instead of hurting attacker).
If FS# is negative: Same but for defender.

So, axeman attacking forted CD2 archer (no FS bonus beyond archer bonus):

Axe#: 0
Archer#: 1
(0-1)=-1

Defender gets one attack.

Axe (5) vs Archer (3*(1+.5+.45)) = 6 for simplicity (say 5% fort bonus).

5 vs 6.

first strike: Axe rolls a 19, archer a 14, so axe 'wins' after the modifier (I forget how that works exactly, humor me).

However, because it's a FS, nothing happens, and we move onto the regular battle.

I believe the combat power is something like (attacker power/defender power)*5 or something to that extent. Again, look at the combat log.

[TriMiro]

OK snoopy posted the same time I did, hence the double post.

I was looking at the wrong thing, the combat odds do indeed show the odds of winning vs. loosing the entire battle. I was looking at the odds below that show the strength + modifiers.

Testing the RNG in that case would be simple, take a sample of 50 battles with different odds and match predicted vs. actual outcomes. I could work out the formula to then test if the RNG is flawed, however, I do not think it would be.

To snoopy again, do you know how damage is computed? Do you know how first strikes are computed? In general, where can I find that information.

[TriMiro]

Again at the same time.

Thanks for the answer, it is really helpful. Also the way I understand it, the FS count only for the fighting units. In Civ III a canon would try to hit a cavalry that was attacking the riflemen if the cannon and the riflemen were on the same spot. In other words, if I have two units defending, only the FS of the fighting unit count.