Great People Points explained

[hhanh00]

And what about this one:
You've build the pyramids, giving you 2 points towards great engineer.
But it will take 50 turns to get one that way.

After 20 turns, using caste system and running a food deficit you hire 6 specialist merchants, for three turns. You grab the money from the merchant specialists, finish the great person not in 50 but in 23 turns and still have a 92 % chance of getting a great engineer.

During 20 turns, you have 1 source towards GE. After that, you have 3 turns with 1 GE source and 6 GM sources. The total is 23 GE and 18 GM. So your chances are pretty much even.

I tried a similar experiment:
7 turns of 6 GM (specialists) + 1 GE (pyramid)
and then 4 turns of no specialist.

The final probabilities where 80% GM - 20% GE. It is explained by 7x6 = 42 GM and 7+4 = 11 GE. Whereas if only the type of sources mattered, you would get 7 GM / 11 GE.

The difference between counting in terms of GP points vs GP sources isn't that big a deal because the number of points brought by a source is fairly the same: 3 for a specialist and 2 for a wonder. And that precludes this tactic.

 

[hhanh00]

Btw, I also tried the 1 priest during 3 turns and then 3 scientists during 1 turn. I get 50/50. I thought it said it should lead to 75/25

 

[BrotherDragon]

I'm not so sure the basic premise of this thread is completely true. I've been able to alter the type of GP I get by changing what I'm researching ie. if I'm researching Philosophy I usually get a Prophet and if I'm researching Engineering I usually get an Engineer though this doesn't always work. In regards to which GP type is best I think that since I try to spread my religion I need a Prophet to build a shrine, I need Engineers to hurry early wonders, I need Scientists to build academies and Merchants get sent to far away cities for LOTS of trade gold. In other words a variety of GPs seems best.

 

[fortytwo]

Nice to know, especially the odds of getting a GP. I always wondered why my Prophet cities had so high of a percentage on Engineers or Scientists .

Req

EXACTLY. I've been looking all over for this answer. I would found 4 religions and NEVER get a great prophet. I thought it was b/c I did not build a prophet-building wonder. But no, even with a prophet wonder, still, whole game, zero prophets. I just had no reason to build priests until now. Awesome article.

 

[Ronnie1]

A minor hijack of this thread, (sorry). Was wondering about the use of Great People. When I first started playing Civ IV, I always saved them to start golden ages, (left over strategy from Civ III I think). But I think this is not the best use for GP's, especially early in the game. Are super specialists a better use early on because of the cummulative effects over time?

 

[Zombie69]

Wrong thread. Try this one, it has everything you want to know :

http://forums.civfanatics.com/showthread.php?t=169059

 

[MestreLion]

As first glance this formula for the odds seemed quite weird for me... but, thinking about it, i quite like it. Because it averages odds per turn, it means i dont have to worry if, for a few turns, i use a lot of engineers to speed up a building when i want to get a Great Priest or something.Those engineers will only contribute with 100% * few turns, not 100% * (lots of engs * 3) * few turns. It means it wont mess too much with whatever i want to generate, as long as i use lots of specialists in fewer turns.

Want to get a Great Priest? Then using a single Priest for as long as you can will give you very high odds of getting one, no matter if once in a while you used a lot of Merchants for gold or lots of Engs for hammers. The single Priest may have very lil odds on those turns, but the 100% for the majority of turns will almost guarantee you will get what you want.

Cool

PS: Is that correct, by the way? Are there any more consequences on the turn-average system i should be aweare of?

 

[jesusin]

Awesome!
I didn't know about the turn averaging. It has important consequences.

Suppose you have a wonder, an excess of 2 food and want to hire a certain type specialist. If you like the GP it would give, use 1 specialist every turn. If you don´t like the GP it would give (you prefer the one form the wonder) it is better to have no specialist for one turn (accumulating food) and 2 specialists for one turn (starving the city). Even best no specialist for 2 turns and 2 specialists for 1 turn, etc...

 

[pixiejmcc]

Want to get a Great Priest? Then using a single Priest for as long as you can will give you very high odds of getting one, no matter if once in a while you used a lot of Merchants for gold or lots of Engs for hammers. The single Priest may have very lil odds on those turns, but the 100% for the majority of turns will almost guarantee you will get what you want.

Really good, useful point MestreLion.

 

[VoiceOfUnreason]

Really good, useful point MestreLion.

Alas, an inaccurate one.

The relevant section of the SDK is CvCity::doGreatPeople(), in CvCity.cpp.

Every turn, two functions are called. CvCity::changeGreatPeopleProgress() is bumping up the GPP counter by CvCity::getGreatPeopleRate(), which is exactly what it sounds like: how many GPP per turn are we generating.

The code then loops through every UnitType, calling CvCity::changeGreatPeopleUnitProgress, and bumping up that value by CvCity::getGreatPeopleUnitRate(). The GreatPeopleUnit rate is being changed by processSpecialist and processBuilding, and in each case the input is the number of specialists/buildings being added or removed from the city.

In other words, if you run 9 scientists this turn, then you increase GreatPeopleUnitProgress( GREAT_SCIENTIST ) by 9.

That means that it doesn't matter how you distribute the specialists - one artist run for 10 turns has the same effect as 10 artists run for one turn....

Except in one bizarre case, which is potentially exploitable, although I've never managed to make it large enough to be notable. It has to do with overflow - the GP points in excess of those used to produce the great person are saved, but the memory of how those points were accumulated is lost when the GP is created.

So imagine that on turn one you run a zillion scientists. So you pop a Great Scientist, but you still have a zillion uncolored points left over. So fire all the scientists, and run an engineer for one turn. Now you've got a Great Engineer, and you still have a zillion uncolored points left over. Run another engineer for one turn. Pop pop pop.

In practice, you can't get that much overflow. But you can shave a few turns off easily enough.

 

[jesusin]

Hi VoiceOfUnreason

You seem to understand the code. Please, would you be so kind as to confirm or reject my 2nd trick in
http://forums.civfanatics.com/showpost.php?p=4567722&postcount=60

Basically, I am guessing that the code changeGreatPeopleProgress() is run, then a GP is generated if the total has reached the threshold. If not, then getGreatPeopleUnitRate() is used to calculate the new probability.

My trick is useless if the code changeGreatPeopleProgress() is run, then getGreatPeopleUnitRate() is used to calculate the new probability, then the threshold is checked to generate a new GP.

Thank you very much.

 

[VoiceOfUnreason]

Bad news, jesusin: both counters are updated before the check against the threshold.


Hey, I just found another useless optimization. GP progress is checked city by city. So it's theoretically possible to get two GP to spawn during the same turn, if the second city is earning more than 100 gpp/turn (urk, or 200 gpp/turn after you've spawned 9 people).

So given equivalent cities, you should put the GP farm in the newer one.

I don't expect to be thanked until someone actually manages it in a game though - currently scheduled for the next full moon in a week with two Mondays.

 

[Xin Yu]

100 GPP per turn only takes 1 great wonder + 1 small wonder + 7 specialists + national epic, and modifiers of philosophy leader, national epic, pacifism. 7 specialists only requires 3 corns = size 10 city. If you have more wonders then the number of specialists can be reduced further.

Not awfully difficult, at least from the face value.

 

[jesusin]

@VoiceOfUnreason : Bad news, jesusin

Well, I consider it good news indeed. The program works as it should. My trick is no good but, who cares.
Thank you very much for analyzing it for me.

@Xin Yu: Not awfully difficult

I thought so at first, too. The main difficulty for me is that my GPfarm tends to be one of my oldest cities, so it is the no-farm-city the one that has to get those 100GPP.

 

[MestreLion]

Alas, an inaccurate one.
(...)
That means that it doesn't matter how you distribute the specialists - one artist run for 10 turns has the same effect as 10 artists run for one turn....

Not according to the OP:

* Assuming normal game speed, no great person generated yet: You hire a priest specialist for 22 turns. Then you fire him, and hire two scientist specialists for 11 turns. Intuitively, one would assume that because you had 22*1=22 sources for a great prophet and 11*2=22 sources for a great scientist, this would give you a 50%/50% chance. But because the game averages over turns, the odds will be 22/33 = 67% for a great prophet and 11/33=33% for a great scientist instead, because you had 22 turns with only great prophet sources and 11 turns with only great scientist sources.
(Thanks to Roland Johansen for suggesting this!)

If what you are saying is correct, then pretty much everything that was said before about odds in this thread is wrong. Both theories are reasonable and have their own logic. As i cant undestand SDK code, i have no clue now. Could you please double check this with Kylearan or anyone else? Can someone (who knows SDK code) please check this (great) example of "1 artist for 10 turns VS 10 artists for turn"? Would both produce the same odds against, say, 1 eng for 10 turns?

 

[Kylearan]

Hi,

If what you are saying is correct, then pretty much everything that was said before about odds in this thread is wrong. Both theories are reasonable and have their own logic.

I haven't followed what has been said in this thread, but want to point out that when in doubt, trust the code readers. My original post was based on trying to reverse engineer the mechanics of how GPPs work, by using experiments and observation. This method is quite error-prone by nature. If the code readers say it's working differently, then I'd trust them more than my previous work (which was based on an older patch too, by the way).

And no, I won't take a look at the code myself. I'm way too lazy for that.

-Kylearan

 

[Mathias]

Once a wonder becomes obsolete, it will no longer produce any GPPs.

Negative. Obsolete wonders continue to produce both culture and GPPs.

 

[CivDude86]

To be fair though that change was made in 1.61.

 

[VoiceOfUnreason]

Hey, I just found another useless optimization.

T-Hawk discovered a useful exploit. During Anarchy, you don't actually need to feed your citizens - you can make everybody a specialist.

So what? Specialists don't generate any GP points during Anarchy, so its a waste of time... EXCEPT that the specialists do generate source points during Anarchy.

So if you need to dilute the effect of your Artist wonders, during anarchy take all of your citizens off of the tiles, and use them as Science specialists (or whatever sort of GP you would prefer).

Exactly the same trick works during disorder when you capture a city.

 

[VoiceOfUnreason]

My original post was based on trying to reverse engineer the mechanics of how GPPs work, by using experiments and observation. This method is quite error-prone by nature. If the code readers say it's working differently, then I'd trust them more than my previous work (which was based on an older patch too, by the way).

The version of this article that's in the War Academy should probably be updated by somebody.

And no, I won't take a look at the code myself. I'm way too lazy for that.

Somebody else, perhaps