Toward Good Humankind City Layouts

Let's explore what a "good" HK city will look like. In a previous thread (Stability and Food: A Quantitative Romp through Scenario 3), I defined SAX to comprise self, adjacency, and multiplier, which we earn from certain Quarters Extensions. Here, I shall ruminate upon the geometric layout portion of SAX, which corresponds to tiling a plane (or subset thereof, haha) with Quarters so as to maximize adjacencies.

• I ignore all Infrastructure families. I will mostly analyze average "base" SAX per hex, which is your SAX with a multiplier of 1X, corresponding to your "level 0" (L0) Infrastructure family of having researched and constructed nothing yet.
• I ignore all terrain, and all Deposits upon them.
• I use the description of Quarters and Infrastructure families from OpenDev.  Knowing that it's pre-Alpha, I do expect some idea drift (and inflation :) before release. For example, we don't know if every Infrastructure family will maintain identical +1 bonus (or +2 for Education) across all levels, so perhaps the concept of "family as multiplier" won't even hold.

Disclaimer: I don't have an optimal solution yet for HK city layout. Most of this section will thus consist of analyze-and-test.

Acknowledgment: I generated all hex grid figures using the free version of Hexographer by Inkwell Ideas (which I found on p.1 of a search for "hexagon grid editor"). I use the ∴ symbol (mathematical or logical therefore) as a hex grid ellipsis, denoting a continuation to infinity in that direction. I added the arrows and their labels with Corel PaintShop Pro.

2. "Sticky" Quarters and Infinite Subsets of the Plane
Farmers Quarter and Makers Quarter both have the form:

• +1 SAX for itself
• +1 SAX per self-adjacency (of an identical Quarter)

This naturally suggests large homogeneous regions of each Quarter. Colloquially, they want to stick together and create large blobs. I shall call them the "sticky" Quarters, and consider only them throughout this section.

2.a  Infinite Plane

First, we obtain a theoretical upper bound on average SAX for sticky Quarters.  Consider the infinite 2D plane, fully tiled with an infinite number of Farmers Quarters :)

Fig.1  Infinite plane of Farmers Quarters

Generally (that is, in every case except this one :), we consider a finite connected region (of sticky Quarters of one type).  Then some hexes are inside the region, and all other hexes aren't.  Loosely following solid geometry, I define a hex as an interior hex of a region if all 6 of its neighboring hexes are also interior hexes. (Yes, it's recursive.)

Fig.2 Interior hex

Lemma 2.1.  Every interior hex has (1+6) = 7.0 SAX, and this is maximal.

Proof: An interior hex has 6 neighbors. For a sticky Quarter, it scores +1 for itself, +6 for its neighbors, so its base SAX is (1+6) = 7. This is maximal because a hexagon has only 6 edges.

Lemma 2.2. The infinite plane has 7.0 average SAX per hex.

Proof: On the infinite plane, every hex is an interior hex.  Conversely, the 2D region in which every hex is an interior hex is, indeed, the infinite plane.  (Programmer's joke: This is what happens when you recur with no base case.)

2.b Infinite Half-plane

Interestingly, we can chop the infinite plane into subsets (that are still infinite), and the average SAX is still 7.0 per hex.  Consider the infinite half-plane:

Fig.3 Infinite half-plane

The half-plane is defined by a boundary edge, which consists of some edges of some hexagons (here, 2 edges each).  It partitions the plane into two regions called interior (inside) and exterior (outside).  While the boundary edge between the two regions is necessarily jagged, we'll look instead at the column of hexagons.  If that column is a straight line of hexagons (i.e. orthogonal to one of the hexagon's edges), we'll call this a "straight border" (of hexagons).  Abusing the terminology somewhat, we label the hexagons incident to the boundary edge as "border hexagons". It follows that a border hexagon necessarily has some exterior neighbors, and therefore is not an interior hexagon.  In fact, border hexagons always score less than the optimal 7.0 SAX per hex.

• In solid geometry, when we talk of points and neighborhoods, there exists a 3rd category for points, called the boundary, which is neither interior nor exterior, but which separates the two.  Then every point neighborhood test uses an open ball of radius epsilon, as small as you(r opponent) likes, so small that it can subdivide even the Planck length and still contain infinitely many points.  But at the higher-dimensional granularity of 2D polygons in a planar tiling, we naturally use polygon edges to define region boundaries, and so each polygon is either fully inside or fully outside: there are no true boundary polygons.  I use the term "border polygon" (note: not "boundary polygon") as a notational abbreviation for the concept of "polygon with both interior and exterior neighbors".

Fig.4 Straight border hex 

Lemma 2.3.  Every straight border hex has (1+4) = 5.0 SAX.

Proof: (The Endless Lemma) Every hex in a straight border has 4 neighbors.  (Fellow players of Endless Legends will surely recognize this.)

Lemma 2.4.  The infinite half-plane has 7.0 average SAX per hex.

Proof: Consider any one row, from the border hex to infinity, as shown in Fig.3.  This row has 1 border hex and (n-1) interior hexes, as n → ∞.  Thus, it has average value

• = [ 5 + 7(n-1) ] / n
• = [ 7n - 2 ] / n
• = 7.0 - 2/n

As n → ∞, the 2/n term vanishes, and we're left with the same 7.0 SAX as for the infinite plane.

From this, we extract our first insight:

1. An optimal (infinite) region of sticky Quarters can have a finite boundary.  Immediately, we may envision two optimal regions, each with a straight boundary, and ... abut their straight boundaries against each other.  Keep that thought in mind!

2.c Infinite Ribbon

In fact, we can add a straight border on the other side of the interior.  Consider the infinite ribbon of width w:

Fig.5 Infinite ribbon of width w

Lemma 2.5.  The infinite ribbon of width w has (7.0 - 4/w) average SAX per hex.

Proof: Each row of width w has 2 border hexes and (w - 2) interior hexes.  The average value is then:

• = [ 5 + 7(w - 2) + 5 ] / w
• = [ 7w - 4 ] / w
• = 7.0 - 4/w

2.d  Finite Quadrilateral

2.e  2xN Stick, c.f. Endless Legends

Lemma 2.8.  The 2xN stick has (5.0 - 3/N) average SAX per hex.

Proof: Every non-corner hex is a straight border hex (by construction), hence it has 4 neighbors for (1+4) = 5.0 SAX.  In EL, 4 adjacencies (to L1 districts) levels up a district to L2, and so every non-corner district levels up.  The -3/N term is the penalty we pay for the 4 corner hexes: each pair scores (3+4) = 7 SAX total, which is a constant -3 below the 5.0 SAX average.  Hence the total penalty is -6 / 2N, which reduces to -3/N.  In EL, these 4 corner districts always remain at L1.

So we could replicate EL cities in HK.  Is this a wise direction to pursue?  For moderate values of N, a 2xN stick layout will yield 4.xx SAX. This is considerably less than the optimal 7.0 SAX, but how does it compare in practice to other "good" tilings? Recall that the full hex ring (see table in previous post) scores 31 SAX in 7 hexes = 4.43 SAX per hex. Let's solve for the value of N at which a 2xN stick is equally as efficient as a full hex ring. We have:

which rounds up to N = 6. Check:

So the 2x6 stick is, indeed, more efficient in SAX than a full hex ring.  But this presumes an enormous 12 sticky Quarters of the same type!  Even Scenario 3's Londonia wasn't that large.  In practice, I think both the 2x6 stick and the full hex ring are equally impractical, and we'll spend most of each game of HK building much smaller subsets thereof.

Finally, note that a partial hex ring under construction is identical to a 2xN stick under construction, from 1 hex to 5 hexes. Only the 6th hex forces you to commit to one shape or the other, either by making it too wide to fit the 2xN stick pattern, or too long to fit the hex ring:

Fig.8 Partial hex ring and 2xN stick, from 1 to 6 hexes

So we should expect that the 2xN stick's SAX formula is consistent with the partial hex ring formula for N = 1 and N = 2 -- and it is.  In fact, they also agree at N = 3!

2.f Insights for Sticky Quarters

Recall that we consider only the sticky Quarters (Farmers, Makers) here.

1. The optimal (theoretical maximum) average SAX per hex is 7.0.
2. An infinite tiling of sticky Quarters can have 2 finite borders, and still be optimal in SAX.
3. A finite tiling can achieve 3.0 SAX at 3 hexes, or up to 4.0 SAX at 6 hexes, and these values are optimal at these sizes.  This looks feasible.
4. The 2xN stick is more efficient than the full hex ring at 12+ hexes.  Alas, this is almost surely impractical.

• [_] Exercise: What about the infinite half-ribbon with 3 finite borders?  Work out its average SAX per hex.

The inchoate idea I get from this is to use sticky ribbons as building blocks, and candy-stripe the plane with parallel ribbons.

Awesome math and job done !!! *thumbs up* :)

However I might add that it was exactly this type of stuff that made Civ 6 city building being so boring too me. It just felt like everything was about min/maxing and tables of math and diagrams. The joy of simply playing and building Your cities felt like it was gone. It is what made me dislike unstacking cities as a system, after having initially being excited about it after I heard about it announced before release. It has made civ 6 turn into almost like Raiding in a MMO RPG where You do spreedsheets of best combination of stats on armour mixed with rotations of skills etc. Turning games into some advanced school or workplace assignement. While me myself too having made such spreedsheets back in the days of Boardgames, Role Playing Games and pen and paper. Even doing that on the early days of PC gaming. I then came to realize it ruins the fun at the same time for me.

I used to watch guide videos of Civ 6 on how to best build triangle shaped farmlands and where to place Your districts to make most use all adjacency bonuses etc. However after a while I didn't feel like I was playing a game anymore in my free time. It more felt like boring and tideous job to me.

I think there is some kind of sweet spot where a Turn Based Strategy game is still fun but also challenging. Go over than line/spot and it becomes more like Grand Strategy and in my opinion not so fun anymore.

That being said, I am still impressed with all Your work and such that You put into this. I just hope that Humankind won't turn into this tideous min/max game that Civ 6 city building became. I hope there is a way for Amplitude to make unstacked cities more fun than Civ 6 managed to do.

Please don't get discouraged from what I write. It's an awesome job You're doing with these tables and math and all. Do keep it up. I'm sure many will appreciate it. Each to ones own of course. I'm sure I too will make some use of it to some degree. 

I will even give You an Upvote and all. 

4. Londonia from Scenario 3

I recreated Londonia's Extensions and terrain from screencaps, for most of its original territory only.  I never thought to screencap its southern territory, so the map stops at row 03.

Right away, we can see the ghastly effects of its starting terrain :)  It's blocked to the SW by the volcanoes, and to the SE by mountains.  Also, it already encroaches its northern territory border, which was a roadblock until we attach the Marble territory at the start of Scenario 3.

Given these limitations, Londonia is not too bad.  It has 12 Extensions:

• 2 Strongholds, Westmoor at [43,08] and Eastgate at [49,10], don't count for FIMS or SAX.
• 5 Farmers Quarters in the east "wing" are optimal for 5 hexes.
• 1 Farmers Quarter at [46,07] is worst-possible (for now).
  • Surely the English player intends to grow a new western wing of 5 Farmers Quarters, in which case this is an investment for the future.  There's just enough room west of Londonia's city center to build a complete 5-hex stick.
• 2 Makers Quarters at [47,05-06] are optimal for 2 hexes, and have an obvious growth path to the SE, into the two starred hexes.  (That's exactly what I did.)
• 1 Research Quarter at [46,06] is optimal for being trapped in a 1-hex dead end.
  • On reflection, I think this choice is pretty good.  This hex could have been a 3rd Makers Quarter, with obvious advantages for Industry, but then you get no Science at all.  You don't want to put Research Quarter(s) in the starred hexes because that completely blocks your Makers stick from growing southward.  Given that the English player chose to settle this close to the volcanoes, and can't grow north because of the territory border, this trapped hex never held much promise.  Getting optimal Science SAX from it is quite clever.
• 1 Market Quarter at [49,06] is ... bad.  Indeed, it's worst-case for SAX.  I don't remember what FIMS it collects -- maybe from the mountain at [50,05]?
  • This could be an awkward compromise for terrain.  It does grab land, enabling an Extension at "Doveria" [50,06] to capture the mountains' FIMS.
  • This Market Quarter could be at [48,06] along the east wing's south straight border for +1 = 3 SAX in Money.  Then it would block the east wing from growing southward, but they were already blocked by the mountain at [50,05].  Actually, [48,06] could be a 6th Farmers Quarter to complete a length-3 triangle.  It's the classic trade-off: -2 sticky SAX for +1 hermit SAX?  (Psst: English are Agrarian, they Neeeeeed Moar Food.)
  • Or it could be at [50,07] along the east wing's east straight border for +1 SAX.  But that blocks the east wing from growing eastward, and that looks like a no-go.
  • It could have been anywhere along the east wing's north straight border, at [48-50,09].  In fact, that looks like a wonderful place to lay down a full line of Market Quarters.

1. Considering SAX only, I would have put the Market Quarter along the east wing's north border.  (But grass/woods is probably terrible for tile Money.)
2. I might move the singleton western Farmers Quarter to [50,07] to extend the east wing into a 2x3 stick.  There's enough room to grow eastward for another +5 Farmers Quarters or so (including this one).  This ultimately depends on the reasonable upper bound of Extensions cap for an Era III city (and whether you keep the same city through Era IV and beyond), which we don't know yet.  If Londonia's Extensions cap will be ~20, then growing east suffices, because we'll run out cap and real estate at roughly the same time.  But if the cap will reach ~35, then it's reasonable to build westward and expect to actually finish a 5-hex west wing.  So I can't judge whether this isolani is an investment or a bad deal.

Fantastic and concise information about basic quarter adjacency and growth layouts given the information from OpenDev. It's already very impressive what can be done before wonders, religious quarters, and culture specific variants are taken into account.

Lord_Funk wrote:

However I might add that it was exactly this type of stuff that made Civ 6 city building being so boring too me... After a while I didn't feel like I was playing a game anymore in my free time. It felt more like a boring and tedious job to me.

@Lord_Funk 

The placement puzzle offered in Humankind is much more complex and dynamic than Civ 6. The setup (from what I have seen in OpenDev) is unlikely to be min/maxed to Civ 6 levels even with perfect knowledge due to culture progression. 

What has been laid out in this guide are simply the vanilla districts without consideration for variants available during specific eras. An early game agrarian (Harappan) or builder (Eygiption) quarter with irregular adjacencies can skew the hard divide seen within Gilmoy's diagrams. Such variation further skews placement puzzles because the quarters are interchanged with different or vanilla ones as cultures become intermingled.

Hypothetically, players could see Harappan farmer quarters (canals) adjacent to vanilla farmer quarters and Haudenosaunee farm quarters (Three Sister Plantations) without commonly shared SAX between farm quarters. 

What Gilmoy has layed out is the foundation of how to evaluate variant quarters compared to standard quarters. Even slight changes such as the Babylonian astronomy house's adjacency/exploitation yield(s) impact a city's quarter placement throughout the entire game. If Gilmoy mapped a city placement with just the astronomy house, the cityscape would already be different. 

I really have to throw in here because all this talk about sticks is just more complicated than it needs to be. Just look at the marginal gains of each extension. There is no reason to confine to building in a stick. The marginal gains are the same regardless of shape unless you double back. Consider the following:
Here we have a stick in progress, already built. So what's our next move?
Well one option is to continue the stick. If we do that then we get two more adjacencies.
Or we could do this. Now we have a bent stick, but still two more adjacencies.
We can just go off in that direction now, with a kink in our stick and continue to gain two adjacencies with each spot.

Or we could just make a lump on the other side. Still only two incremental adjacencies.

Now I can continue with this logic and make a snake. Build order is red, orange, yellow, green, blue, purple, repeat. Note that with each extension I gain two more adjacencies so this snake is identical to a stick.
However there is a better option. If I double back you can see that red only gets me two more adjacencies. However when I build orange, yellow, green, blue I gain three adjacencies each time.

Now consider this: I had a little stick, I doubled back, and orange and yellow got me three adjacencies instead of two. What next?
Well I can just double back again. Blue only gets me two adjacencies, but purple got me three. Am I done now, since I can't double back on the left? No.
Red gets me two, but orange gets me three again.

So in conclusion, sticks are dumb. Snakes are equivalent and allow you to follow terrain, and zigzag blobs are better. Ideally you make a hybrid of a snake/zigzag blob that follows the natural terrain, winding through plains or hills, doubling back when possible, and then put "hermit" quarters along the edge or on borders between hills/forest and plains.

Lord_Funk wrote:

However I might add that it was exactly this type of stuff that made Civ 6 city building being so boring too me... After a while I didn't feel like I was playing a game anymore in my free time. It felt more like a boring and tedious job to me.

Tainted wrote:

@Lord_Funk 
...The placement puzzle offered in Humankind is much more complex and dynamic than Civ 6. The setup (from what I have seen in OpenDev) is unlikely to be min/maxed to Civ 6 levels even with perfect knowledge due to culture progression...

@Tainted Hmmm, I found the the only real test we had of this in scenario 1 Babylon was too limited to get a feel of how this all comes together in an acctual gameplay. If it is as You say, I fear it might be too tideous and advanced. Borderline or even full blown grand strategy in the sense that it won't have a natural fun game flow, but must be planned in every small detail until it feels like sitting in an office with spread sheets rather than just play a game and have fun. That is how I felt with Civ 6 as far as the unstacking of cities system goes. It was what turned me off to that whole system/mechanics with districts taking up tiles and have adjacency bonuses etc. Hopefully it will not feel like that in Humankind, but I musdt admit Your reply made me a bit worried that it will.

I don't know about others, but as an adult with all the tideous stuff You already have to do with possibly supporting Your family with kids, maybe being a single parent even, workplace and tight schedules, bosses that are on Your back, or if You are a boss the preasure to get a project to tie together with finances etc. Once You have time for just Yourself and Your hobby, like playing a game, You really don't want to have that feel like just another job. You really don't feel the need to challenge Yourself to some breaking point to prove Yourself on Your free time, as You get enough of that during all the other obligations in life. I found Civ series 1, 2, 3, 4 and all the way up to and including Civ 5 was perfect in that sense. Advanced enough to be intriguing and even challenging at times, not silly easy, but at the same time not so advanced that it gets boring. Civ 6 crossed that line in that it was both silly as the AI was totally broken with all the new systems, as well as feeling like having to sit with spread sheets to plan Your cities. That is the reason I don't play Grand Strategy, or a game like EVE Online. They require that You basically dedicate Your life to it. No idea how anyone with a family and a job could ever find time for such games (unless you're a game dev, full time streamer or youtuber) ???

Hopefully Hukmankind will balance it well and not become like that.

Don't get me wrong those are brilliant games, EVE Online and Grand Strategy I mean (civ 6 is just bad). But just not for me. I could never find the will nor time to invest in them, as I never found them fun, just tideous. At the same time I don't want to sit and place silly easy games either. Balancing a game and finding that sweet spot when desigining it is the key to success I think.

It's so frustrating to be at a point in life where You can acctually financially afford hobbies like gaming (not through your parents wallet), buying games, a decent PC, nice monitors etc and upgrade it regularly and such, but less and less games are designed for You, but seems to be more and more aimed at people who do nothing else than gaming 24/7 with no other real life obligations, or so it often seems anyway. MMO RPG's are definetly more and more designed that way, like a full time job for truly addicted MMO RPG Gamers. That could be one of the reasons why they are dying slowly but surely one after another, as most people can't live like that. The ones that can spend their time 24/7 in a game often can't afford it financially themselves as they are home 24/7. Only a few MMO RPG's will survive as there are only enough people that can live and play like that to support very few games. Compared to the golden age of MMO RPG's. I don't want Turn based Strategy Games to go the same way and only become for the few that can play 24/7 and invest all their time and lives into it, as if it was a job. No I don't want Turn Based Strategy games to be silly easy, but at the same time not needing to be a stay home unemployed rocket scientist with spread sheets to play them either, but acctually be fun to play.

That is what I am missing in Civilization 6 and hope for Humankind to be able to provide. That would be a game and a game series I'd be happy to pay for and play (when I have time to spare) for years and years to come.

Lord_Funk wrote: @Tainted

 If it is as You say, I fear it might be too tideous and advanced. Borderline or even full blown grand strategy in the sense that it won't have a natural fun game flow, but must be planned in every small detail until it feels like sitting in an office with spread sheets rather than just play a game and have fun. 

Hopefully Hukmankind will balance it well and not become like that.

@Lord_Funk 

I completely understand where you are coming from about unpacked cityscapes. I would like to address how I think Humankind solves such an issue with a few bullet points.

 -- Civ 6 cities may only build one of each district (without take-backsies) putting too much importance on initial placements and planning (thus the pin system).

Humankind allows multiple of each quarter to be placed and "take-backsies." The science/civic production cost increases heavily incentivized dropping districts first (which I loath with a passion) and getting back to them later strategies. 

 -- Civ 6 district bonuses only took into account limited adjacencies plus others requiring build charges (additional production investment).

Humankind quarters "exploit" the surrounding tiles making it abundantly clear how good a quarter already is without considering or planning for other quarter adjacencies.

The quarters in humankind have more in common with Civ's unique tile improvements (kurgan, Nazca line, etc.) than districts. It is abundantly clear upon placement how well they will interact with their surroundings and how easily they are changed. Players could even say suzerain dependent improvements becoming unbuildable (due to conquest, lost suzerain status, etc) is similar to culture specific quarters. 

As stated before Humankind's approach (in my opinion) appears more complex and dynamic than Civ 6, but this does not mean it is tedious, unapproachable, or boring. Quarters in Humankind seem more transparent and straightforward in their placements as you know what your getting without the fuss of planning everything 60 turns out.  

Tainted wrote:

Lord_Funk wrote: @Tainted

 If it is as You say, I fear it might be too tideous and advanced. Borderline or even full blown grand strategy in the sense that it won't have a natural fun game flow, but must be planned in every small detail until it feels like sitting in an office with spread sheets rather than just play a game and have fun. 

Hopefully Hukmankind will balance it well and not become like that.

@Lord_Funk 

I completely understand where you are coming from about unpacked cityscapes. I would like to address how I think Humankind solves such an issue with a few bullet points.

 -- Civ 6 cities may only build one of each district (without take-backsies) putting too much importance on initial placements and planning (thus the pin system).

Humankind allows multiple of each quarter to be placed and "take-backsies." The science/civic production cost increases heavily incentivized dropping districts first (which I loath with a passion) and getting back to them later strategies. 

 -- Civ 6 district bonuses only took into account limited adjacencies plus others requiring build charges (additional production investment).

Humankind quarters "exploit" the surrounding tiles making it abundantly clear how good a quarter already is without considering or planning for other quarter adjacencies.

 

The quarters in humankind have more in common with Civ's unique tile improvements (kurgan, Nazca line, etc.) than districts. It is abundantly clear upon placement how well they will interact with their surroundings and how easily they are changed. Players could even say suzerain dependent improvements becoming unbuildable (due to conquest, lost suzerain status, etc) is similar to culture specific quarters. 

As stated before Humankind's approach (in my opinion) appears more complex and dynamic than Civ 6, but this does not mean it is tedious, unapproachable, or boring. Quarters in Humankind seem more transparent and straightforward in their placements as you know what your getting without the fuss of planning everything 60 turns out.  

@Tainted Thank You for Your reply !!! That does sound alot better. I'm glad it seems You truly understand what I meant. It is very hard to express a feeling, idea or anything really in a foreign language and only via text online with not being able to react to the others replies at each moment if a misunderstanding arise. I even forgot to mention what You brough up with the increase in cost of placement of districs tied to science and civics and how one just tries to rush to place it and then build and worry about it later. I also loathe that with a passion. That to me shows that You really get what I meant and tried to explain. Which also makes me hopeful that Your more indepth understanding and explanation of Humankinds way of handeling the unstacking city system is correct and thus alot more fun that Civ 6's way of doing it is. That one can more easily get an overview in the here and now where to put an extension/quarter without having a spread sheet and planning 60 turns ahead. That gives me hope that it will be a fun game and not like some tideous spread sheet planning job in the office. :)

As a side note, why would it take longer for a more advanced society with modern machines and such to build a factory than a less advanced or evolved civilization ??? I mean today You can build a shoppingmall or storage facilities and even factories in no time with building it from pre-made parts and such. That the more research and more advanced You become scientifically or evolved civic wise the longer it takes to build a district never made any sense to me. I understand that the production cost can not be fixed either, as then later in-game as You get more production in Your cities each quarter would take 1 turn to build (and that would be way to over powered of course), but it should definetly be better balanced as to how more expensive it becomes, not sky rocket the cost like it did in Civ 6 (especially on marathon and epic speed) and not via science and civics penalty at all. The implementation of that whole system was a big miss in Civ 6, not a hit at all in my opinion.

I would like to say city building is never a static planning problem. It's a dynamic planning problem. Some effects early on might worth much better effects that comes later, as your early advantage can accumulate lots of fame & eliminate a competitive AI. It's the same logic for civ6: the earlies bonus is the best. I used to (and still love to) build a +12 theatre square in civ6 (+13 if machu pichu) but it turns out I'm usually already leads AI greatly if that 6 wonders are completed on my capital.

For Humankind, early exploitation and infrastructure is extremely important. A Harappans capital with 6-7 river forest tiles within one hex of their city center and a lumber mill can probably be better than the majority of cultures with careful city planning. And you probably never get a chance to finish your plan if you have a lucky and aggressive Harappans neighbor like me who will send you infinite number of units before you can even build your districts.

That's why I also rate Zhou's emblematic district better then Babylon's astronomy house if you got some mountain regions. It's not hard to have one +21 confucius and a +16 confucius school in your capital by turn 20, when your babylonian neighbor is still buried in ancient darkness. Zhou's districts produce remarkable science without needs of researchers, freeing them some extra worker and farmer which can produce an army to take their babylonian neighbor's wonderful cities.