Elite Troops die easier

[sumthinelse]

Originally posted by Qitai
It is possible for a fast elite to die while a fast veteran survive. This is due to how the retreating works. FYI, a unit retreats if it gets a "unlucky" roll.

Qitai, you are right again.

I downloaded the posted game and that does seem to be happening. It takes a 5/5 elite longer to get down to 1 hit point, so when the random number is applied to determine if it retreats, it's a different random number than the one that determines whether the veteran or regular retreats.

Ozy is right that in this particular instance, of the 3 stacked knights, if you attack first with the elite, the elite always dies and the vet and reg units survive.

However, if you attack first with the unstacked knight 2 squares away, then attack with the stacked elite, and then attack with the stacked regular knight, the elite retreats and the regular dies!

I believe that if you made 100 such trials with different random numbers, where the defender had equal strength and defense bonuses the elite would survive more often because it has a higher probability of retreat (unless Ozy edited the unit characteristics).

 

[etj4Eagle]

Originally posted by sumthinelse


Qitai, you are right again.

I downloaded the posted game and that does seem to be happening. It takes a 5/5 elite longer to get down to 1 hit point, so when the random number is applied to determine if it retreats, it's a different random number than the one that determines whether the veteran or regular retreats.

Yep keeping track of the use of the random number is important for verifying these reports (had forgotten about the retreat calculation). I also think that moving units through some terrain might use a random number as well (I know when I was doing my leader generation test I accidently moved a unit the wrong way in verifying a test and then started to get different results).

But again even with retreat you can verify that there is no special penalty for an elite unit by looking at the round by round results prior to when the unit has the chance to retreat. Those rounds will be the same.

 

[Ozymandius]

What is significant though is that strategy must take this unllucky roll into consideration when attacking.

If you attack first with the vet and save the elite, you also win in the mountains plus you can now attack and take, the Greek city of Artemesium to the northwest with the two sleeping knight.
This battle would be lost if attacked prior to the elite battle.

My point is that attacking with an elite unit first can result in several defeats where a weaker unit attack can lead to several victories. This means that for us who do not play GOTM competitions, the save/reload feature is important to overcome the "elite getting killed syndrome".
Maybe elite troops just refuse to retreat whereas ordinary grunts recognize the futlity of an attack and save their skins so that they can fight another day

 

[Qitai]

It is also a good strat to always use a weaker unit to fight first to wear out the defender before sending in the elite. This way, your chance of elite surviving and winning is higher and thus having a greater chance to generate GL.

From my observation, GL generation is a purely dice roll thing with no relation to whateverso. So, keeping them alive and winning becomes the key to generating GL. The vet can die since you can produce tons of them from barrack. I have been able to consistantly generate 4-5 GLs per game recently, after piecing all the info on this forum together and playing in such a way that my chance of generating GL is maximized.

 

[Qitai]

Also, I don't think moving units uses RNG. Building a new city does use a RNG though, for reasons that puzzle me.

 

[TheNiceOne]

Hi Ozymandius, thanks for the effort, but as sumthinelse and etj4eagle points out, your save doesn't prove anything.

What the save shows, is that the 4 first combat rounds are identical regardless of you using the elite or the veteran, those are W,L,L,L. After the 3rd loss, the next random number is used differently in the two situations: The veteran uses it to calculate whether it withdraws, while the elite uses it for the next combat result.

Now, the chance of a veteran unit withdrawing is 67%, while the chance of the other knight winning against the pikeman is around 50% (maybe less), so the fact that one is success and the other a loss with the same random number is quite likely.

Then, in case of the elite unit, the 6th random number is used to see if it withdraws, which fails. If you had used the veteran first, then this 6th random number would be the first combat with a second knight. This combat must be a loss since the same number was a failed withdrawal - and this holds regardless of you using the elite or the second veteran.

The conclusion is that there is no hocus pocus going on here - merely the totally random numbers hitting you.

My point is that attacking with an elite unit first can result in several defeats where a weaker unit attack can lead to several victories. This means that for us who do not play GOTM competitions, the save/reload feature is important to overcome the "elite getting killed syndrome".

And I still maintain that this is totally random. This time the veteran survived, but if your 5th random number had been a bit worse (thus the veteran failing his withdrawal) and the 8th a bit better (where the elite now loses his last HP), then the elite had survived and the veteran died.

So I still don't think there is any elite getting killed syndrome. I'm sorry for not being more specific when I first requested an example, but for the example to prove anything, then the same random numbers must be used all the way. Here, the vet and elite fought identically until the withdrawal chance came, and after that they use different random numbers.

So please be patient - but to get a proof, you need to show a save game where withdrawal don't clutter the picture, i.e., a game where the attacker don't have more movement than the defender.

 

[sumthinelse]

Originally posted by TheNiceOne


Now, the chance of a veteran unit withdrawing is 67%, while the chance of the other knight winning against the pikeman is around 50%

According to civ3edit the elite retreats 66% and the vet 58%. But that does not refute your general idea.

 

[jimmymango]

NiceOne (and anyone else who seems to think that saving and reloading proves anything)

You can't demonstrate a statistical phenomenon by a single event. Saving and reloading a specific situation means absolutely nothing. (It's a sample size of 1).

What I've noticed from playing many,many,many hours is that Elite units die easier.

They don't *always* die, they just lose more battles than veteran units it seems. I don't particularly have the time or energy to compile statistics on my battles, but it was noticed by other people on this forum, and I don't think people just make things up to post on forums.

 

[etj4Eagle]

Jimmymango you misunderstand what the save will prove. Yes multiple events are needed to prove something that occurs at a certain percent of the time. However, in disproving a rule only one event is needed. Its been a while since I took Algebra, so I forget the name of this principle.

The current working postulate is that the combat sequence is not effected by the experience level of the units in question (except wrt retreat percentages). This thread was begun by claiming this was false, and so a save game was asked to prove it. As this is a rule differance and not a percent differance only one instace is needed to prove that the rule fails.

Hence as a conter-example has failed to be provided we can not accept the conter-argument and remain with the postulate that combat is not effect by the experience level of the units in question (except wrt retreat percentages).

If however the claim was that veteran units only retreat at a 10%, than many events would be requried to prve or disprove that point.

 

[etj4Eagle]

Just a few further items on this topic. As there are mulitple percentages coming into play it is conceivable that while in general elites do not die faster than veterans (the postulate of this thread) that in specific situations they might.

I worked out the probabilities for situations where the defender as 1,2 and 3 HP left when the veteran unit checks for retreat (now these are under the assumption that retreat is only checked once, if it is checked more than once, the formulas will differ).

Below are the fomulas and then the conclusions:
let W = prob to win the round w/ W=1-L=A/(A+D*(B+1))
Rv = chance for a veteran to retreat and Rv=1-Fv
Re = chance for an elite to retreat and Re=1-Fe

in such case the chance of a veteran dieing are:
1 HP: L
2 HP: L*Fv*(2-L)
3 HP: L*Fv*(L^2-3L+3)

for an elite unit they would be:
1 HP:L^2
2 HP:L^2*Fe*(3-L)
3 HP:L^2*Fe*(3L^2-8L+6)

Now using Rv=58% and Re =67% (I appologize if these are incorrect, but I am currently away from the game for a few weeks)

the 1HP and 3HP equations do not have a solution within 0<L<1 such that an elite unit is more likely to die. For the 2 HP case, L>0.71 So if you have less than a 29% to win a round and the attacker has to 2HP when the retreat option comes up for a veteran unit, the elite is more likely to die.

Now some with a bit more stat background probably could more easily derive the equations starting at the base case of fully healed units. And from there get what L would have to be. But as you can see it is far from a general case that elites will die easier, but it can occur in specific circumstances.

 

[TheDS]

Originally posted by sharky
if you keep on talking about this all save and reload , you can cause millions of deaths.
if hitler would find out about it, he would just reload 1942, wont attack russia, and control the world.
so be quiet, you found this amazing realistic system of saving and reloading, but there is no need to talk about it anymore.
be quiet.

In Eye of the Beholder, we called this "Cast the Save Game Spell" and "Cast the Restore Game Spell". It is since a part of my gaming language. Does that date me?

Ozy:
We're ALL whining here! Except maybe the people making fun of us...

Anyways, some one mentioned attacking with weaker units to soften up the target. This is not always a good idea! What happens if your target beats you, and then gains experience? You've just made him tougher to beat. The problem is even more critical when you make Vets worth 5 and Elites worth 8; you can turn a marginal defender into a real tough guy by throwing weak units at him. (And it's really neat to have an elite Warrior ambush an unsuspecting enemy regular swordsman or fend off a couple regular Archers.)

 

[TheNiceOne]

jimmymango, etj4Eagle has answered already, but I'd like to answer as well: We're not discussing a statistical phenomen, but an absolute phenomen:

Ozymandius says that something can (and has) happened, while I say it cannot happen. If Ozymandius can show 1 occurence of this, then he has shown that it can happen and that I am wrong.

If we were talking about how often something happened, then we would need a lot of samples, but proving that something is not zero can be done with one example.

Ozymandius one example so far was not valid (explained by me above) though, so we'll have to wait for a definite answer.

Unfortunately, I cannot prove that it never happen, so I'll have to argue that it is indirectly proved by the lack of proof by Ozymandius

 

[TheNiceOne]

etj4Eagle, there is an error in your interesting and otherwise excellent mathematical examination.

For those of you who don't have much patience for mathematical formulas I'll give the conclusion first: There are no combination of HP and W/L probabilities where an elite has higher chance of dying than a veteran.

Now on to the meaty stuff: The 2 HP formula for the elite unit has two errors, one mathematical and one logical.

The mathematical error is very easy to spot if you substitue L for a real number (say 0.8), in which your formula gives the result that an elite dies 140% of the time! The error is that (3-L) should be (3-2L). Now, the logical error is that if the elite starts with a win, it will not roll for withdrawal (since the defender will only have one HP), so the final formula becomes a bit more complicated:

2 HP:L^2*(Fe*(2-L)+1-L)

Anyway, neither your corrected formula (with 3-2L) nor my formula gives any solution to the equation for when an elite unit has a higher death probability. So there's no single battle where it would be statistically better to use a veteran than an elite unit.

 

[etj4Eagle]

Originally posted by TheNiceOne
etj4Eagle, there is an error in your interesting and otherwise excellent mathematical examination.

For those of you who don't have much patience for mathematical formulas I'll give the conclusion first: There are no combination of HP and W/L probabilities where an elite has higher chance of dying than a veteran.

Now on to the meaty stuff: The 2 HP formula for the elite unit has two errors, one mathematical and one logical.

The mathematical error is very easy to spot if you substitue L for a real number (say 0.8), in which your formula gives the result that an elite dies 140% of the time! The error is that (3-L) should be (3-2L). Now, the logical error is that if the elite starts with a win, it will not roll for withdrawal (since the defender will only have one HP), so the final formula becomes a bit more complicated:

2 HP:L^2*(Fe*(2-L)+1-L)

Anyway, neither your corrected formula (with 3-2L) nor my formula gives any solution to the equation for when an elite unit has a higher death probability. So there's no single battle where it would be statistically better to use a veteran than an elite unit.

Thanks for the correction, I had forgot about that 1 HP rule. (I had been a little concerned that it showed up, as my feeling was that a veteran should never be better). So with your corrections, we now know that the math also does not support that elites die easier.

Hence until an example is provided showing that the game rules behave differently than they should on average an elite unit will die less often than a veteran unit.

 

[EdwardTking]

"Elite Troops Die Easier"


Interesting. It is my perception too that my Veteran's do better than my Elite troops. I cannot prove it; and may have just been unlucky. However the fact that I and others have not proved it, does not mean that it is not so.

 

[etj4Eagle]

Originally posted by EdwardTking
"Elite Troops Die Easier"


Interesting. It is my perception too that my Veteran's do better than my Elite troops. I cannot prove it; and may have just been unlucky. However the fact that I and others have not proved it, does not mean that it is not so.

But there is a difference. We have rules that state how combat occurs and from this we can derive the various probabilities. Hence assuming these rule are followed, this is just an illusion of how we remember things. Since no one has found the rules being violated, we know that it must just be an illusion.

 

[Ozymandius]

I object strongly to that last comment.
I posted my game to show that in certain circumstances the mathematical formula gets my elite killed. No excuses. The vet survives, the elite is transported to memory heaven.

If the hit points calculator lets him die, the formula is wrong. No matter how often you try, that attack results in the death of the elite if he initiates the attack. The formula contradicts perceptions & common sense in this instance. In reality rookie death rates surpass veteran troops. Just because it does not happen all the time does not mean that our perception is in error.

 

[etj4Eagle]

Originally posted by Ozymandius
I object strongly to that last comment.
I posted my game to show that in certain circumstances the mathematical formula gets my elite killed. No excuses. The vet survives, the elite is transported to memory heaven.

If the hit points calculator lets him die, the formula is wrong. No matter how often you try, that attack results in the death of the elite if he initiates the attack. The formula contradicts perceptions & common sense in this instance. In reality rookie death rates surpass veteran troops. Just because it does not happen all the time does not mean that our perception is in error.

Yes you can point to an occurance here and there where an elite unit will die and a veteran unit will live when retreat is allowed. However, what we are saying is that if you take a sample of 100 veteran units and a sample of 100 elite units, more of those veteran units will die than elite units.

The rules and statistic support this trend, the fact that in specific cases the opposite outcome occurs does not invalidate the statement that Elite units are more likely to survive than Veteran units.

The probability of a Veteran living when an elite would be killed happens less often than the occurance of an elite winning and a veteran dieing.

 

[Zachriel]

Originally posted by Ozymandius
I posted my game to show that in certain circumstances the mathematical formula gets my elite killed.

Sometimes elites will lose when veterans would have won -- even with the same randomizer seed. Your very interesting post has clearly demonstrated that.

This does not mean that "elite troops die easier." The real odds are not changed by our sudden realization that sometimes a saved randomizer seed will still result in an elite dying before the veteran. Due to the random nature of combat resolutions, sometimes elites will lose, sometimes veterans, but elites will tend to be more survivable in the long run.

Reloading to change an unfavorable result is still reloading.

 

[Chieftess]

I've had those wierd things happen too.

Mod. Armor losing to a spearmen...

And..

2 knight armies losing to a mustketman, however, 1 lone knight defeats the 2 defenders! (one only had a 1 HP loss).

I even noticed that elites die quicker. So, now, I only use elites against units with 1 or 2 HP.

BTW, I've had vet and elite infantry (and tanks) LOSE to 1 HP conscripts!