Only 1 leader possible from any given elite

[etj4Eagle]

Originally posted by Lt. 'Killer' M.


what this reminds me of is Falcon 4.0. On some systems, it wouldn't run properly, you'd always crash on landing.
On others it ran smooth as silk. So it got some outside input into whatever, maybe a variable of some kind, that lead to a Bug appearing, while other systems didn't show that.....

I 'solved' my problem accidently when I had to replace my main board.....

Those system dependent bugs probably have to be the most annoying to crush. Especially when the developers don't have a system that creates that kind of bug. There should be a way to test for this kind of behavoir in the game. One possibility would be for me to make another test map. And then record the battle progression and see if it behaves similarly on your system. (To simplify things I would look only at victories between events like promotions and leader generations).

Assuming no system specific bug (other than maybe in seeding the generator) are two tries should play out exactly the same. (perserve the random seed of course will be on).

 

[Lt. 'Killer' M.]

Originally posted by etj4Eagle


Those system dependent bugs probably have to be the most annoying to crush. Especially when the developers don't have a system that creates that kind of bug. There should be a way to test for this kind of behavoir in the game. One possibility would be for me to make another test map. And then record the battle progression and see if it behaves similarly on your system. (To simplify things I would look only at victories between events like promotions and leader generations).

Assuming no system specific bug (other than maybe in seeding the generator) are two tries should play out exactly the same. (perserve the random seed of course will be on).

and then we'd have to do the same thing again with seed off, after all the leader droughts appeared in the tests I did with seed off......

Would that be too much work for you?

 

[sumthinelse]

Originally posted by etj4Eagle


One possibility would be for me to make another test map. And then record the battle progression and see if it behaves similarly on your system.

You could try my scenarios that I created and uploaded on this thread a few pages back. It has the random seed off, but you could use the Gramphos editor to toggle that. If there is a long wait at the start, then you won't get a leader in the 1st 25 attacks.

It has 25 elite modern armor in position to attack 25 regular warriors. There are 2 saves, the same except one has the heroic epic and the other doesn't.

Without the epic, I got a leader in 1 (!), 34, 19, 1 (!), 3, and 33 tries, which is 6/91, about 1/15. Yes I know it's a small sample.

With the epic I got 19,15,3,2,13,11,22, which is 8/75, about 1/12.

Killer, I am going to add a section to Tutorials, Reference, & Guides. about leaders eventually but I have to go to the embassy tomorrow.

 

[kring]

I have had very few Leaders myself. Since I automatically use the Leader to either build an Army or (usually) complete a Wonder, that's not the problem. Also, since they are so rare, using a ELITE is not the problem since they are not that many of them.

Based on the number of replies I have seen between CivFanatics, Apolyton, 1BC, and about 9 other sites, it probably does average 1/16 and 1/12; it is just that some get them much more frequently than others; some less so.

The fact that I rarely get more than a couple of Leaders a game is offset by those who get the 1/8 or 1/10 ratios.

 

[sumthinelse]

Originally posted by Lt. 'Killer' M.

I always ahev long waits for the first leader. Like 35, 40, 28 fights

Killer do you have a saved game that you think behaves like this?

An idea I had was that if you use fewer units to attack with you are more likely to get "stringy" results where the RNG or something about leader generation appears to be "stuck" because somehow the units have something unique about them that affects the outcome.

My example was flipping warped coins. If you flip one warped coin 100 times you might get a 70 tails/30 heads distribution because the coin was heavier (or rounder) on one side than another. But If I have 10 coins and flip them 5 times each (total 50 trials), if some of them are warped on the "heads" side and others are warped on the "tails" side the unique properties of the individual coins might cancel each other out, and I might get 23 tails/27 heads, which is nearer the expected result of 50% tails, 50% heads, even though I have a smaller number of total trials.

OK, if this were true then , since there are fewer units at the start of a game, then you would expect to get "rougher" results at the start of a game than later, when you were using more elite units.

And another explanation is that you just had bad luck, of course .

 

[Lt. 'Killer' M.]

@ sumthinelse:

I get that behaviour also when I start a fresh game, then reload with seed off thus doing the attack with the same unit over and over. If I then reload an older game that has had leaders, then go back to the new one, it's better. Just a feeling though, tests with significance for this would take far too much time. But I do get the feeling somthing is strange here, and I don't think it's a feature (as the blocked elites). I played a game a day this week, always Immortal rush on tiny world with 16 civ. In none of the 6 did I get a leader fast (that is in less than 16 fights) and only one came before win No. 20. The rest took awfully long.
In none of the games did I get more than 3 leaders, despite over 200 elite wins per game! And I DID check for used elites, I didn't count their fights!
The sole exception was a game where I got a 4th leader on defence that died in the next attack.

So I'm back to the point where we once started: do I get fewer leaders because I keep using the one and same troop type all the time??????

 

[sumthinelse]

Originally posted by Lt. 'Killer' M.


tests with significance for this would take far too much time.

Yes, we have a dilemma.

If I use my scenario attached in this forum, I get the expected 1/16 and 1/12 results and it doesn't take long to do, because it's a tiny world with 25 elite attacks ready to go where the attackers always win. (You should try it on your machine to see if you get the same results. I'll bet we have different hardware and I am running win2000).

But if we suspect the problem (when I say "problem" it might mean that it takes a long time to get a leader or it happens very quickly) occurs when there are just a few elite units (1 or 2 or 3?), what do we do? Keep starting new games and recording results? The last time I started a game it was in 1.17f and I loaded it into 1.21f before I got the 1st leader. I got the first leader (playing my 1.17f game in 1.21f) in 1300 BC after only 11 elite victories, without the epic wonder, of course.

So I'm back to the point where we once started: do I get fewer leaders because I keep using the one and same troop type all the time??????

Unfortunately (as you know), having one elite unit attack less than 50 times is a small statistical sample. I think you are saying that you got 5 initial leaders in - what- about 120 tries? 150 tries?, with one initial leader created on defense in less than 16 tries?

 

[Sanaz]

Originally posted by sumthinelse
Yes, we have a dilemma.

How about setting up two tests :

(1) 50 units, movement 1 and no blitz, high hit points and attack, and then attack one with unit at a time until a leader is produced

(2) 10 units, but with Blitz and movement 100. The same unit could attack over and over until it produces a leader (likely less than 100 victories).

This should cover the two test cases of many units winning one at a time, and very few units winning repeatedly.

 

[Lt. 'Killer' M.]

@ sumthinelse

I tried your game (w/o HE), and the first time I got the expected number of leaders: 15,8,24,13,19,17,3,18,14,5,20,17,8,19

then, I quit windows, rebooted, restarted civ3 and started that game right away (w/o HE)
200 tries, queue was 33,17,44,1,1,2,18,19,11,5,22

so again, a bad start....


similar thing in a fresh game I started afterwards (!) 1st elite win (defensive, no less!) = leader!
then, average of 1 leader every 12 wins without HE, average of 1 in 10 with HE!


so I went and replayyed the same game after a restart of the enitre system, and got a bad string, I conquered the world without getting a single leader!

Probably it is just bad luck, but it was a very strange conicidence....
So again,

 

[sumthinelse]

Originally posted by PaleHorse76

I've suggested this to someone else for his findings and thought I would suggest it to you guys also:

I suggest that you post your findings in Tutorials, Reference, & Guides.

OK, here it is:

http://forums.civfanatics.com/showthread.php?postid=317257#post317257

 

[Lt. 'Killer' M.]

sumthinelse, wait 5 min then check for new threads!!!!!!! there's something you won't believe either!!!!!!

 

[sumthinelse]

Originally posted by Lt. 'Killer' M.

I tried your game (w/o HE), and the first time I got the expected number of leaders: 15,8,24,13,19,17,3,18,14,5,20,17,8,19

then, I quit windows, rebooted, restarted civ3 and started that game right away (w/o HE)
200 tries, queue was 33,17,44,1,1,2,18,19,11,5,22

These don't look too unusual to me.... Since the odds of one leader is 1/16, the odds of 2 leaders in a row is 1/256, so I'm not surprised you got that at one point.

so I went and replayyed the same game after a restart of the enitre system, and got a bad string, I conquered the world without getting a single leader!

How many elite victories?

 

[Lt. 'Killer' M.]

Originally posted by sumthinelse


These don't look too unusual to me.... Since the odds of one leader is 1/16, the odds of 2 leaders in a row is 1/256, so I'm not surprised you got that at one point.



How many elite victories?

my point was the 33, 17,44 string I had. I always see that5 when I start Civ fresh, never when I quit a game then start another. I know it isn't statistically significant, but still it is strange.

how many victories? I quit counting somwhere around 50, then just killed off 6 more civs.... So I'd say around 100. Bad day, today
But this really was bad luck, I guess.

Did you see the AI unit chaet thing????? Could you check it out please???? I'm busy adding pics right now...

 

[kring]

Originally posted by sumthinelse


Yes, we have a dilemma.

If I use my scenario attached in this forum, I get the expected 1/16 and 1/12 results and it doesn't take long to do, because it's a tiny world with 25 elite attacks ready to go where the attackers always win. (You should try it on your machine to see if you get the same results. I'll bet we have different hardware and I am running win2000).

But if we suspect the problem (when I say "problem" it might mean that it takes a long time to get a leader or it happens very quickly) occurs when there are just a few elite units (1 or 2 or 3?), what do we do? Keep starting new games and recording results? The last time I started a game it was in 1.17f and I loaded it into 1.21f before I got the 1st leader. I got the first leader (playing my 1.17f game in 1.21f) in 1300 BC after only 11 elite victories, without the epic wonder, of course.



Unfortunately (as you know), having one elite unit attack less than 50 times is a small statistical sample. I think you are saying that you got 5 initial leaders in - what- about 120 tries? 150 tries?, with one initial leader created on defense in less than 16 tries?

The problem here is the same, not a big enough sample. I have personally seen 2 Leaders close together, that was offset by a long string of no Leaders. A good sample would require several thousand replays. Often times, people see something that is not there; an apparent pattern, that is nothing more than coincidence and not a pattern.

Unless someone shows me coding that indicates less than a 1/16, 1/12 or 1/32, 1/24 probability or a different probability if you do things a different way, I am willing to take them at their word. Not worth my time to do several thousand tests to prove it. I would rather play the game or work on my mods.

 

[sumthinelse]

Originally posted by kring


The problem here is the same, not a big enough sample

Killer, I think kring is right that you need a bigger sample. But how would you do that? You would have to play 100 games? I guess you will play 100 more games eventually . But you have to promise to write down carefully how long it took for the 1st leader in all of them, not just the ones where it took a long time.

 

[Lt. 'Killer' M.]

sumthinelse:

Right, it will take alwfully long, and I will have to be carefull not write down if it was a 'fresh' game or if I played another before, and how long it took for the first leader and if I only used one troop type or only UUs or mixed or what.... and all that, so I guess somthing of statistical significance will take years But I will play 100 more tgames, no sweat!

 

[kring]

This is a bump since I think this is a good thing for people to know about creating GL. I know there is another thread out as well, but it doesn't hurt to keep them both going, since not everyone goes to both fora.

 

[sumthinelse]

There are really 3 threads. The other one is http://forums.civfanatics.com/showthread.php?s=&threadid=24371

 

[kring]

Three? I knew about the one you linked. Is there another after that one? I bounce between CFC, Poly, and 1BC so much it is hard to keep track of what I saw where.

 

[kring]

Here is a game where an artillery with lethal bombard turned on creates a Leader. This was found by a player at 1BC during the 1BC June GotM.


"There's a stack of artillery in the northern part of yellow Egyptian territory. Right-click on the stack, and choose the third artillery unit from the top of the list. (So not the one that has the ring around it already.) With the chosen artillery kill the Cavalry that's near the Russian town; and Peter the Great will emerge; just like he did when I killed the Cavalry with an elite Cossack of my own. (Which is what I'll again do when I restart the turn from the autosave-)-"