Granary or settler?

ThinkTank

RL Addict
Joined
Feb 2, 2007
Messages
1,085
Location
An inaccessible possible world
This thread continues a discussion that was started in the thread Favourite wonder in posts 43, 45, 46 and 47.

The issue, at least to me, is to understand the role of granaries in efficient REX. In REX you want as many cities as possible as soon as possible. Do you need granaries for that? And if the answers is "it depends", what does it depend on?

Now here is an answer to that question:

To decide if a granary is useful, determine what increases growth (measured in surplus food per turn, or sfpt) the most: a granary, or a settler?
A settler increases growth by the amount of surplus food the next city-site can grant.
The granary doubles growth of the city you are building it.

If the current city produces 5 sfpt, a granary is worth 5 sfpt! And since it costs 60 shields, that is 12 shields per sfpt.
If the current city produces 4 sfpt, a granary is worth 15 shields per sfpt.
If the current city produces 3 sfpt, a granary is worth 20 shields per sfpt.
If the current city produces 2 sfpt, a granary is worth 30 shields per sfpt.

If the future city-site grants 2 sfpt, a settler is worth 15 shields per sfpt.
If the future city-site grants 3 sfpt, a settler is worth 10 shields per sfpt.
If the future city-site grants 4 sfpt, a settler is worth 7.5 shields per sfpt.
If the future city-site grants 5 sfpt, a settler is worth 6 shields per sfpt.

So as you can see if the future city can output 2 sfpt, a granary in the current city is only worth it if the current city already has 5 sfpt, and is almost equal if its has 4 sfpt. (the granary will also cost maintenance, and the city will also add shield and commerce output, in addition to unit support, with 4 sfpt a settler is still a better choice.)
If future city sites have more than 2 sfpt, even a 5 sfpt doesn't justify a granary, and a settler should be build instead, until all hight food sites are settled.

I disagree with these criteria.

I have to disagree. The problem with this argument is that it considers the choice between a settler and a granary as a one-off decision.
It depends on the circumstances if that is right.

At the start of REX, in a food poor starting location (only 2 sfpt locations),
but with a lot of room to expand to (say large or huge map), a granary first will shorten the production cycle of many settlers to come,
even if it will delay the production of the first or first few settlers. So the real value of the granary (at least for REX) must depend on the number of settlers that will be built in the city.

At the end of REX, in a city that will maybe only build 1 settler before expansion is over, the considerations by MAS are valid.

In general I think the right perspective is something like the following. A granary first will always pay of in the long run. It must be able to calculate how long that run is, that is, how many settlers that is; the answer will likely vary with shields and food production. Suppose that n settlers is the turning point. Then if you think REX is going to last longer than n settlers build a granary first, otherwise don't build one.

In order to give an exampe of what I mean, consider a map that only has 2spft city locations, we are at turn 0 of the game, so at the start of REX, so we only have the capitol. Should we build a granary first, or start building settlers right away? According to the criteria described by MAS, you should not build a granary.
According to me, it depends on the number of cities you will be able to grab in REX. The reason I think this is the case in general is as follows.

Mechanism 1: building a settler first has the advantage that you get your second city sooner, as well as everything that is produced by the second city. This city can also build setllers, making you grow faster. This mechanism reinforces itself for what is produced by any city founded by a settler produced by the second city, since these are also sooner, and so on.

Mechanism 2: building a granary shortens the production time for settlers.

I think that mechanism 1 gives you a better initial growth, but that mechanism 2 compensates for not having mechanism 1 in the long run. By which I mean that there is a break even number x such that if the number of settlers you need becomes larger than x, mechanism 2 will get you the settlers faster.

I will illustrate this in the scenario sketched. I am assuming a map with only BGs. Only 2sfpt cities ever. This means we can do 2 growth in 20 without a granary, 2 growth in 10 with a granary. I also assume that we can improve tiles fast enough that we can do 3s(shields) at size 1, 5s at size 2, 7s at size 3, and so on. I will compare two strategies that only differ in the first city building a granary or not. All subsequent cities founded will be building settlers (without a granary). I will assume that a settler takes 5 turns before it founds a city.

Strategy 1: no granary ever. The earliest settler that can be produced is at turn 20 (growth to three). After that, a settler every 20 turns. All cities founded later follow this strategy.

Strategy 2: granary first in first city. All later cities follow strategy 1.
Since we cannot produce 90 shields in the first 20 turns I assume that in the first city we produce the first settler at turn 30 (growth to 4), then a settler every 10 turns.

In the attached Excel sheets I have calculated all settlers produced and towns founded and settlers produced by towns founded and so on for the first 100 tuns or so:



After 105 turns we have 17 cities with 3 settlers on the move in the no granary strategy; 23 cities and 4 settlers on the move in the granary strategy. The break even point appears to be at 7-8 cities.

You can find the complete Excel sheet here:
 

Attachments

  • granary_2sfpt.zip
    8.5 KB · Views: 49
Strategy 1: no granary ever. The earliest settler that can be produced is at turn 20 (growth to three). After that, a settler every 20 turns. All cities founded later follow this strategy.

Strategy 2: granary first in first city. All later cities follow strategy 1.
Since we cannot produce 90 shields in the first 20 turns I assume that in the first city we produce the first settler at turn 30 (growth to 4), then a settler every 10 turns.

There is a bit of a hole in your argument - if you only have 2 sfpt locations, then you will be building much more than just settlers. Assume your city has 5 spt. Strategy 1 gets a settler plus 70 spare shields every 20 turns. Strategy 2 gets 2 settlers plus 40 spare shields every 20 turns. I recognize that cities are the most important thing in the early game, but obviously you will do something with the extra 30 shields - most likely build military to capture cities :evil:.

I'm not really disagreeing with you - I find granaries to be very useful. In your example of only 2 spft locations, I would typically build 1 settler on growth to size 3, then a granary mainly because you will have way too many shields before your population grows back to be able to build another settler, and a granary is usually the single most effective thing you can build with those shields at that point (although that's somewhat dependent on level - on the lower levels, I might be content with 2 cities and an archer rush).
 
I agree that you should account for shields produced and not spent on settlers; in the details of the spreadsheet you can find just that (sheet 2).
Summarizing: in the 20 turn no granary cycle you have 54 shields to spend on other builds than the settlers; in the 10 turn granary cycle you have 30 shields to spend on other builds. That will account for military. Also both scenarios have room for building a couple of warriors at the very start, for exploration for example, which is realistic I feel.

[Edit] My number of shields differs from yours. The reason for that is that the granary first scenario naturally ends up in a size 2 to size 4 cycle; the no granary scenario natuarlly ends up in a size 1 to size 3 cycle. That is why my shield numbers for the granary scenario are better than yours. You can certainly MM the no granary scenario into a size 2 to size 4 cycle too; but this will add a 10 turn delay to the settler production sequence.
[/EDIT]

I am not in anyway committed to the 2sfpt assumption; I just wanted to make any assumption that enables a real calculation. We should look at other assumptions on growth potential but I expect that this will change the numbers (growth rate and break even point), but not the general tendency: there will still be a break even point after which the granary first strategy is winning.
 
Thank you, I like the research you've done. At what point then does the break even point become so far that it's negligible?

It's probably so high that no granery is just as effective on standard and smaller maps with 5spft, but on larger maps then the granery is even more effective, possibly.
 
Thank you, I like the research you've done. At what point then does the break even point become so far that it's negligible?

It's probably so high that no granery is just as effective on standard and smaller maps with 5spft, but on larger maps then the granery is even more effective, possibly.

We will need to add calculations based on different assumptions on the surplus food per turn situations. That may indeed lead to the conclusions you suggest. Certainly a large open space to expand into is favourable for the granary scenario; and a small space favours the no granary scenario.
I think this is not surprising since a lot of players would share that intuition. But putting numbers on how large the open space must be can be useful I feel.
 
Granary vs. settler is kind of like those investment brochures where they always have the example of a younger person who saves $2,000 per year for 10 years and then stops vs. a person 10 years older who saves $2,500 per year and never stops. In the investment brochures, the younger person always ends up with more money, but there are 3 possible outcomes to this type of situation depending on the parameters you set: the interest assumption, the dollar amounts saved per year, the difference in their ages, and the time horizon:

A) The older person catches and passes the younger person before they both reach retirement age.
B) The older person would eventually pass the younger person if they lived forever, but they would certainly both be dead before it happens.
C) The older person never catches up to the younger person.

Granary first is the equivalent of delaying starting to save, but then saving more money per year later. Not only it is possible that the granary will catch up too late to be relevant, it is also possible that the granary will never catch up – in particular a high food and low shield start.

As usual the answer of which to build first is it depends. It depends on 4 things:

1. Food available for other cities (interest assumption).
2. Food in the capital (dollars saved per year).
3. Shields in the capital (age difference – i.e. how long before the granary completes).
4. Map parameters (time horizon – how many total settlers you are building).

MAS’ theory (which looks like it may be a plagiarized version of WackenOpenAir’s theory :mischief:) looks at #1 and #2 primarily.

I think SirPleb said that he would build a settler first if the second town would have more sfpt than the capital – also using #1 and #2, but drawing the line in a different place.

I think it was DaveMcW who said that food is not the most important factor, but how many BGs you have is – if you have 2 BGs, then granary first; if you have one BG, then 1 settler before the granary; if you have no BGs, then don’t build a granary at all - he emphasized parameter #3.

ThinkTank seems to be emphasizing parameter #4 primarily.

So everyone’s right :), but I think it will be very difficult to come up with a general rule that incorporates all 4 parameters.
 
We had actually had a similar thread recently. You can find it here.

After that thread's discussion, I felt that the overall goal of REXing was being overly simplified. The goal cannot just be to simply fill in as much land as fast as possible with tiny settlements.

Such an empire forgets that corruption effects make cities settled outside of a "core radius" too corrupt to contribute shields toward the empire. Certainly overly-corrupt cities contribute commerce/science if you set them up as specialist farms. Plus, they can claim strategic and lux resources, deny the AI settling space, and a host of other uses. Definitely a bigger empire in Civ is a better empire. BUT, again these overly-corrupt cities contribute next to nothing in regards to shields.

This means the ideal empire is a large, productive core with numerous amounts of farms. This a well-advertised fact :p and not rocket science. IMHO though, strengthening the core by building infrastructure starts before all the available non-corrupt land has been been filled i.e. before the main REX phase ends. That mean I place greater emphasis on specializing my cities, such as setting up settler factories earlier on. Read on to find out why...

Just to complicate the picture Chamnix paints a bit further (good summary, btw :thumbsup:), I see at least a couple more variables in play.

-Commerce generated by citizens in core cities
-Extra shields not needed for settler production generated by citizens in core areas

For this example, assume the city is not on a crazy high food spot like 3 FP wheat tiles. ;) -->

Say the city has these specs:
- 5fpt
- 8spt on average

This city makes a natural 4-turn settler factory. It can gain 1 pop point every 2 turns with a granary, and can easily make a 30 shield settler every 4 turns. Without a granary, the city gains 1 pop point every 4 turns.

If we set up a settler factory with a granary in this core city early on (who would set one up in a corrupt city anyway? :rolleyes: ), then that city will have a higher average population over the same amount of turns than a city without a granary.

For example, we might run this city from sizes 4-6. That means the population never drops below 4 and gets back to size 6 in just four turns. On the other hand, say we have no granary. The city builds a settler when it hits size 6 (dropping it to size 4). It will take twice as long to recover those pop points, a full eight turns.

Remember that if you've roaded all the worked tiles for that city (and you should!), then those citizens are each making commerce every turn. I'm not going to attempt to quantify it because their are too many different possibilities, but it's safe to say the city with a granary makes a significant amount of commerce more than the granary-less city. This is simply due to the population recovering in size more quickly and then putting those extra citizens to work.

The same thing works with shields. The city without the granary will make less shields in a settler-building cycle because the average population will be lower. Often times these extra shields can be put to good use building units in between settler builds (especially when the city is not a 4-turn factory).

Lastly, if you set up settler factory cities, you allow other cities to specialize in building units. Specialization is truly the mark of an advanced civilization. :p

So, my conclusion is that granaries have more importance than simply allowing or not allowing a certain amount of settlers to be built in a certain time frame. They also give the empire more uncorrupted shields and commerce by keeping more citizens at work, and they allow specialization of cities. I agree that getting the core cities built quicker can often be a better choice than building the granary right away but I'm betting that the number of core cities where it really is better are fewer than we might think.
 
Good input guys. My strategy for this depends totally on the landscape. But basically I will always build 1 settler before a granary in my capitol. Sometimes my capitol does not even get a granary in which case it is the second city that builds a granary (sometimes both). If I discover any AI that are way too close to me I switch any granary in progress to a barracks and concentrate on whatever offensive units I can muster in a short period of time. Warriors have proven to be excellent "Settlers" in the past.
 
Good input guys. My strategy for this depends totally on the landscape. But basically I will always build 1 settler before a granary in my capitol. Sometimes my capitol does not even get a granary in which case it is the second city that builds a granary (sometimes both). If I discover any AI that are way too close to me I switch any granary in progress to a barracks and concentrate on whatever offensive units I can muster in a short period of time. Warriors have proven to be excellent "Settlers" in the past.

I somewhat agree, as most of my line of play have centred around having few but highly productive cities for intense military action that would, among other things, make room for my following expansion.

The research and analysis in the previous posts have been insightful and educational to me, at the very least. Thank you :)
 
Granary vs. settler is kind of like those investment brochures where they always have the example of a younger person who saves $2,000 per year for 10 years and then stops vs. a person 10 years older who saves $2,500 per year and never stops. In the investment brochures, the younger person always ends up with more money, but there are 3 possible outcomes to this type of situation depending on the parameters you set: the interest assumption, the dollar amounts saved per year, the difference in their ages, and the time horizon:

A) The older person catches and passes the younger person before they both reach retirement age.
B) The older person would eventually pass the younger person if they lived forever, but they would certainly both be dead before it happens.
C) The older person never catches up to the younger person.

Granary first is the equivalent of delaying starting to save, but then saving more money per year later. Not only it is possible that the granary will catch up too late to be relevant, it is also possible that the granary will never catch up – in particular a high food and low shield start.

As usual the answer of which to build first is it depends. It depends on 4 things:

1. Food available for other cities (interest assumption).
2. Food in the capital (dollars saved per year).
3. Shields in the capital (age difference – i.e. how long before the granary completes).
4. Map parameters (time horizon – how many total settlers you are building).

MAS’ theory (which looks like it may be a plagiarized version of WackenOpenAir’s theory :mischief:) looks at #1 and #2 primarily.

I think SirPleb said that he would build a settler first if the second town would have more sfpt than the capital – also using #1 and #2, but drawing the line in a different place.

I think it was DaveMcW who said that food is not the most important factor, but how many BGs you have is – if you have 2 BGs, then granary first; if you have one BG, then 1 settler before the granary; if you have no BGs, then don’t build a granary at all - he emphasized parameter #3.

ThinkTank seems to be emphasizing parameter #4 primarily.

So everyone’s right :), but I think it will be very difficult to come up with a general rule that incorporates all 4 parameters.

For me it has more to do with the ratio of shields to food - I will build a granery if I am shield heavy relative to food, my normal ratio is if I have 2 more shields than food at pop 4 I will almost certainly build a granery If I have more food than shields then I won't. The rest falls into the "it depends" area
 
Thank you all for your comments; there are a lot of good arguments there.

So everyone’s right :), but I think it will be very difficult to come up with a general rule that incorporates all 4 parameters.

Difficult, but probably not impossible. A good strategy in these kind of cases is to first try to understand the factors in isolation, and only then start thinking about the combined mechanism. I plan to continue this research, but my first steps will be to write a bit of software that enables a quick calculation of scenarios because doing that all manually takes too much time. I will post that software and any subsequent findings here for those of you who are interested.

On a different note, the useful analogy that Chamnix makes with investment makes me believe that there are peole out there who analyze these sort of mechanisms all the time. What we need is a student in, say, economics, operations research, or mathematics or some such subject, who likes civ and is in terrible need of a good master thesis subject. Maybe we should rename this thread to "Free master thesis subject"?

But basically I will always build 1 settler before a granary in my capitol. Sometimes my capitol does not even get a granary in which case it is the second city that builds a granary (sometimes both).

That is very similar to the normal strategy I usually take: settler first, then granary in whatever has the best food of my first two cities, barracks in the other. I feel this gives me flexibility, but I also think that this strategy is not optimal in terms of REX.
 
I've read this thread yesterday and I've been working on my own calculations, but I'll give a short reply now. (cause its taking to long)

There is no denying math, sure, in this scenario, its better to build a granary by turn 105: 30% more cities. But there are 2 big assumptions that normally do not apply in a random epic game:
#1 you know beforehand that you have room to expand. In a random epic game, you do not know this until after exploring.
#2 5 turn to settle means you are using a xxCxx|xxCxx pattern. I usually don't.

But I see you already stated that yourself in post #12, by saying you prefer to build settler-->granary for flexibility.

I'll post the result of my own calculations in a while...
 
#1 you know beforehand that you have room to expand. In a random epic game, you do not know this until after exploring.

I agree. But the theoretically optimal strategy assumes we know the map, even if in practical play we don't. This is exactly what makes exploration so important; as well as finding ways of postponing critical choices until sufficient information is in. The scenario I calculated assumed the initial build of a couple of warriors for exploration.

But your general point is valid: exploration may not provide sufficient information early enough for making the correct decision. In the two scenarios I calculated the break even point is 7-8 cities and we could create MM scenarios in such a way that at turn 19 we need to decide to finish the settler build or change to a granary. I think it likely that in 19 turns we can decide if we have at least 7-8 city spots. But this may fall out differently for other scenarios.

#2 5 turn to settle means you are using a xxCxx|xxCxx pattern. I usually don't.

Neither do I. This was not what I was thinking. I wanted to make an assumption that was fairly realistic but also simplifies the math. I would expect early settlers to take less turns to found a town (say 3-4), but later settlers to take longer, certainly later settlers founded by central core towns(even if supported by roads by then). 5 turns looked like a reasonable average. But we can vary this and see what the differences are.
 
Nice discussion, for some reason everytime I see the name DaveMcW or SirPleib I think of the 'Randolph Scott' scene in 'Blazing Saddles', as if we should all bow our heads and remove our hats.

But back on topic, my primary consideration for a granary should be what the city is going to be used for. If I see either a worker or settler factory potential (a food bonus + a couple of BG) then it'll be a settler then a factory. If there's a lot of food but few shields (a FP wheat and few BG) then settlers without a granary until a factory is set up elsewhere then the city will turn to popping out workers. If there's lots of shields but limited food (2sfpt) then units while getting big enough for a settler then barracks and units.
 
There is a point #3 I forgot to mention: Workers.

I can see myself getting 4 settlers before I start on my 2nd worker (starting worker being 1st) but not 17 towns.

(Truth be told, I never did bother to calculate whether it is better to first build cities or build workers alongside. My gut said I needed workers earlier)
 
Geez, MAS why you have to complicate things with workers! :lol: j/k

I agree workers are "mission critical". Every turn an unimproved tile is worked you are missing out on gold, sheilds, and/or food. I believe there is a balance between expansion, infrastructure, military units, and workers. I do not use spreadsheets or that much math to find the balance, I just "fly by the seat of my pants" so to speak. This could be why I do not play higher than Emp.

Edit: added thought

1) I see some guys are arguing for the theoretical max expansion -- it is important to know this.
2) Others are arguing for the applied max expansion. (applied = in game, don't know map, AI is out there)
I believe it is important to understand the first in order to apply it to the second.
 
It's a lot easier to find a spot that can generate the 5spt & 5sfpt that is necessary for a 2-turn settler factory. After Despotism a pair of irrigated FP and a pair of mined hills are all you need.

In the early part of most games I play, my priorities are exploration, gettting a second city (one with a extended plan) and developing my local area. To do so the early argument often is after the first explorer (scout or warrior) is it going to be a worker or a settler?
 
Lastly, if you set up settler factory cities, you allow other cities to specialize in building units. Specialization is truly the mark of an advanced civilization.

I really think this is the crux of the argument. It should be obvious that building a granary first will be theoretically better in most cases. The problem is reconciling the theory with practice. The real value of a granary in a suitable settler-pump location is that it frees up other cities to produce the things they produce best. Have a bunch of hils with enough food to support them? Build a rax and crank out military. Have a location with a small food bonus, but not enough to cut it as a settler pump? Workers. High-commerce cities (say on a river but also with gems or gold)? Build libraries or markets. Highly corrupt cities? Build artillery or just wealth. The real key is to make sure each city does what the terrain suits it to and thereby minimize waste.

That same thought process underlies the emphasis workers get in these forums. Every turn a worked tile is unimproved, or every unit turn spent hiking overland instead of cruising along a road, is a turn in which your empire has wasted some of its potential. The higher the difficulty, the less of this waste you can have and still have a shot at a win.

Setting up a settler pump is not about achieving a theoretical maximum on an ideal infinite map. It is about maximizing the resources at your disposal. On some maps, that may mean not having any granaries and building up your military to settle by conquest, instead. On other maps, it may mean you build an early ganary or two and get 15 quick cities with barely enough units to defend them.

It is important to know that the early granary theoretically pays big dividends. It is equally important to recognize how your particular game start differs from this theoretical ideal.
 
Top Bottom