Don't Waste Worker Turns

[MarigoldRan]

Oh, you know one reason I hate automated workers?

Building forts.

Instead of mining that hill or farming that grassland, you could have like 10 workers move across your empire to build a fort on a bronze tile within your cultural boundaries, even though you already have 3 or 4 bronzes. Ugh.

 

[JammerUno]

Just queue worker commands by holding shift.

You can have them running around a city chopping and putting down the improvements you want in the order your want them in seconds, and then you can forget about them until they're done. It saves time in playing and helps you think about your improvement order and timings to boot.

 

[Nick Carpathia]

Does anybody have an opinion on how valuable a settler turn is? I'm thinking specifically about cases where I road to my target city spot, including the spot itself if it'll save a turn. My main reason is that, if barbs come, I'd rather put the worker in danger. Kinda dumb though; if I save time, I'll put less stuff into danger altogether, so it's better to chop.

That's a damn good question. My thoughts are... if you're going to be culture-restricted by a monument chop or a missionary to catch up to the city (to pop the borders to unlock the food improvement) then wait. If you're getting free culture from CRE, then settle immediately. If you're settling food in first ring, then settle immediately. I like to spawnbust around the cities I'm settling anyway, so barbs aren't a biggie. Plus, if your worker is spending its first move walking onto the proposed city tile, and it spots a barb in range, then it can always run away.

How are these thoughts?

 

[krikav]

If the city can't do anything productive yet, I often have the settler on sleep.
Only when a worker is ready to improve the new citys tiles, or do a monument chop or something, I settle.

Creative is different (as mentioned), since you start to build culture right away.

 

[vicawoo]

Depends when you plan on comparing them.

 

[ben-jammin]

If the city can't do anything productive yet, I often have the settler on sleep.
Only when a worker is ready to improve the new citys tiles, or do a monument chop or something, I settle.

Creative is different (as mentioned), since you start to build culture right away.

There are of course situations where a city is worth settling without any workers ready to immediately improve. Financial coast, oasis, and floodplains will at least get you growing.

 

[civvver]

This may have been mentioned but don't command workers in stacks. What I mean is, if you need an improvement completed quickly and want to use multiple workers that is fine by command each one individually. If you group select say four workers and have them build something and it finishes in not exactly the amount of right worker turns the extra workers will sit idle. It's kind of like when you group select units of different movement points and have them move somewhere- they only move as fast as the slowest guy. So if an improvement takes 6 worker turns and you selected a group of 4 workers and have them improve the tile, on the second turn 2 workers will complete it and the other two will just wait there.

I used to do this a lot when building railroads across my empire til I realized what was going on. Now I just queue individual workers up to complete long railroads and they often work on the same tile, but the won't waste turns doing so.

 

[krikav]

This may have been mentioned but don't command workers in stacks. What I mean is, if you need an improvement completed quickly and want to use multiple workers that is fine by command each one individually. If you group select say four workers and have them build something and it finishes in not exactly the amount of right worker turns the extra workers will sit idle. It's kind of like when you group select units of different movement points and have them move somewhere- they only move as fast as the slowest guy. So if an improvement takes 6 worker turns and you selected a group of 4 workers and have them improve the tile, on the second turn 2 workers will complete it and the other two will just wait there.

I used to do this a lot when building railroads across my empire til I realized what was going on. Now I just queue individual workers up to complete long railroads and they often work on the same tile, but the won't waste turns doing so.

This have not been mentioned I think.
Good to point out, I overlooked this for a long time when I started playing.

 

[JustBen]

Rivers and coast can be used for trade routes before Sailing as long as they lie within your cultural borders. It's a poorly documented feature that is only obvious if you pay close attention to trade routes early game. You can freely switch between river and road without a city as a junction.

Rivers that lie on your border require testing that I haven't done in quite a while. I think you need a road on your side to establish trade routes to the other. Corners might behave differently than edges. Just be careful if you're microing early trade routes along a riverine border.

Edit: The example I'm thinking of is when you have a river connecting your first and second cities, but they are far enough apart that their cultural borders to do not touch. If you build a road along that part of the river in neutral territory, you will need to road 2 more tiles to complete the trade route. The river inside your border does not have access to the road outside your border, until you add a road inside too.

 

[vicawoo]

So if you settle a city n turns early, you could look at it as:
-you get those n turns of extra production
-or say T turns down the road, you get an extra n turns of max production (the anarchy argument)
To better explain that, it's kind of like the way you learn to sum arithmetic series in school, that is, comparing the offset:
0 0 0 1 2 3 4 5 5 5 5 5
1 2 3 4 5 5 5 5 5 5 5 5
So in this example 3 extra turns gives us 3 extra turns of a max 5 production.

The problems with the latter argument, let's say our late game city generates 100 hammers per turn, but this doesn't feel like the cost early.
Second, lets say our n turn early city also produces another settler n turns early, which later produces another late game 100 hammer city, and that our early city and the new also produce more settlers, etc. Potentially we could say the cumulative effect gets out of control.

The only way I could figure out how to do this was to use the first method, or to prorate the increase over the number of turns we're looking ahead.

The first argument is easier, so settling early generates n turns of 2 food, 1 hammer, 1 commerce, - something maintenance, and an extra tile yield. From that standpoint, chopping a forest (20 hammers) to get a settler 2 turns faster isn't that great (although the city could have produced something else in those 2 turns, which gets back into the second argument).

 

[Um the Muse]

Thanks for laying it out like that, vicawoo. Two comments on your analysis: the offset gets "reset" so to speak when you hit the happy cap and only applies at the margin anyway (i.e. if settling earlier gets you the Nth citizen earlier than what you actually did, the implied costs are what that citizen would've given you).

The other comment is that when you have a recurring but shrinking benefit like what you described, you can approximate its value by using a geometric series. I'm specifically talking about the power series related to the time-value formula under economics. I think that the ratio in the denominator would be the production the earlier settler will give you over the total amount of production used to produce the next settler.

So, three things: marginal benefits, Time-Value of a settler and the Time-Value of a worker. Four things: MB, TV of a city, the TV of city development and NOBODY EXPECTS THE SPANISH INQUISITION!

 

[VoiceOfUnreason]

The only way I could figure out how to do this was to use the first method, or to prorate the increase over the number of turns we're looking ahead.

I think what you're really looking for here some flavor of present value.

The value of settling a turn early (ignoring jump points - like the AI blocking the location) really ought to be the sum of the profits that you make on each subsequent turn, discounted by a factor that represents the time required for the investment to mature.

"Loan me $99 today, and I'll pay you $100 tomorrow" is a decent deal for you.

"Loan me $99 today, and I'll pay you $100 a year from now" is not nearly as exciting. Sure, you come out ahead, but not by so much that it's worth the hassle.

0 0 0 1 2 3 4 5 5 5 5 5
1 2 3 4 5 5 5 5 5 5 5 5

On the turn in bold, the raw difference is 5-2 = 3. The present value of that difference is something like (5-2) * exp[ - 4k ]. 4, because we're considering a profit 4 turns from now.

The k value is essentially arbitrary - you choose some discount rate that feels reasonable, and you don't worry about it. For instance, you might decide that a difference of 1 hammer 100 turns from now should equal .01 hammers now.

.01 = 1 * exp[-100 k]
ln(.01) = -100 k
ln(10^-2) = -100 k
-2 ln (10) = -100 k
ln(10) / 50 = k
0.04605 = k

Using the numbers from your example, instead of 15 hammers total difference, the value would be close to 13 hammers.

Probably not a good trade for the 20 hammers of forest + the worker turns.

 

[VoiceOfUnreason]

Double post: pilot error.

A few extra thoughts, that should discourage anyone from going further down this trail.

1) There are some additional calibration points available when trying to choose a suitable value for k. For instance, if you believe in Worker First Almost Always, then the value of getting your initial improvements sooner had better be more than the alternative of training a warrior and growing first. Likewise, you need to make sure that training a settler now is better than waiting for one more turn again, or you never get the second city out. And to make sure that settler first isn't always the right answer.

2) The value of hammers decay while you are training a unit/constructing a building. The price of a warrior is 15 hammers, the cost is 15 hammers plus the interest you lost while training him. Pretty much any economic move you make is going to be exchanging a short term cost for a longer term gain (Cottage suck, why do we ever work them? Because Towns don't suck.) In short, on a short enough time scale, pretty much every build is a negative

5) More good news - none of that matters because it considers merely cost. What we actually need to be examining is marginal cost - how does each move (properly contextualized) fare in comparison to the real alternatives that are available. You're not even really getting a measure of profit, because the inputs to the calculation are arbitrary. All you get is a ranking, with some fuzzy boundaries where choices are too close to call.


What's this got to do with winning? Not a lot; the amount that it matters falls off the more you think about it. I've got a formula for that here....

 

[vranasm]

hmm I think I start to see how some people could eventually feel when I post "straightforward" copy of some function from source code of civ everytime I read some deep thought VoU+vic posts with maths...

As for settling the settler... I read somewhere that city 2 is always net gain and thus I wouldn't delay it much.
As for later cities...it depends... but if you build the settler then not settling him feels dumb, since that basically means you could have used the production of capital for something else.
If the number of turns you feel you need to wait with settling exceeds for example 5 turns, then I am pretty sure the capital had enough time to build worker instead the settler, if for example happy cap was issue and you needed to stall growth.

What I find lately in SGOTM, there is some following pattern going on in these games (Emperor difficulty) and that is how quickly the team meets 4-city empire. That feels/is some magical number tied to the city maintenance spike.

4-city empire is optimal from maintenance point of view for quick teching. This number of cities should block enough land for later claims, but is worth delaying a bit the expansion for techs to catch up a bit.
This of course could be influenced by resources, so if you have gems, gold. Bigger happy cap through ivory, fur, there could be argument that this magical number is higher then with more generic land.

Not sure on immortal/deity, but I saw a lot of games from AZ where the magical number is either 3 or 4 for his rushes.

 

[krikav]

A thing that could fit into this thread, is improvements on resource tiles, that are not specificially designed for that resource.

For example, a farm on a sugar or banana tile.
Mining pigs/sheep hills are also a nice example.
I have also built cottages on calendar resources many times.

Sometimes a mine on a pig-hill is just the best improvement, as in the case of a city with very few hills and a over-abundance of food.
But other times, one just mines the pig to avoid the costly tech AH.

Calendar resources often get utilized in other ways than plantation, becuase plantation arrives much later than other worker techs.

 

[vicawoo]

Thanks for laying it out like that, vicawoo. Two comments on your analysis: the offset gets "reset" so to speak when you hit the happy cap and only applies at the margin anyway (i.e. if settling earlier gets you the Nth citizen earlier than what you actually did, the implied costs are what that citizen would've given you).

The other comment is that when you have a recurring but shrinking benefit like what you described, you can approximate its value by using a geometric series. I'm specifically talking about the power series related to the time-value formula under economics. I think that the ratio in the denominator would be the production the earlier settler will give you over the total amount of production used to produce the next settler.

So, three things: marginal benefits, Time-Value of a settler and the Time-Value of a worker. Four things: MB, TV of a city, the TV of city development and NOBODY EXPECTS THE SPANISH INQUISITION!

Adjusting logarithmically is probably more reasonable than the naive linear approach I took. However, I have to ask what assumptions make it valid in economics, and do they apply in civilization? In economics, interest, inflation and all that creates close to true exponential growth.

Civilization it's kind of like that, but horizontal growth is limited by city maintenance (which empire wide increased per city linearly with cities and distance, so overall quadratic), and vertically it's semi-linear in population (10+n), but techwise it's harder to say.

---

I hated these problems so much, which is why almost all the calculation contributions I've done focus on maximizing the simpler tile yield per turn, or translating it (like a forest chop) into sustainable yield/turn (1/5 of a settler).

Finally either approach tends to indicate that over the long run, more cities (with reasonable maintenance) are better - more so when you can translate hammers to research, which is why I'm philosophically at odds with absolutezero or mylene's super early tech approaches. Assuming you can get away with being greedy.

 

[JammerUno]

^if your analysis stops at a bland 'more cities is better' conclusion there's really nothing I can disagree with.

However:

1. inflation does not require an exponential system, you could simply have declining inflation as the system grows
2. civ games tend to have exponential patterns anyway

That leads to the equally bland conclusion that early game advantage will lead to midgame advantage in citycount, which in turn supports 'more cities is better'. The objective is getting to a certain number of cities which gives you victory (simplified). Whether you go from 4 cities at 1ad, to 10 at 1000ad to victory in 1500ad or from 12 cities at 1ad to 12 at 1000ad to victory in 1500 ad makes no difference.

That does imply that worker management in the early game is more important than in the late game, which in turn means we're in luck, because it's less work in the early game.

 

[dalamb]

If you group select say four workers and have them build something and it finishes in not exactly the amount of right worker turns the extra workers will sit idle.

Is that still true with the BUG mod? I think I've seen a stack finish and the one with move points left wake up; you might still need to unstack it to get it to do anything, but you don't need to lose turns.

 

[AutomatedTeller]

With something, extra workers are ready to go. (so if you set 2 workers on something that takes 5 worker turns, of them will be ready the turn the improvement competes).

I dunno if that's Bug ( I run Buffy, which has bug) or BtS or what.

 

[Um the Muse]

What did you mean by "ready to go" though? If it means that the extra worker wakes up, that's not part of the base game. If you mean, "I can click on the extra worker, seperate it from the stack and send it on its merry way if I happen to notice this happening" then it is indeed a part of the base game.