New player, some thoughts/questions

[Valkrionn]

Valk : I think 127 Gp is a large enough sample for probability to be represented. IMO, his numbers are interesting. And the laws of probability should have kicked in.

On the Healer note : I noticed in on ecity I captured that had a great healer settled : +6 so a GH gives you health for 3 more populations points... meaning having a GH is well worth a GS or a GM... (3 pop is as a mean rule : 2 2 + 1-2 food:or more ) 2 are not to be counted as they are for the pop. a GH is worth AT LEAST 6 + the bonus food:or more ) times 3.

....No, it is not. Did you read my post?

Without 'New Random Seed on Reload', numbers in civ4 are not truly random. Even with that option, the only time they are close to truly random is when you reload the game. Otherwise, you can get in situations where the seed leads to a string of bad luck quite easily.

You want a sampling that will show the laws of probability? Turn on that option, and save/reload for the same specialist around a 100 times. THEN it will more or less match what the odds say.

127 GPs says nothing about any probability, as we don't know what chances each had.
To see if a certain outcome is random, you would have to know the individual chances.
Besides, 127 isn't really enough to do anything but get an idea (no definitive answer, and you can't actually prove that anything is random, only that it most likely is or is not).
The RNG of CivIV have been tested on these boards several times, and none of those have shown it to be screwed (Even though Obsolete claims that Great Bards are favoured). I don't think anyone have tested the GP distribution though.

A great healer isn't that good.
If you are above your health limit, each pop (man, I've been playing too much Victoria 2 ) cost 4 , so a Great Healer amounts to a net gain of 1.5 pop.

It's frustrating to micro GP/specialists, mainly because there is no "avoid or ban this GP/specialist" button.
It's annoying in vanilla CivIV, and it's annoying in any mod, and that is a feature that should have been included in the game the moment Great People were introduced.

Healers have been improved somewhat, as the math does show that they are useless (Healer, not Great Healer, though both got improved). Healer is now 4 1, GH is 6 3.

I disagree : 127 is large enough that a majoritary GP (more than 50%) should appear at roughly 50% chances, and minoritary (less than 20% should appear at 20%.. and middle means if I understood correctly : not the best GP (higher %) but not the least (smaller %).
but
Sarisin reported :
15 of them have come with the highest percentage of type, yet 67 came with the lowest percentage. The other 45 fell in the middle somewhere.

It means : more than 50 % were of the lowest percentage comprised between 45 % at best and 5% with a mean of 20% (taking my games into account to have a rought value for percentage)
less than 12 % were of the highest percentage, comprised between 100% and 40% at minimal value.
then , 35 % for values between 10% and 35 % (when you have 3+ gpp types) seems ok.
Anybody can see that 40-100 % giving 12 % of GP and 5-45% giving 50% of GP is broken.
and 127 GP is a high enough population to apply the law of statistics.

you are almost right but not totally.
in your calculus a GH PAYS IN FULL for 1.5 pop. meaning it pays AT LEAST (2, 1, 2) x 1.5 = 4.5/ + 3 that is better than a settled GS or settled GC.
But in reality a GH pays the of 3 pop, meaning it allows your city to rougly grow 3 pop more so that you can again fall into your 1 pop = 4 equilibrium...
you have for 12 pop, + surplus food so a 15pop city can be reached: with a settled GH your pop can almost always grow to size 17-18 depending on the city terrain.

The only way my reasonning is wrong is when you don't have any more tile to put a new citizen to work. Then a new pop cannot pay for its own 2. So there, a GH is worth only 1.5 specialist. and depending on the civics and wonders it might be worth less than a GC/GS but it might not be as a GH still brings some in.
BUT in any case, when your city cannot work any more tiles, I don't think one has to be angry that the city cannot become so much bigger. It is normal IMO.

Again: When working with the same seed, nothing is guaranteed.

Sarisin might be up to something there.
According to this strategy article, "the number of GPPs will determine when a great person will be generated, and the number of sources will determine what type it will be".
But the tooltip/progress bar percentage prediction in Rife is not working this way, but reflects the proportion of produced GPPs of each type.
This could easily explain the discrepancy observed by Sarisin, as "minority" GP types tend to be sourced by buildings with 1/2 GPP each, which therefore would be underestimated by the percentage prediction.

That was just dumbed down text. A random number is compared against the total amount of GP points. The more 'typed' to each GP, the more likely that GP is; If you have multiple 0.5pt sources, then you have less GP points for it. It works exactly as the bar shows.

 

[tokala]

Without 'New Random Seed on Reload', numbers in civ4 are not truly random. Even with that option, the only time they are close to truly random is when you reload the game. Otherwise, you can get in situations where the seed leads to a string of bad luck quite easily.

You want a sampling that will show the laws of probability? Turn on that option, and save/reload for the same specialist around a 100 times. THEN it will more or less match what the odds say.

Seems the mentioned strategy article is severely outdated, and BTS changed the GP generation mechanism to the more intuitive way: "In other words, where 20 turns of oracle plus 20 turns of scientists used to leave you at 50/50, now you are at 40/60 favoring the scientist."

Did a test set up as proposed by Valk for 120 turns with 20/60/20% for Bard/Merchant/Priest, got 24/68/28. Looks like the system is working just fine

 

[Swinkscalibur]

Not to be annoying, but as a math guy is there a reason that in 120 turns there are only 118 gp's?

 

[Tsathoggua]

You want a sampling that will show the laws of probability? Turn on that option, and save/reload for the same specialist around a 100 times. THEN it will more or less match what the odds say.

Even this would not be enough, because there would be a simple correlation between seeds due to the fact that they'd all be moments on the clock in close sequence.

The best you could do for true randomness is to send the save file off to 100 people and tell them to open it some time in the next day, or couple of days. Even then, your ignorance of the system clock values is not so much pure randomness as it is a justification for *assuming* pure randomness.

All of this would be pointless of course because as Valk has stressed, true randomness simply does not exist within the machinery of a computer.

 

[Valkrionn]

Pretty much, yes.

 

[tokala]

Not to be annoying, but as a math guy is there a reason that in 120 turns there are only 118 gp's?

Corrected

lost 2 reloads while adding it up

And while it might be not truly random, it looks way better than Sarisin's sample. And as each iteration took roughly half a minute, the times should be not to similiar.

 

[Sarisin]

That is simply the laws of probability. Get a larger sample size, and it will work out as it should.

The random numbers in Civ are NOT random... In fact, no computer produces random numbers. Ever. They follow a formula involving a 'seed'; each new rand is in sequence from the old one. Based on your seed, you may consistently get unlikely results.

That's the whole point of 'new random seed on reload'; Each time you reload the seed is reset based on the system clock. If you have that on, and reload, say, 100 times, the results should reflect the odds the game gives you.

I don't have that option selected.

What do you consider a suitable size sample? You will likely recall I have been bringing this up some time, so I would guess if I logged every time the sample size would be close to 1,000 or more with similar results.


Again, you are missing my main point, V, and that is the strategy element of what is supposed to be a strategy game. What good does it do you in this type of game to plan and strategize if you know you are going to be whipped by some random element almost every time?

Throwing random elements into a game are necessary to be sure, but I think there comes a point when you have to decide if this will be a game of strategy or just dice-rolling.

Edit: After going back and reading further comments...

All I can say is what I said when I brought up the Blight problem in 1.30...I am simply reporting what I am seeing in my games. I would encourage other players to take notice in their next game when GPs are produced looking at the various odds. You don't have to be anal like me and keep a notepad nearby to track the results. Again, it is not so much the issue of the game being broken (not in the bug thread), but perhaps just bringing up one way it is a waste of time to strategize in a supposed strategy game, when your results usually are unexpected. There are plenty of random things in the game that I don't mind at all, my starting position for example. You just go ahead with that (or restart, I suppose) and play your game. However, when you plan your city to produce a certain type of GP, do a pretty good job of creating strong odds in your favor, and then get skunked time and again, IMO that randomization takes away from the strategy element of the game.

Certainly not a game breaker, and I just love this modmod and appreciate the work that goes into it, but whenever this happens in my games, it just makes me grit my teeth and say, well, something not nice. I'm sure we all have things in the game that cause this during game play, and this is likely at the top of my list.

Thanks for reading.

 

[Valkrionn]

I'm not going to rewrite the entire method of gaining Great People. I'm just not. It was designed random, so it's going to stay that way.

You say it enhances strategy when you can plan for a specific specialist. I say that it could too easily unbalance the game when you can choose your specialist, even when you have a higher chance of another.

As I said, without that option selected (and reloading repeatedly for the SAME specialist, to avoid other factors such as the exact odds varying from specialist pop to specialist pop) you may see what look like odd results. There is nothing that can be done about that.

 

[northeaster345]

I don't have that option selected.
Throwing random elements into a game are necessary to be sure, but I think there comes a point when you have to decide if this will be a game of strategy or just dice-rolling.

Yes, it can be extremely annoying to get that Great Physician which was only 10% chance.

But I've always just accepted that there's randomness in Civ.
If you want the worst, just look at combat, f***ing str 5 orcish archer killing Chalid!
It can be annoying to micro, but it also gives value to not just letting the computer play the game for you

 

[PPQ_Purple]

I always hate the randomness in Civ. It just felt out of place.
If I had the skill I would completely remove all random effects from the game.

This said, I would also remove any and all automation functions.
But that is a case of personal preference I guess.

 

[dunedainjedi]

Yes, it can be extremely annoying to get that Great Physician which was only 10% chance.

But I've always just accepted that there's randomness in Civ.
If you want the worst, just look at combat, f***ing str 5 orcish archer killing Chalid!
It can be annoying to micro, but it also gives value to not just letting the computer play the game for you

Valk has pretty much already stated this, but if you really have a huge issue with having a bad roll you can check new random seed and quicksave the turn before you generate a GP. If you get a healer you can reload. Hell you can even use this to prevent bad combat rolls. Is it a little annoying and a waste of time? Yes it is, but this option really only exists for people who don't like bad luck/rolls and want to win protect their heroes/super promoted units and get great sages/engineers/whatever instead of healers. Personally, when I am using it, I feel like I'm cheating and tend to reload every single bad roll which really lengthens the game, but the option exists. It's there. Use it.

 

[Sarisin]

Valk has pretty much already stated this, but if you really have a huge issue with having a bad roll you can check new random seed and quicksave the turn before you generate a GP. If you get a healer you can reload. Hell you can even use this to prevent bad combat rolls. Is it a little annoying and a waste of time? Yes it is, but this option really only exists for people who don't like bad luck/rolls and want to win protect their heroes/super promoted units and get great sages/engineers/whatever instead of healers. Personally, when I am using it, I feel like I'm cheating and tend to reload every single bad roll which really lengthens the game, but the option exists. It's there. Use it.

Again, respectfully, you are missing my point.

Why have a strategy element in a game such as specialization in a city to produce a certain type of GP when all you will get is based on luck?

Why not just take away the whole GPP system and just say OK this time you get a Healer, or this time you get a Great Engineer? Make it purely random. Actually, that's almost what it is now (however, I don't think you can get a certain type of GP if you are generating zero pts of that type).

I don't know, to me it is like taking out another strategic element of the game say deciding which type of unit to build and instead just randomly giving you units. In this case you are deciding to 'build' a certain type of GP in a city, but you are basically just randomly given one.

I agree that fooling around with that variable is cheating, and, frankly a big waste of time. Not that the AI doesn't cheat mind you.

In sum, V has said it would involve too much to make a change and I certainly accept that. If you have to choose between revamping everything or leaving it as is, it just wouldn't be worth it to make that change.

 

[Sarisin]

I always hate the randomness in Civ. It just felt out of place.
If I had the skill I would completely remove all random effects from the game.

This said, I would also remove any and all automation functions.
But that is a case of personal preference I guess.

I remember having this discussion with Kael and his FFH. One of the purest strategic games is chess. No combat odds, random Great People, random mutations or Haunted Land effects, or things like that. You won't see a Pawn fight back against a Queen capturing it and winning the day randomly.

As I said, some randomness is good (like starting position). I guess each player has his/her own way of looking at the amount of randomness that is acceptable in a game. Zero like chess, or more like RiFE.

 

[Calavente]

Valk.
IMO you misunderstood both my comments and Sarisin.
We are not adverse to randomness.
We are not asking for chosing the GP... but one can normaly be assured that a city specialized with priests (15-20 gpp) but with 2gpp bard and 2gpp scientist, should have AT LEAST 1 Prophet out of the next 3 GP. and not 1 bard, 1 scientist and 1bard.
That is what we are speaking about in terme of "strategy" and "planning for a great specialist"

BUT IMO, if you should have 10% of having a GH out of 5GP : getting 1 is almost normal, happenstance of 10% kicking in, getting 2 is bad luck, or coincidence, getting a third is very bad luck and not normal ; it devlops into a pattern, or ennemy action.
I understood perfectly the mechanism of the random seed, contrary to your opinion.

BUT, in 127 GP creation, out of 9 different games, spaced on so many turns (like 400turns minimal per game, so many actions and evolution of the seed), there should still be some link between the nature of the GP and the odds.
It's true that there could be some pattern appearing that is not totaly random.. as a 60% chance GP appearing only 40% of the time or as much as 80% of the time.
BUT I'm not saying that for the 127 GP, 60% odds should have given 60% of the GP...
I'm saying that even with the random seed not changing, the 60-70% odds should give more GP than the 15-30% odds.
I'm not asking for 50-80% of the GP vs 5-45% as one should normaly think .
I'm just even wondering if a 51/49 ratio even exists (ratio between [GP that should appear with 65% odds] and [GP that should appear with 20% odds]).
And in Sarisin case it doesn't. He had a 12/50 ratio instead if a 51/49... and not speaking of the ratio one could plan for : 60/20.

I can only compare with combat odds.
I am totaly in agreement with the odds.
a 90% chance combat odds for 1 unit means my units wins around 5-7 combats before dying (well it's normal... once it dies, it cannot even fight the next 3-5 combats in order to fullfil the 90%odds)
BUT on the whole, civilisation-wide I lose 1combat out of 10, or more precisely, 8-12 out of 100, when I have 90% odds.
meaning : on the length of 1 game (not even speaking of many games) the combat odds are accurate. And they use a "not so random seed".

Here is my ignorance kickin' in : If it works for combats odds, why can't it work for GP ?

EDIT : I appologize Sarisin, for speaking in your place. Maybe your don't agree with what I said... well in this case, just say it

 

[Andvare]

Here is my ignorance kickin' in : If it works for combats odds, why can't it work for GP ?

I would be very surprised if the RNG for combat is different from that for GP selection (it aint).

 

[Valkrionn]

Valk.
IMO you misunderstood both my comments and Sarisin.
We are not adverse to randomness.
We are not asking for chosing the GP... but one can normaly be assured that a city specialized with priests (15-20 gpp) but with 2gpp bard and 2gpp scientist, should have AT LEAST 1 Prophet out of the next 3 GP. and not 1 bard, 1 scientist and 1bard.
That is what we are speaking about in terme of "strategy" and "planning for a great specialist"

You have a string of bad luck, which is far more memorable than when you get the GP you want. In Sarisin's case, a particularly bad string, that is still possible, statistically, and will eventually even out.

I cannot 'assure' you'll get without rewriting the entire mechanic. It is simply not worth that amount of work, when the current mechanic works.

BUT IMO, if you should have 10% of having a GH out of 5GP : getting 1 is almost normal, happenstance of 10% kicking in, getting 2 is bad luck, or coincidence, getting a third is very bad luck and not normal ; it devlops into a pattern, or ennemy action.
I understood perfectly the mechanism of the random seed, contrary to your opinion.

If you understand it, then I should not have to repeat this again: With any computer based 'random number' system, you can get a string of similar results. Sarisin was notably unlucky. That is all.

A second test, by a third party, under controlled circumstances that should remove all odd results showed an even distribution, should have eliminated all argument. I do not see how you can still make that argument.

Do we know the odds for each of Sarisin's GP pops? No, we do not. The 'major' chances could have been as low as 60:40, or hell, 51:49. There is no way to gage how far off his odds were.

BUT, in 127 GP creation, out of 9 different games, spaced on so many turns (like 400turns minimal per game, so many actions and evolution of the seed), there should still be some link between the nature of the GP and the odds.
It's true that there could be some pattern appearing that is not totaly random.. as a 60% chance GP appearing only 40% of the time or as much as 80% of the time.
BUT I'm not saying that for the 127 GP, 60% odds should have given 60% of the GP...
I'm saying that even with the random seed not changing, the 60-70% odds should give more GP than the 15-30% odds.

Again: No, there does not necessarily need to be. He only stated the ratio of his GP. There was no statement about the odds. Without which, we cannot calculate how far off he was.

Which is why I'd said that using new random seed, and reloading the same GP with known odds often enough would yield results that followed the predicted curve. An experiment which Tokala carried out, and found to be true.

I'm not asking for 50-80% of the GP vs 5-45% as one should normaly think .
I'm just even wondering if a 51/49 ratio even exists (ratio between [GP that should appear with 65% odds] and [GP that should appear with 20% odds]).
And in Sarisin case it doesn't. He had a 12/50 ratio instead if a 51/49... and not speaking of the ratio one could plan for : 60/20.

You have the exact ratio it shows for each individual GP. He has no such 60/20 ratio! There is no ratio in the game for multiple GP!

I can only compare with combat odds.
I am totaly in agreement with the odds.
a 90% chance combat odds for 1 unit means my units wins around 5-7 combats before dying (well it's normal... once it dies, it cannot even fight the next 3-5 combats in order to fullfil the 90%odds)
BUT on the whole, civilisation-wide I lose 1combat out of 10, or more precisely, 8-12 out of 100, when I have 90% odds.
meaning : on the length of 1 game (not even speaking of many games) the combat odds are accurate. And they use a "not so random seed".

.....Combat odds use the exact same random number generator. Everything in Civ uses the same calls to the same random numbers.

You are being completely mislead by a false statistic. There is no real prediction able to be made about his numbers. BECAUSE WE DO NOT KNOW THE ODDS.

This is like keeping track of every battle your units are in, but only recording win/loss vs predicted result. It is very possible to have an anomalous spike, because of those 60:40 battles.

Here is my ignorance kickin' in : If it works for combats odds, why can't it work for GP ?

EDIT : I appologize Sarisin, for speaking in your place. Maybe your don't agree with what I said... well in this case, just say it

For the last time: It does work for GP.

The data being used in these claims is incomplete and misleading. An independent verification of the odds closely matched predicted results. There should not be anything more that has to be said, here.

 

[Calavente]

I prepared a looong answer. On the otherhand I decided it would end with blood-spilling. And I deemed the subject unworthy. So I deleted it.

So I'll only say :
I totaly disagree with you on the "mathematical" analysis, wether it is about your jugement on mine, sarisin, or even yours.
Your point about random seed is totaly accurate, but you also totaly missed my point.
100+ data points is considered a reasonnable statistical sampling.
I've worked with far smaller sampling for things that had a lot more financial weight, even more scientifical weight.
If you start by accepting a 3rd party measurement, and using it as a basis for refuting Sarisin's or my analysis, you are using 1 result that goes your way for refuting another result that doesn't go your way.
To refute Sarisin's analysis about 127 GP, one should not test once another 127 GP creation, but at least 5-10 times more. Else you have no more evidence that the test that goes your way is more representative than the one that doesn't go your way.

On the other hand I understand that the work is not worth the prize, especially as I'll never write a line of code about it.
So I won't pursue the matter any further.

 

[Fluesterwitz]

There is no way to gage how far off his odds were.

Maybe the problem is based on the %-display leads computer-illiterate ppl. like me to erroneously infer probabilities where non are implied. Would it be useful / possible / worth the effort to change just the display?

Somewhat OT, IMO the very idea of a "Great" person being plannable seems fancyful, I like to think about them as wildcards.

 

[Valkrionn]

......I'm not sure how much clearer I can make this.

Sarisin's data is not valid. Without knowing the odds of each individual GP pop, we cannot use his data to produce any kind of analysis on the odds.

It is not that I am ignoring his data in favor of another analysis. It is that his data is incomplete and unusable.


Once more: His sampling is invalid as it stands. We need to know the odds associated with each sampling to calculate how far off his results were. It is WELL within the bounds of probability if many of his odds were 60:40 or 70:30. This is why I said that it needs to be done on the same GP pop, so that we work with known odds. This was done, and reflected the curve.

If you would like to refute this result, conduct an experiment of your own, keeping with these requirements:

• Odds for each pop must be known. The random number only operates within each individual GP pop; Any over all comparison is invalid without this.
• Need a large sampling (at least 100, preferably more)

 

[Valkrionn]

Maybe the problem is based on the %-display leads computer-illiterate ppl. like me to erroneously infer probabilities where non are implied. Would it be useful / possible / worth the effort to change just the display?

Somewhat OT, IMO the very idea of a "Great" person being plannable seems fancyful, I like to think about them as wildcards.

What exactly do you mean? What probabilities are implied that aren't there?

My post was simply in reference to the fact that we do not know what his odds were. The displayed odds are accurate, but without knowing them, we can't know how unlikely any individual result is.