[SPOILERS] Suboptimal Goes Waltzing Matilda

[suboptimal]

Turn 8

Take a look at the scores, I’m the last to 2 population.  No matter.  That will change soon enough.  Irukandji switches to a settler, that’s due in 9 turns.  In the west I send the warrior onto the next hill over.  The western shore of the small lake has coffee with jade to the north.  



There’s no fresh water or coast immediately visible.  The coffee might make for a good spot to settle, but I’d need to see what else is west.

The scout moves northwest onto the hill:



Coffee, coast, fish and scout.  The scout will move W onto the hill next turn (unless the barbarian beats me to it) so that I can see the full coast line.  It does look like I’ll be able to get a nice compact core going here with red dot to the west and a blue dot between the wheat tiles.  That spot gets me the Breathtaking coastal tile by the mountain for a Campus or Commercial Hub and leaves the just-found coffee available as a potential site later on.

[Ichabod]

Have you considered buying the spices tile or do you have a plan for the early gold?

Regarding your planned second city, it's worth noticing that once you settle on the forest, the appeal of surrounding tiles will go down, because forests raise the appeal of tiles. Perhaps that can alter your plans. Besides, I wouldn't be so focused on settling on fresh water considering you have the Australia Coastal bonus. That's 6 housing for Coastal cities, that you can easily increase with farms and pastures. It's way more than you need in the early game.

[suboptimal]

Have you considered buying the spices tile or do you have a plan for the early gold?

I plan on buying the spices as soon as I have the gold available (next turn) in order to get the 3/2 tile into service.  Early gold after that will be held for future uses - a combination of slinger -> archer upgrades and a trader as soon as I have the gold and Foreign Trade completed.

Regarding your planned second city, it's worth noticing that once you settle on the forest, the appeal of surrounding tiles will go down, because forests raise the appeal of tiles. Perhaps that can alter your plans. Besides, I wouldn't be so focused on settling on fresh water considering you have the Australia Coastal bonus. That's 6 housing for Coastal cities, that you can easily increase with farms and pastures. It's way more than you need in the early game.

You make a good point about the tile appeals.  Both of the tiles that are Breathtaking appeal are +4 but that's the minimum.  I'll have to look at some other options (settling on the hill currently occupied by the warrrior would be a good spot - could buy the closer horse tile and then the culture bombs get me the other horse and the wheat.  The coast tile can be purchased for a Harbor if I so desire.  Fortunately lumbermills and harbors don't decrease tile appeal.

Regarding the coastal bonus to housing I've definitely got the housing bonus in mind.  The current area supports settling on fresh water just because of the layout. The other consideration about being close to rivers while coastal is that rivers provide +1 Appeal to adjacent tiles (for a +1 yield bonus) and I think (but haven't checked for sure) that the +3 housing bonus stacks with the fresh water bonus (not that housing is hard to come by).  

Edited to add strikethrough and the following: Settling the blue dot city between the wheat was primarily to get the Breathtaking tile into the 2nd ring for a Campus, but if I shift Red Dot southwest I can just go coastal and put that tile and the wheat first ring.

[suboptimal]

Follow-up to Cultural Expansion

I took another look around to see if I could locate anything in the Civ 5 code that could cause the picker to flip-flop between tiles and glean some further insights.  Here's the update:

- Tile Flip-Flopping: the tile scoring routine returns a list of tiles that have the lowest score.  If there's one tile at that score, that's the one the picker takes.  If there are multiple tiles at that score the picker randomly picks a tile.  Obviously this can change whenever the game is reloaded or the routine is called during the pre-turn sequence.

- Bonus resources are valued less than strategics & luxury resources.  1 is added to the influence score for bonus resources.  

- An additional evaluation for unclaimed tiles is the number of non-bonus resources and natural wonders in adjacent unclaimed workable tiles.  For each strategic or luxury resource or natural wonder that is in an adjacent unclaimed tile AND the tile can be worked by the city (within the 3rd ring) the influence is decreased by 1.  

Second Edit, HUGE Edit

When I played Turn 9 my capital had expanded to the eastern citrus tile and selected the spice tile to go for next.  Using the MP-based scoring method the NW citrus should have been next.  This leads me to believe that it's not MP-based.  The use of a pathfinder in the Civ 5 code leads me to believe that the distance is equal to the ring count (all 2nd ring tiles have initial "influence distance" of 2) unless the path has to go around an impassible tile (mountain or natural wonder).  Note that the "impassibility" doesn't matter if the tile is the "destination" and it won't matter if there's a lone mountain or a single-tile natural wonder. 

A back-of-the-envelope rescoring using this method for influence distance and accounting for the adjacent unclaimed resources gave the eastern citrus a score of 90 and the spice and NW citrus a score of 91.  Based on this the choice of the initial citrus was correct and the picker would have flip-flopped between the spice and NW citrus if I didn't buy the spice. 

I will re-write the initial detailed post to be more...concise...and show the new scoring around Irukandji.

[suboptimal]

Turn 9

Irukandji grew to the citrus tile 2E of the city center.  The tile picker has selected the spices as the next tile to go for.  For those who read through my two posts on the possible tile picking algorithm this implies that it is absolute tile distance.  That means that when I buy the spice tile...now...the tile picker should be selecting the NW citrus at the start of next turn.  The citizens in Irukandji realign to work the spice tile and the SE forest.  Growth in 7 turns.

The warrior goes around the jade to the north and then heads west.  Cornflakes has definitely roughed up the coastline because I can’t remember seeing anything like Red Dot Cove on a "standard" Inland Sea map:



Settlervision shows that there is an outer city ring that runs along the line formed by the NW stone, mountain and the grassland hill forests at either end.  Hopefully this is another city-state.  Moving to the NW stone next turn should reveal that city center. Back at Irukandji the scout moves west to the hilltop to reveal the coast at the end of the river.  Nothing special in terms of resources.  Time for some dot maps.  In the east:



District placements for Irukandji and Blue Dot are marked.  The harbor at Blue Dot, if placed,  would provide an intrinsic +3 and bonuses of +1 to the campus and +3 to the Commercial Hub.  The tile for the campus has a forest on it; however the forest doesn’t affect its tile’s appeal, just those around it.  At Irukandji the placements are straightforward.  



As discussed earlier in response to Ichabod the Red Dot location needed to move because putting it over the forest reduces the appeal of all adjacent tiles by 1 and the Breathtaking tiles would drop to Charming.  Likewise I can’t chop the two forest tiles marked “No Chop” as they’d also reduce the appeal of the surrounding tiles.  I can, however, build Lumber Mills.  The harbor here would be +3 and provide +1 to the Campus and +2 to the Commercial Hub.  

If I can manage all of the Campus & Commercial Hubs these three cities could generate 11 and 11 from unimproved districts relatively early in the game and give me a trade route capacity of 4.  Two Harbors would add another 11 and 2 to the yields. 

[suboptimal]

Cultural Expansion & Tile Purchase Costs - “Final Version” (for now)

I’m calling this a final version as I’d like to have a fixed base to work against for the duration of this game.  On to “the rules”.

Gold Cost to Buy Tiles

It’s been observed that gold cost goes up the more techs and civics are researched.  After looking at some saves this formula has turned up yet again:

FLOOR(BaseCost * (1 + Multiplier * FLOOR(100 * MAX(CompletedTechs / 67, CompletedCivics / 50))/100))

For 2nd ring tiles the BaseCost = 50 while for 3rd ring tiles BaseCost = 75.  In both cases the value for Multiplier is 4.

What Tile is Being Picked

Calculate the “tile influence” as follows.  The tile with the lowest score is the next tile to be picked.  If multiple tiles share the lowest score the picker will select randomly each turn.

1) Find the distance to the tile from the city center and multiply by 100.  If there is an impassible tile between the target tile and the city center the distance is the shortest number of tiles to get to the target.
2) Add 100 if the tile is beyond the 5th ring.
3) If the tile is a water tile (coast, lake, ocean) add 25.
4) If the tile contains a barbarian camp add 100.
5) If the tile contains an improvement subtract 5.
6) If the tile contains a natural wonder or a resource subtract 105.
7) If the resource is a bonus resource, add 1.
8) Add up the total yields of the tile and subtract the total.  All yields are equal, so that a 4 tile has the same yield as a 2 / 2 tile (both are yield 4).
9) Count the number of tiles that are adjacent to the target tile, are in the 2nd or 3rd ring of the city but not yet claimed and contain a strategic resource or a luxury resource.  Subtract this count from the score.
10) Count the number of tiles that are adjacent to the target tile but not yet claimed and contain a natural wonder.  Subtract this count from the score.
11) Add the terrain influence modifier: 1 for grassland, plains and ocean, 2 for desert, tundra, snow and coast, 3 for mountains
12) Add 1000 if the target tile is not adjacent to any tile already owned by the city in question.

Note that as tiles are claimed by the city the scores may vary due to changes in the value of Step 9.  

I’ll be scoring all of my cities 2nd ring tiles using the above method and we’ll compare the scored results vs. the actual results, tile buys notwithstanding.  When a tile gets rescored because of expansion I’ll note the tile’s new score in the turn report (and update the score in the post with the original scoring).

Here’s Irukandji scored using this method as of my end of Turn 9.  As of right now there are no resources adjacent to the southernmost tiles visible in this picture.



If the game or the city goes long enough there could be a test of the “conventional wisdom” that all 2nd ring tiles are taken before all 3rd ring tiles.  The scoring suggests that the 3rd ring citrus should be taken after all other 2nd ring resource tiles are claimed but before any other 2nd ring tiles.  

I will keep track of successes and failures of these rules in edits below as the game progresses.
-----------------------------------------------------------------------------------------------------------------------------------
T10: Failure.  The scoring picked the NW citrus as the next tile.  The tile picker in-game is taking the wheat tile.
T11: Success?  The picker swapped from the wheat to the SE rice, both of which scored identically.
T13: Success?: The picker seems to have scored both rice tiles and the wheat tile identically (it's flip-flopping between the three)
T20: Success?: Despite never showing the tile in the GUI Irukandji culturally expanded to the NW citrus as predicted by the scoring.  The GUI is now showing the SW rice tile.  As long as that continues to flip-flop amongst the three bonus food tiles all is well.  Cassowary's scoring is here.
T38: Wheat tile at Irukandji is picked - picker was flipping between two rice tiles and the wheat, bought one rice and it grabbed the wheat. Picker has settled on the remaining rice tile. After that it should go for the 1/1 plains tile that is adjacent to the horses and jade (scored 197). If the resource adjacency bonus only counts once then it'll be flip-flopping between that tile and the four tiles that scored 198.
T45: Cassowary takes the iron tile (win). Its next tile is the 1/1 plains NW of jade. The tile should is scored at 199 once the jade is controlled. I have to check if there's flip-flopping because there's three tiles at 198 if the jade matters, one at 198 and two at 199 if it doesn't.
T50: Founded Mouse Spider and scored it out. Should go citrus->citrus->rice (if available). Cassowary is dead set on the same tile NW of the jade and has not flip-flopped.
T51: Citrus at Mouse Spider should be scored one lower. Tile picker has selected the western unclaimed citrus.
T56: Irukandji and Mouse Spider both expand to the expected tiles and select the remaining lowest scoring tile for the next expansion.
T60: Blue-Ringed Octopus founded & scored. Fish tile should be first.
T61: Blue-Ringed Octopus is choosing the fish first.
T65: Mouse Spider has grabbed all citrus, now choosing the 1/2 plains to the east.

[Cornflakes]

I've seen several cases where a 2nd ring water tile with a resource is either picked over a 2nd ring land tile with a resource. According to your list, it looks like the land tile would always beat out the water tile due to the +25 to water tiles (for an example see my round 2 duel vs. TheArchduke for an example where both 2nd ring Fish were picked over 2nd ring Rice).

[suboptimal]

I've seen several cases where a 2nd ring water tile with a resource is either picked over a 2nd ring land tile with a resource. According to your list, it looks like the land tile would always beat out the water tile due to the +25 to water tiles (for an example see my round 2 duel vs. TheArchduke for an example where both 2nd ring Fish were picked over 2nd ring Rice).

Hmmm...interesting.  I'll need to go take a look at that game.

[suboptimal]

Turn 10

Start the turn, Pottery is complete.  Start Writing.  I see in the score that Emperor K and The Archduke both completed a tech during Turn 9 so they’re running slightly ahead in science.  In the north the scout heads for the hills, literally, but doesn’t quite get there.  After moving 1 tile NE there are religious city-state borders in the distance.  At this point I was going to move E onto the hill top to see the desert but instead move onto the coffee tile.  I meet Kandy but don’t get the first-meet bonus.  They would like me to train an archer.  The lack of the first-meet envoy confirms my suspicions about our current layout and also means that my northern opponent has been scouting in this direction.



In the west the warrior moves onto the stone and has its line of sight to the probably city center blocked by mountains.  That’s going to take a couple more turns to get sorted.  
-----
In terms of the tile picker test I’ve had the first failure.  The tile picker is taking the wheat (a bonus resource on a 2 / 1 tile) over the NW citrus.

[suboptimal]

Turn 11

After opening the save I took a very close look around in the west to see if I could make out any borders hiding in the fog.  Sure enough, I could just make out the red dotted line of a military city-state’s borders.  The warrior crosses the river and finds Valetta.  I get the first meet bonus, they’d like me to send a trade route and I get the inspiration for Political Philosophy.  



Interestingly I find them even though I can’t see the city center directly.  I think that has to do with the fact that I’m adjacent to their borders.  The warrior will begin making its way back towards Irukandji next turn.  Up in the north the scout heads northeast.  The primary objective is for the scout to find my opponent.

With the extra production bonus from Valetta the settler is now due in 5 turns, along with the population.  Once the settler is complete I’ll be able to switch to working one of the citrus tiles for slightly faster growth while maintaining a good build pace.  I’ll probably build a military unit or two before getting a Builder out since I’m delaying Animal Husbandry and Mining to keep the campus costs down.  

The settler will take 8 turns to get to and settle Red Dot.  Once settled I’ll buy the closest horse tile and build my first Builder there.

Internationally Archduke has grown another population.  

After some thought I’ve redone the northern dot map by moving Blue Dot to a coastal location closer to Red Dot.  This puts the Campus tile in the first ring and gives me the option of building the harbor in its original spot (still marked on the map) or adjacent to the putative Red Dot harbor.  It also lets me fit in Green Dot further up the coast.  That will serve as the coastal chokepoint against any northern rush.  The first military unit out of Irukandji, if not chasing barbarians, will scout the eastern side of that mountain range to see what’s there.



One more note – the tile picker at Irukandji has now flip-flopped to the SE rice.    I had the two rice tiles and the wheat scored identically (93) but the citrus scored lower.  In some regards, this makes sense.  In others, it doesn't.