Easy veteran->elite

[hifisoftware]

Yestarday I found a guranteed way to upgrade my tanks from veteran to elite status. I just started a fight with Germans. I had tanks and Germans were little behind in science, so they only had infantry. I took couple of their cities and started waited for their counter attack. They send a lot (50-100) of units toward captured cities. Most of the units were infantry with some rifleman, swordsman. First biggest stack got on the mountain just in front of the city. I renforced the city defence and started encircling atacking stack with fortresses with several infantry fortified inside. Computer realized it and withdraw (I was suprised at the AI smartness here). I took mountain position.
Here comes an easy upgrade part. On the next turn computer moved one of it's stacks toward another city. He put that stack on plains (no defence bonus), so it was easy for me to attack. I soften this stack somewhat with atrillery and then started sending tanks to destroy units. Apparently if tank wins 2 battles in a row it always upgrades. I had a lot of tanks (about 30) and most of them got upgraded to elite status in two turns. I also got several leaders really fast but it was while attacking different places, so I am not sure if this strategy is good for getting leaders, but at least there will be plenty of elites. You can't use this strategy with units before tanks (I think) because you need to attack twice in a turn (to win twice).

 

[punkbass2000]

It has nothing to do woth attacking twice in one turn. If you were militaristic, I would think you would upgrade even more often than this.

 

[Lovro]

Originally posted by hifisoftware
You can't use this strategy with units before tanks (I think) because you need to attack twice in a turn (to win twice).

Knights or cavalries in an army will do the trick, IIRC.

 

[hifisoftware]

Originally posted by punkbass2000
It has nothing to do woth attacking twice in one turn. If you were militaristic, I would think you would upgrade even more often than this.

Sure it does. Here is a easy comparison that shows it:
In the beggining of the war I used 10 tanks to take city with 4 infantry defending. Since I had to get into enemy teritory I only had one move to attack. I lost one tank. Several retreated while loosing and four of them actually killed infantry. Only one of them got promoted to the elite. So 4 won and one got promoted. 25% chance of pomotion.
When I attacked twice in a turn every single tank got ptomoted. 100%. I did not check this on few tanks. I got practically all of my tanks promoted (maybe 25 got promoted). (These that are still veteran were used to atack some units inside of the enemy teritory, so they could only atack once).
Difference is just too dramatic to ignore. If I was a militaristic I could probably get more leaders, but even in this game with practicaly evey unit being an elite I got 3 leades in about ten turns.

 

[punkbass2000]

Those are not statistically significant results, and I won't accept your hypothesis without 1000 tests, minimum. I think you're just experiencing a statisitical anomaly, especially with that 1 in 4. Seriously, how could 1 in 4 mean anything? With only 4 tests, you results would have to be 0, 25, 50, 75, or 100%. If you had got 3 out of 4, you would be swering up and down not to attack twice in one turn. In addition, your 'statistics' are not comparable because 100% out two attacks is not the same thing as 25% from one attack. I find this hypothesis unlikely at best, so unless it is confirmed by Firaxis or you do a lot more testing, I doubt anyone will accept it as fact.

 

[McBain]

I would like to reiterate punkbass's point. If any of those 100% tanks were preexisting the the orginal attack, then your stats are even more skewed. In this case your upgrade is 1 out of 2 attackes when you when 3 attacks in a row.
1000 tests may seen excessive to establish a trend however. 100 should be enough to formulate the hypothesis, while at least 100x that should prove/disprove it.

 

[Loopy]

Originally posted by punkbass2000
It has nothing to do woth attacking twice in one turn. If you were militaristic, I would think you would upgrade even more often than this.

Actually two successful attacks in the same turn a single unit increases the odds of a promotion tremendously. This is fairly well known and one of the benefits of having the blitz flag.

As for Cavalry or Knights in an army, they're not nearly as useful for two reasons. First, you get more HP by adding units than via promoting within. Second, elite units within armies aren't supposed to be able to spawn leaders. Besides, as documented, an army doesn't grant the blitz flag, it just lets the army continue to blitz if it already has the flag.

 

[punkbass2000]

Originally posted by Loopy


Actually two successful attacks in the same turn a single unit increases the odds of a promotion tremendously. This is fairly well known and one of the benefits of having the blitz flag.

Well known? By whom? How has it been confirmed?

 

[hifisoftware]

There is no way to prove it. Even if a billion (or any number for that matter) of tanks get promoted right after another theory of propability tells you that it really does not mean anything. You can't prove anything mathematicaly by trying it bunch of times. But then again random number generator in PC is anything but random, so 30 results in a row is enough to establish a pattern.
Your mistake is that you try to apply statistics which is too unpresise to care about. Ever wonder why two different medical trials come to opposite conclusions? Well that because they use statistics. Theory of propability tells you that statistical analysis can come with wrong answer every time for n number of years. So trust me never relly on what statistics tells you.
Plus apparently this is well known. I am wondering though if this got added in the last patch.
P.S This has nothing to do with Civ though. Just try it and you'll see.

 

[Lovro]

Originally posted by hifisoftware
But then again random number generator in PC is anything but random, so 30 results in a row is enough to establish a pattern.

There is no such thing as a random number generator in PC.

Random number generators do what people program them to do, not what they decide to do.

 

[alexander dumas]

Why is everyone smacking on each other about statistics? The core of the original post stands: get better units through a combination of attacking on home turf and use of artillery. That's not that hard.

 

[punkbass2000]

That was not the 'core' of the post. His thesis was that attacking twice in one turn causes promotions.

 

[punkbass2000]

Originally posted by hifisoftware
There is no way to prove it. Even if a billion (or any number for that matter) of tanks get promoted right after another theory of propability tells you that it really does not mean anything. You can't prove anything mathematicaly by trying it bunch of times. But then again random number generator in PC is anything but random, so 30 results in a row is enough to establish a pattern.
Your mistake is that you try to apply statistics which is too unpresise to care about. Ever wonder why two different medical trials come to opposite conclusions? Well that because they use statistics. Theory of propability tells you that statistical analysis can come with wrong answer every time for n number of years. So trust me never relly on what statistics tells you.
Plus apparently this is well known. I am wondering though if this got added in the last patch.
P.S This has nothing to do with Civ though. Just try it and you'll see.

I know there is no way to 'prove' it. There is no way to prove anything. However, if you could demonstrate it with a large number of units, it is almost certainly true. If not, then it is almost certainly not true.

 

[Evincar]

Originally posted by punkbass2000


Well known? By whom? How has it been confirmed? [/B]

Punkbass, almost everybody knows that 2 victories in the same turn mean promotion. Well, you want a prove do you?

Try it 1000 times. It will work 1000 times. End of subject.

Ahhhhhhhh, I love studying engineering

I forgot to tell this: If your unit is promoted after the FIRST battle, your unit won't be necessarily promoted after the second. However, if your unit is not promoted at first, promotion is guaranteed after second battle. This means that, after two battles, your unit has promoted AT LEAST once (two if you are lucky )

 

[punkbass2000]

I refuse to accept the argument 'almost everyone knows'. Everyone 'knew' that the Earth was flat, once. I'm not saying it isn't true, but that I won't buy it without proof.

 

[Evincar]

Originally posted by punkbass2000
I refuse to accept the argument 'almost everyone knows'. Everyone 'knew' that the Earth was flat, once. I'm not saying it isn't true, but that I won't buy it without proof.

No no no no no no no no no...etc

Nobody "knew" the Earth was flat. They just decided it because...well, because they liked Earth being flat.

However, I "know" what I said because i've seen it happens once, twice, again, and as many times as you want to try it

 

[JamieE]

The above claims that you cannot prove anything through statistics are incorrect. In this case you would test the null hypothesis that the probability of getting an elite by atacking twice on different turns is equal to the probability of attacking twice on the same turn. A simple t-test would be suffice.

You are correct though that an n of 4 is too low to say anything conclusive. An n of 1000 would almost certainly be overkill though, given the heuristic observations that I and others have seen. I would guesstimate that a sample size of ~10 would give you a 95% confidence.

The reason that epidemiological (medical) studies often report conflicting results is that they are trying to tease out minute differences in probaility between groups in situations where there are a) millions of possible confounding variables and b) ethecal constraints that make it impermissible to do experiments.

End of math geek rant

 

[punkbass2000]

I think 10 is too low. It might hint at the idea, but 95% confidence is too much, IMO.

 

[TheNiceOne]

There's actually a very simple way to prove it if this theory is true:

1) Be sure that the game is started with "save random seed" on.

2) Start with a situation where you have two veteran tanks ready to beat at least two enemy units. Neither of those tanks should have been in battle earlier.

3) Attack twice with one tanks,

4) Reload, then:

5) Attack once with each tanks.

This will use exactly the same random number to calculate whether the second win will promote the attacking tank. If 3) results in a leader, but 5 don't then were having shown that attacking twice in a row increases the chance of being promoted.

 

[Allemand]

If a unit attacks twice in one turn or is attacked twice in one turn, and it wins both times, it is promoted. This always works.

Moreover, if the unit wins two defensive battles (getting promoted) and is attacked a third time and wins, then it is promoted again. This always works.

The only exception is that this doesn't apply to getting leaders.

Rather than argue with each other, just try it in your games. It always works.

I usually cultivate strong barbarian camps, keep them isolated from the other civs, and send teams of spearmen out to train against them. I love it when there is a massive uprising. I get lots of promotions.