Culture matters
SP's only job should be to maximize its culture... I haven't ran numbers on this but Forge, Levee will probably be worth their cost.
SP will go (?Forge>Levee?>JT?)>Hermitage>Cathedrals. We can start building some fail gold in Hermitage if we want any. (Turfan?)
Also we need a few more missionaries (Hindu, possibly backups for Jewish/Buddhist) and build some 3 temples of each of these religion in other cities. Jewish is already covered...
Spy production
Assuming we can get SP to ~10k culture total (Shaka has 1,000+ accumulated) (for T190 victory, we'll have 33 turns to do this, 9k/33 = ~273 average). With Hermitage, 3 Cathedrals and possibly FS that's +350% multipliers so 100 base culture = 450 cpt for instance so it looks reasonable.
Pyramids already provide 12, religion will add more and each artist will grant 6 base after Sistine is built. I think this is a reasonable level to achieve, but obviously the more we get going here, the less EP we need.(opportunity cost of scientist vs. spy specialist... about 1.5 times more beakers from scientist)
About 10k EP converts to 5k beakers loss in comparison. This isn't too worrying as I think beakers will not be our limiting factor in the end game. So the ~12k we need means 6k beakers, adds 2T to research in total, plus the cost of Jails+IAs (300H, let's say 15 cities another 4.5k beakers)... so we're adding about 3T to research to generate this culture through espionage.
However, we need ~35 successful spy missions to bring SP from 9,000 to 50,000. Success odds are about ~80% for culture missions, so minimum 35/0.8 = 44 spies. In SGOTM16 I used a safety factor of ~1.5 whereas ~80% converts to 1.25. So around 55 spies to be safe. They have a tendency to be caught...
If we have a few cities start building spies when we have some spare hammers, the end game will be much less a panic. Something like 20 spies over the next 20 turns will be great, and then we can rush a bit the last few.
Settlers
As I previously said, we need to build somewhere between 8~12 settlers to bring Joao and Mansa up to +8 status. Gandhi can be left as UN opponent.
After warring with Mansa for the Wine, and hopping out of Religion, our relations will be around -3 or -4. We'll gift him 5 resources for 5 turns to gain +2. OB after 25 turns grant +1. Each 5 techs gifted is +1.. we can get about +4 this way.
-4+2+1+4 = ~3 so for +8 we need to give him 5/1.5 = 4 cities. A cushion from FM for +1 perhaps as well.
For Joao, after current war and outside religion, we'll be around -11 (!) a lot of work here...
+1 from late HR switch
+1 from OB
+2 from resources
+2 from gifts
---
~-5 ouch.
8--5 = 13
13/1.5 = ~9 cities and that's cutting it close. We'd presumably liberate some of his old cities. (same for MM)
It could be just easier to use GK however we don't have much room near his capital so it seems quite difficult. (I see around 6 in his area, plus we can liberate some old cities... needs double check).
So another 1200 beakers gone... still puts us roughly in the 3T more research time frame... now when we will factor in Paratroopers (160H a piece)... maybe around 10 of them and a few bombers, we'll reach the 4th extra turn. This all leads us towards T190 I think.
City improvements
We will re-evaluate next turn for Levees (some cities will not be worthwhile...). Given that there is a lot to build (Levees, Factories, CP, Jail, IA, spies, settlers, Paratroopers), the positive ROI period becomes a lot shorter. Let's cut off beaker multipliers at 10T ROI, big max. (cost/beaker bonus < 10 or bust).
Exception: Synan is fine as it doesn't get to build a Levee. etc.
City growth
Ultimately, all cities should be near the point of starvation once we win to maximize research, EP, hammers... e.g. coastal tiles are very weak and we should avoid them once we've whipped the production boosts.
Control of the water
Honestly, this seems like a really minor point at this time. Caravels don't pillage so we must only sink galley/trireme if they pop up. Capture of Roskilde seems not that advantageous anymore... about 14C base and maybe 30b/turn. We'd get more from going after Gandhi instead.
Worker improvements
Once unimproved tiles are correctly done, pre-improve all riverside tiles (left 1T from completion). When we hit State Property (or just about when), then switch all of them in quick fashion.
Shaka
The faster we finish him, the faster we can swing our troops vs. Joao. However, if we wind up using GK after all (needs testing) then we will just move to Gandhi right away. Or GM... really not sure.
~~~
Sorry it's all in-organized ... just writing stuff as I think about them. Food for thought.