[SPOILERS - NO PLAYERS] The Final Clusterfuck: The PB38 Map and Lurker Thread

[Erwin]

I ran the map through Novice's tool. Here is some analysis of the starting positions based on Novice's tool and my gut feeling. I did ignore strategic and luxury resources this time, they're easy to adjust later on. In the land and food tiles count, islands haven't been taken into account.

Very bad:
Start H: 128 usable land tiles, 244 food, 5 neighbours, bad island access
Start V: 110 usable land tiles, 247 food, 4 neighbours, good island access
Start W: 133 usable land tiles, 237 food, 4 neighbours, mediocre island access

Bad:
Start Q: 156 usable land tiles, 305 food, 5 neighbours, bad island access
Start C: 130 usable land tiles, 216 food, 3 neighbours, mediocre island access
Start B: 108 usable land tiles, 210 food, 2 neighbours, good island access
Start X: 121 usable land tiles, 220 food, 3 neighbours, good island access
Start O: 128 usable land tiles, 232 food, 3 neighbours, good island access

Below average:
Start U: 143 usable land tiles, 301 food, 3 neighbours, bad island access
Start D: 132 usable land tiles, 279 food, 3 neighbours, good island access

Average:
Start M: 165 usable land tiles, 296 food, 4 neighbours, mediocre island access
Start R: 154 usable land tiles, 278 food, 3 neighbours, mediocre island access
Start A: 144 usable land tiles, 258 food, 2 neighbours, mediocre island access

Above average:
Start J: 157 usable land tiles, 292 food, 3 neighbours, mediocre island access
Start T: 178 usable land tiles, 385 food, 4 neighbours, bad island access
Start F: 174 usable land tiles, 350 food, 3 neighbours, bad island access

Good:
Start I: 176 usable land tiles, 337 food, 3 neighbours, good island access
Start S: 172 usable land tiles, 337 food, 3 neighbours, good island access
Start P: 189 usable land tiles, 361 food, 4 neighbours, good island access
Start G: 170 usable land tiles, 313 food, 3 neighbours, good island access
Start L: 185 usable land tiles, 326 food, 4 neighbours, good island access
Start N: 188 usable land tiles, 362 food, 4 neighbours, good island access

Very good:
Start E: 206 usable land tiles, 450 food, 3 neighbours, bad island access
Start K: 220 usable land tiles, 381 food, 4 neighbours, mediocre island access
Start Y: 190 usable land tiles, 359 food, 2 neighbours, good island access

The output from novice's script (without islands and coast):
https://pastebin.com/nCxcngFA

[RefSteel]

Thanks for looking at that, Erwin! That also broadly agrees with GJ's tool that the latest move I tried is insufficient to rein K in and gives the land it takes to the wrong parties: See attached table if you'd like to see details. (Hilariously, the forum will allow zips but not rtfs...)

My inclination right now is to go with the original start positions Cornflakes posted and try to improve the balance by modifying the coastline and/or adding a major lake or mountain range or two. (Likely close-ish to K's southwest.) If no one else has a go at it or thinks of something better, I'll try it tonight by editing the base map directly so that if my changes turn out to help, they can be used with different stat positions and/or orientations. Identifying the start locations is still easy without the actual starts because porting signs between maps is trivial in a WBSave file.

[Cornflakes]

If we do something like this:

   

Sinking about 40 tiles to the west of K, and turning the area to the SE into a barren desert would probably go a long way towards cutting down K's land area back to the average. Sure the desert will still get counted as someone's land, but it won't have much impact. B, C, and D will need a little bulge of land to maintain 5 tiles off the coast. 

After making the above changes, we can sprinkle in some additional food at the starts with less land to compensate (4th-5th ring so they can't pick it up with a 1-2 tile capital move). Also if we make the extra food Deer it will require extra tech to bring online and not have as much impact on throwing the early game out of balance at a 2nd city site.

[RefSteel]

Sounds like a good plan. I think I would hold off on any changes to the region marked "desert" at least until the other initial changes (e.g. sinking the land marked "sink," drawing start rivers, adding horses, filling out the mentioned bays, etc.) to the map are made, as they might affect balance in unexpected ways, and it's an easier change to make than to revert.

(Also, looking again at the balance report I just posted, moving K west and L southwest didn't really reduce K's land in any meaningful way, so that reinforces the idea that moving the coast is better than moving the start.)

[RefSteel]

Another note: If we swap starts instead of rotating them, here's how I would do it:

Code:

Pos   BFC   New position
E     G     H
F     P     T
G     X     R
H     F     E
J      Y     O
O     M    G
Q     V     F
R     N     J
S     J      Y
T     H     Q
V     O    W
W    I      S
Y     U     V

[EDIT: The late edit I put here is now in a post below instead.]

[Cornflakes]

Another note:  If we swap starts instead of rotating them, here's how I would do it:

Code:

Pos   BFC   New position
E     G     H
F     P     T
G     X     R
H     F     E
J      Y     O
O     M    G
Q     V     F
R     N     J
S     J      Y
T     H     Q
V     O    W
W    I      S
Y     U     V

This is fine with me. Just to be clear, the intent is to leave the starts oriented as originally shown in the screenshot with no rotations necessary?

[RefSteel]

This is fine with me. Just to be clear, the intent is to leave the starts oriented as originally shown in the screenshot with no rotations necessary?

Yes, exactly. Er, assuming I did it right.

The main advantage of swapping start positions instead of rotating them is that the starts will look more familiar to the players and match their sandboxes. This also avoids certain weird river artifacts caused by rotating a start with Mapcad, but the artifacts are easily correctable of course. The main disadvantage: It's arbitrary rather than random. (Specifically, it's the first satisfactory solution I found to get rivers pointing in useful directions while minimizing the number of starts moved.) I could compile all the reasonable solutions (of which the above is just one) that give all the starts good river facings, and we could randomly select among those solutions, but that would take significantly more time.

Now, for the most part, these starts are just letters and collections of tile configurations to me - there are too many of them for me to keep track of whose is whose; I even had to keep looking up which start corresponds to each letter, in addition to which position corresponds to which BFC, to create the list above. I honestly couldn't tell you which of them belong to favorites like mackoti and TBS, and even with GJ and Dark Savant, all I remember for sure about their starts is that they both have mangroves (and I have no idea off the top of my head who got the other three). Nevertheless, preserving the randomized positions and rotating them appropriately might be better just to ensure there's no possible appearance of bias on my part or the map team's in general. I'm happy either way.

[Cornflakes]

After making the above swaps, I ended up with a few duplicated starts. After deleting the duplicates I ended up with no copies of I, M, N, P, or X. The reason being that the set in the middle column (which is being deleted) is not an exact match for the set in the right column (which is being added).

Was the intent to just randomly assign the "missing" starts to the positions (A, D, K, L, N) which had a duplicated start deleted?

I have no problem with arbitrarily-but-not-randomly locating the starts. As you say, there are too many to keep track of and we're not locating them based on who has which start.

[Cornflakes]

I have a few minutes now. I'm going to lock D and L into positions B and C and incorporate those into the base map. Having two starts on the map will allow it to load without randomly assigning start locations. And those made sense since we need to add some land there. I'll take a stab at deleting coast to the west of K and adding some at D and L.

[RefSteel]

After making the above swaps, I ended up with a few duplicated starts. After deleting the duplicates I ended up with no copies of I, M, N, P, or X. The reason being that the set in the middle column (which is being deleted) is not an exact match for the set in the right column (which is being added).

Ugh, my fault for incompletely labeling the columns.

Column 1 = Original start position
Column 2 = Actual BFC label (e.g. A is player 0 from the 15-starts save, B is player 1 from the same save, etc.)
Column 3 = New start position.

I can get the actual numbers from one of your images upthread actually; sorry for not doing that the first time.