leader drought

[Sebastian]

1) The chance of getting a leader desperately needs to be scaled to the size of the map.

2) If you go 100 elite victories without getting a leader, all it probably means is that you got 2 leaders in 102 elite victories. Big deal. Saying you got a drought of 100 (.16% chance) is a big difference from 2 in 102, which is much more common... though still less than the average (about 6). Both are true, but one is slightly misleading. =P

 

[MuddyOne]

A lot of people bet a dollar that worse odds than you are reporting will happen. The lottery, a tax on people that can't do math.

What are the odds on getting hit by lightning? Last I heard the worlds record was a person was struck 7 times at different locations and times. None of the times did the times did he seem to be in any particularly dangerous spot.

A particular professor has a 99% success rate at determining whether students recorded results at flipping a coin 100 times, or wrote what they though the results would be. Randomness is often "streaky".

 

[Lovro]

Originally posted by Sebastian
2) If you go 100 elite victories without getting a leader, all it probably means is that you got 2 leaders in 102 elite victories.

If I understand what you are saying, then I have to disagree

If you hadn't gotten a leader in 100 elite victories in a row, the odds of getting 2 leaders out of the next two battles are 0.39%.

 

[MuddyOne]

He is saying on combat one you get a leader, combat 2 through 101 no leader, and on combat 102 you get a leader.

I have had 100 combats with no leader is a correct statement for the above situation, but is misleading.

 

[etj4Eagle]

Originally posted by MuddyOne
A lot of people bet a dollar that worse odds than you are reporting will happen. The lottery, a tax on people that can't do math.[/quote.]

One of my cheme professors loved to describe the lottery as a tax on stupid people. Especially since they also comitt fraud in declaring the value of the jackpot. Any other group would have to discount the future value of the money stream, but the government just pretends a dollar ten years from now is worth the same as a dollar today.

What are the odds on getting hit by lightning? Last I heard the worlds record was a person was struck 7 times at different locations and times. None of the times did the times did he seem to be in any particularly dangerous spot.[/qoute]

Unlikely events do occur and people also worry about and complain about events with a 1/million chance of occuring.

A particular professor has a 99% success rate at determining whether students recorded results at flipping a coin 100 times, or wrote what they though the results would be. Randomness is often "streaky".

Yep, people make random numbers "too" random. There are even programs written on this principle that audit documents based on the random number principle. By there very nature random numbers are streaky, otherswise how do you get those 1/100 and 1/500 and 1/100000000 occurances to occur.

 

[sumthinelse]

Originally posted by Lovro
If I understand what you are saying, then I have to disagree

If you hadn't gotten a leader in 100 elite victories in a row, the odds of getting 2 leaders out of the next two battles are 0.39%.

Or the odds are zero of getting a leader on the next 2 tries. When the odds are zero on a given turn, then the probability of getting a leader out of 1000 elite victories is zero. That is the kind of thing I am observing. On some turns, the odds are 1/16 on a given turn. On others, zero.

 

[sumthinelse]

Originally posted by Sebastian


If you go 100 elite victories without getting a leader, all it probably means is that you got 2 leaders in 102 elite victories. Big deal. Saying you got a drought of 100 (.16% chance) is a big difference from 2 in 102, which is much more common... though still less than the average (about 6). Both are true, but one is slightly misleading. =P

I'll make the same challenge to you I did before. I'll send you my saved game. You load it and attack 3 archers with 3 elite horsement. Reload and do it as many times as you like. After 20 victories, you would think the odds of generating a leader is about 1/4, after 40, 1/13. Try it and see how long it takes you. OK? I claim that the probability of generating a leader on this turn are 0.

 

[Zachriel]

Originally posted by sumthinelse


I'll make the same challenge to you I did before. I'll send you my saved game. You load it and attack 3 archers with 3 elite horsement. Reload and do it as many times as you like. After 20 victories, you would think the odds of generating a leader is about 1/4, after 40, 1/13. Try it and see how long it takes you. OK? I claim that the probability of generating a leader on this turn are 0.

When you reload, you are reloading the same randomizer "seed," so the combat results will be the same. So you are correct, if it is all unsuccessful the first time, you can reload forever and never get a Leader. But if you have 1000 unique combats, the results will be non-zero.

 

[sumthinelse]

Originally posted by Zachriel


When you reload, you are reloading the same randomizer "seed," so the combat results will be the same. So you are correct, if it is all unsuccessful the first time, you can reload forever and never get a Leader. But if you have 1000 unique combats, the results will be non-zero.

Didn't you notice that 1.21f changed the way the "seed" works? I am not playing a scenario -- you have to edit the .bic file to make the seed change in a scenario. Did you actually test the game, or are you telling me what you believe? Load a saved game, a regular game, not a scenario, and try a combat between fairly evenly matched units, like spearman vs. spearman in the open (to make the results more likely to be different). Reload and redo the combat 20 times and tell me that you get the same result every time.

I can also send you a saved game where the creation of a leader is possible. If you retry the combat many times, the leader will appear after a random number of turns, sometimes 3 turns, sometimes 12 turns, sometimes 28 turns. Almost certain to get a different result each time you reload.

Or maybe you are saying that you have seen the game's source code, and sometimes it uses random seeds for leader generation and sometimes it preserves the seed?

 

[sumthinelse]

Originally posted by MuddyOne
He is saying on combat one you get a leader, combat 2 through 101 no leader, and on combat 102 you get a leader.

I have had 100 combats with no leader is a correct statement for the above situation, but is misleading.

I realize that some posts complain about a string of losses, and these posts may ignore the times that their units won a string of seemingly improbable victories. I agree that it is very easy to be subjective about such results.

There might be some strange combat situations in which the probabilities are not what they are supposed to be,
but as far as I know the combat results (just who gets hurt in combat. not talking about the creation of a leader) are
more or less random, and somewhere in the ballpark of where they are supposed to be.

"Occam's Razor" means reducing your assumptions to a minimum.

Suppose you were open minded in this matter. Suppose you did not assume you knew what the odds were of creating a leader, that the probabilities might change from one combat to the next, and that you did not assume that it was even possible to create a leader in any specific elite victory.

I can choose any starting point and any ending point I like in my testing, if I record the results carefully and honestly. It is logical to begin in the first elite victory after I use up a leader, since while a leader exists no leader is supposed to be generated. I choose as my ending point the first elite victory that creates a leader.

The previous "leader drought" was 155 elite victories, probability 1/22,000 assuming that the game actually does use 1/16 for leader creation.

In my current game, I have 278 elite victories in a row without creating a leader. Probability: 1/62,000,000 assuming that the game actually does use 1/16 for leader creation. In fact, I don't even know if it will be possible to create a leader in the rest of the game. I am just playing the game, and carefully recording the results. I am NOT ignoring results that do not fit my hypothesis (which is that certain combat situations cannot create a leader).

How would the game behave if the probability of creating a leader was zero on some combat situations? Exactly like what I am seeing. What is the simplest explanation of what I am seeing (Occam's Razor again)? You might say I am unlucky, but that is getting hard to believe. Killer reported 1000 elite victories without a leader in his game -- ever. Odds: 1 in (1 with 28 zeroes after it). I think the mass of the sun is something like 10 to the 18th power (quick calculation, may be a little less than that). How big do the numbers have to be to convince you?

 

[Lovro]

Originally posted by sumthinelse
Didn't you notice that 1.21f changed the way the "seed" works? I am not playing a scenario -- you have to edit the .bic file to make the seed change in a scenario. Did you actually test the game, or are you telling me what you believe? Load a saved game, a regular game, not a scenario, and try a combat between fairly evenly matched units, like spearman vs. spearman in the open (to make the results more likely to be different). Reload and redo the combat 20 times and tell me that you get the same result every time.

>cut<

Or maybe you are saying that you have seen the game's source code, and sometimes it uses random seeds for leader generation and sometimes it preserves the seed?

The game seed *is* saved with the game. Reloading will (should?) result in the same results over and over, as long as you don't change anything in the order of actions (i.e. "wait" with your spearman, do something else that involves randomness and then come back to attack the spearman).

This has been rather obvious since the release of the game and even more in v1.21 where you were given the option "Preserve random seed", which is turned on by default. Turning it off before starting the game will result in different combat results when reloading.

AFAIK, of course

Could you tell us which units are involved in your leaderless combats? Are they from different ages? Have you tried starting a new game and setting up a similar situation? How many leaders have you had in that game so far?

 

[sumthinelse]

Originally posted by Lovro
The game seed *is* saved with the game. Reloading will (should?) result in the same results over and over, as long as you don't change anything in the order of actions (i.e. "wait" with your spearman, do something else that involves randomness and then come back to attack the spearman).

This has been rather obvious since the release of the game and even more in v1.21 where you were given the option "Preserve random seed", which is turned on by default. Turning it off before starting the game will result in different combat results when reloading.

AFAIK, of course

Could you tell us which units are involved in your leaderless combats? Are they from different ages? Have you tried starting a new game and setting up a similar situation? How many leaders have you had in that game so far?

(Reediting on 11 May)

The real problem is that there is an apparently undocumented requirement for creating a leader. If you attack with an elite unit that previously generated a leader, the probability of creating a leader by that unit is zero - forever.

OK, now I see what you mean about the seed. I loaded a 1.17f game into the 1.12f patch and it turned OFF the random seed by default. Therefore I assumed incorrectly that if you start a NEW game in 1.21f that the default was "ramdom seed" OFF, since I could not find a "random seed" option for regular games in the editor. Then I started a new game and noticed the new start option for "preserve random seed."

This may mean that if you load a 1.17f game in 1.21f, you always get new combat results every time you reload.

Or is it possible that the outcome of whether you get "remember randon seed" is itself random ???

This is what I wrote before:

There is only one option I can find in the editor for "preserve random seed", and it is for SCENARIO ONLY! I am playing a regular game.

 

[Lovro]

If you manage to recreate the exact same battles, the results will/should indeed be the same every time you play.

However, once you change a thing in your game that involves randomness, you will change the random seed and the results will seem 'random', although it was your "fault".

For example, if you had 5 actions that included randomness in this order:

1-2-3-4-5

...and then you decided to include another in the midst of it like this:

1-2-intruder-3-4-5

...then the new action would get the random number originally meant for action 3, while action 4 *might* get the random number initially meant for action 3. I think.

Another example for this is, say you block one of the AI's units' path with one of your own units. The AI's algorithms might include randomness, so your path blocking may result in your saltpeter source being exhausted

We don't know how civ3 handles random numbers so we can't know for sure.

 

[Lt. 'Killer' M.]

'default' buggy?????


I used to have trouble with setting governors as 'default', then switching them off - never worked.

Now a friend showed me his Civ3 where if he un-checked 'Preserve random seed' it was still on but the next game whatever he did it was off (even if checked!!!!)

sumthinelse: you got that too, maybe?????

 

[JoeM]

Hi,

Slightly different tack which =DOCTOR= reminded me of.

I've started using a tactic to produce multiple leaders in a single turn, it's only useful when there's plenty of railroads around though.

Once a Great Leader is generated I use it up straigh away that turn then initiate another Elite combat and *always* get another leader. I haven't tested it ad infinitum, and with the amount of troops I usually have I haven't had the means to generate more than four leaders in a single turn, however I always generate repeated leaders when I have the means to do so.

Can anyone confirm/refute this behaviour? (v1.17f)

 

[Lt. 'Killer' M.]

Originally posted by JoeM
Hi,

Slightly different tack which =DOCTOR= reminded me of.

I've started using a tactic to produce multiple leaders in a single turn, it's only useful when there's plenty of railroads around though.

Once a Great Leader is generated I use it up straigh away that turn then initiate another Elite combat and *always* get another leader. I haven't tested it ad infinitum, and with the amount of troops I usually have I haven't had the means to generate more than four leaders in a single turn, however I always generate repeated leaders when I have the means to do so.

Can anyone confirm/refute this behaviour? (v1.17f)

WOW!

My leader generation seems normal (that is repeat leader do happen, but rarely), but this really takes the cake!

Conclusion from all I´ve heard:

The RNG is system dependend!

 

[damunzy]

Originally posted by Killer
The RNG is system dependend!

That is what I suspected

 

[Hurricane]

I have read several independent posts were it was confirmed with very good statistical proof that the RNG in Civ 3 was not broke, and that it was NOT system dependant. So I don´t think there is any need to debate this one more time. THE RANDOM NUMBER GENERATOR WORKS PERFECTLY WELL.

Now, if a person regularly does not get a leader for hundreds of battles, something strange is clearly going on. The ones I can think of:

* The RNG doesn´t change the seed even if it is supposed to. (Note that a pre v1.21 saved game will be re-seeded under v1.21.)

* The Available leader flag is switched even if the leader is used.

* Bad luck

If I understand Sumthinelse right, all these conditions have been checked and bad luck is the only unclear one. But that sure seems strange. Care to post a few savegames?

 

[sumthinelse]

Originally posted by Hurricane


Now, if a person regularly does not get a leader for hundreds of battles, something strange is clearly going on. The ones I can think of:

* The RNG doesn´t change the seed even if it is supposed to. (Note that a pre v1.21 saved game will be re-seeded under v1.21.)

* The Available leader flag is switched even if the leader is used.

* Bad luck

If I understand Sumthinelse right, all these conditions have been checked and bad luck is the only unclear one. But that sure seems strange. Care to post a few savegames?

As of now I believe that the RNG is OK, and the seed seems to be working like it's supposed to, that is, new seed every time I reload.

I can upload some saved games, but I think I have an explanation for this behavior, maybe. Something Killer said gave me an idea. He said that he had better luck creating leaders attacking with a variety of unit types. I was using elite horsemen exclusively. Then the idea came to me: all of the elite units I was attacking with might have produced leaders before, and maybe you can get only one leader per elite unit, ever. So I created a new horse unit, got it to elite, attacked, and... presto! a leader in 3 tries. I will reload the game to see how many times it takes to get a leader next time.

If this checks out, it is damned tedious to try to determine which elite units have produced leaders and which ones have not. All the horse units look identical to me except for hit points and movement left. They should make elites that have produced leaders visibly different.

What is the url for uploads here?

 

[sumthinelse]

Originally posted by Killer
'default' buggy?????


I used to have trouble with setting governors as 'default', then switching them off - never worked.

Now a friend showed me his Civ3 where if he un-checked 'Preserve random seed' it was still on but the next game whatever he did it was off (even if checked!!!!)

sumthinelse: you got that too, maybe?????

(Reediting on 11 May)

I loaded a 1.17f regular game (not a scenario) in patch 1.21f, so I didn't get to select the "random seed" option. All the combats, etc. are random every time I reload. Then I started 2 new 1.21f regular games, one with random seed on and the other with random seed off. In both the new games the random seed option seemed to work as expected. But why is the default ON in a new game and OFF in a game loaded from 1.17f? I think they should have made a .bic file option for games loaded from 1.17f.

I think I may have solved the mystery about the leaders, though, and I got my idea from you. The post that explains was the one before this one. Thanks!