Here's my first set of scattered post game thoughts.
Re-reading my early posts
I've reread my initial posts on this thread and some of it is amusing. My favorite is my naive initial hope that I'd make a dent in one rival with about 50 Mounted Warriors. Yeah, right
Early Sid AI expansion
I had the most rapid initial expansion I've ever had, building four towns by 2630BC, 28 turns into the game. I did that by being agricultural, having lots of cattle, and not building any military units at the start. So how does that compare to a Sid AI's start? Here's the map at 2630BC from the replay window:
On watching the replay further there seems to be a strong relationship between available land and eventual AI strength. Here's the replay map at 1000BC:
In the above map you can see that Egypt (yellow) has long since run out of expansion room (she'd nearly filled her island in the previous map at 2630BC) and is the smallest AI at this point. She ends up being the weakest AI. Zululand (dark gray) is already at war with America (blue) and will win the fights near their common border. They subsequently expand to the unclaimed areas to their east and their north and will become the strongest AI. The Inca (red) haven't quite finished expanding to fill their island and will become the second strongest AI. All of the other AIs have recently reached their limit of easy expansion and will end up being the mid-strength AIs.
Monarchy vs. Republic
I've compared these two governments by starting a revolt at a particular date, playing with minimal actions through anarchy, saving the last turn in anarchy, then playing the next turn switching to each government and comparing the result.
Comparison at 310AD. This is shortly after the end of the Golden Age and about mid-way through the war with Mongols. I had one owned luxury at that date and 3 imported ones. War weariness was at stage one with one rival, i.e. would become worse later on.
In Monarchy I used a 10% luxury rate and 5 entertainers. 118 free units were allowed and I was paying 69gpt for additional units. Net income with those settings (and no research) was 162gpt.
Republic needed a 40% luxury rate and 5 entertainers. 73 free units were allowed. After disbanding 25 older units I was paying 178gpt for additional units. Net income with those settings was 93gpt, 69gpt less than Monarchy.
Comparison at 1080AD This is part way through my war with the Zulu. I had four owned luxuries and one imported. I made peace with Zulu and America before running this comparison, to make the anarchy easy to play through. Then compared Republic afterward in peace and at war again with Zulu. I didn't count entertainers in these, just made the trade-off I'd use if playing that situation. I.e. more entertainers in the Republic war case to avoid using an insane luxury rate.
Monarchy: 0% luxury, 26gpt unit support, net income 447gpt.
Republic at peace: 10% luxury, 316gpt unit support, net income 486gpt.
Republic at war: 40% luxury, 316gpt unit support, net income 134gpt, 313gpt less than Monarchy.
It seems that Monarchy was definitely the way to go for this game. The massive unit count I was supporting would've hurt Republic during most of the game.
Communism
I wanted to do a detailed comparison of Communism vs. Monarchy at a couple of stages in the game. But I've hit a problem with that. Doing a good comparison would require playing a fair bit with a different approach. All cities should get Courthouses, and perhaps even Police Stations. And an SPHQ should be built. At the point when I learned Communism, I didn't have two techs which would be required to build SPHQ. And I didn't learn them until late in the game when a comparison is of less interest. (Less because my city count had grown very high, probably higher than I'd have done if in Communism. And the result doesn't matter as much after the tough wars.)
If I'd gone for Communism, the best thing might have been to wait a bit longer to retake the Great Library, until I saw two opponents building Universal Suffrage. Or I could take it at the same date but the play immediately following that would be different. I could begin the switch to Communism in 820AD. After anarchy, I'd immediately start full speed research toward Espionage for a SPHQ. And I'd build Courthouses everywhere and perhaps Police Stations. More Barracks of course. Instead of using funds to purchase armies as I did in Monarchy, I'd build more Cavalry (using the higher total production), and use the higher numbers of them to produce more leaders, and get my armies that way. In the meantime I'd use money first to research to Espionage, then to steal more techs. I can't be sure how things would subsequently play out but I think I'd end up being able to reach the Inca sooner even though I'd have attacked the Zulu later.
Anyway, the comparison I'd love to have would be along that sequence at the point of having Courthouses everywhere and just having built SPHQ. It would take rather a lot of play time to make that comparison!
What I have done is used my 1000AD revolt sequence (used to compare to Republic at that date) to also do a basic comparison of Communism.
Monarchy: 26gpt unit support, net income 447gpt, 521 shields.
Communism: zero unit support, net income 549gpt, 672 shields.
Net income is not much improved in Communism. But the increase in production is nice, about 30%. The number of cities I had at that date was still not very high - over time production would definitely still be rising as the empire grew in Communism. And there'd be a big increase in income and production once Courthouses and SPHQ were built. I don't have a good feel for just how big those increases would be. It does seem clear that Communism would've been very strong.
Placement of the FP in Conquests
I wanted to get a better feel for the difference, in practice, between Forbidden Palace locations closer to vs. farther from the Palace. The results from this test are not absolute in any sense. They're specific to this particular map's geography, and to what improvements I've built in various cities. Still, I think it is useful information.
To test this I started from a save I had at 580AD which had an unused leader. I made peace with everyone, put all cities under governor control, and moved the leader to the home region. I abandoned my original FP, then at 650AD rushed an FP in various cities, played each case forward to 660AD, and noted the results. The following image shows the cities I tried. The Palace is circled in green, the five cities I rushed FP in are circled in red:
At this time I had courthouses in most of my southern cities, including the ones near Arabia (orange.) The newer cities toward Arabia were generally smaller (typically size 5) than the cities in the core region. I think that the five locations I tried for the FP are representative of locations I might actually have used, progressively more distant from the Palace, ranging from distance 4 to 19. In the game I used the second of those locations, Oka, at a distance of 8 from the Palace.
The results were surprising to me.
With no FP, my gross income was 768gpt and corruption was 293gpt, for a net income of 475gpt (before maintenance, unit support, etc. - this is the net gold which can be used for all of those purposes.) Net production was 286spt.
I built an FP at each of the five locations and measured increased commerce as a percentage increase in net income (gross - corruption), and measured increased net production as a percentage. Results, with the FP locations numbered from 1 to 5 in increasing distance from the Palace:
Code:
FPcity +gpt +gpt% +shields +shields%
1 35 7.4% 34 11.9%
2 35 7.4% 34 11.9%
3 42 8.8% 33 11.5%
4 50 10.5% 32 11.2%
5 43 9.1% 34 11.9%
I don't understand the odd discrepancies between gold and shields increase, and why shield increases are larger and more uniform. Marketplaces are probably a factor, and the strengths of various towns in terms of income vs. shields are also probably a factor. I think there must be at least one more factor.
At any rate, two overall conclusions seem possible:
1) The Forbidden Palace isn't providing a huge boost. (It is a lot less than I expected.)
2) The exact placement of the Forbidden Palace is not very important. Building it sooner probably helps one's game more than building it in a better location.
Safe Beachhead Towns
Will the AI attack armies which are blocking access to a town if that is the only way to reach the town? The answer is no.
To test this I went back to my 1525AD save, the time when I first had this thought. Instead of forming my first Funnel of Doom, I moved one Cavalry Army so that the Inca could not reach my beachhead town at all by land, then ended the turn. As I expected, the Inca didn't make any effort to attack the armies surrounding the town.
So it seems that beachheads can indeed be established simply by landing a settler and a few armies, then founding a town and deploying the armies to block all land access to the town. The AI won't attack.
In this game a better approach to getting a high score would probably have been to claim a luxury in this fashion as early as was possible from each of Inca, Aztecs, and China.
How much does the AI fear Armies?
I'm still poking around this and a couple of other issues. A bit of preliminary information on this subject:
It seems that armies are not what the AI is afraid of. It looks like there's a formula used to decide whether the AI will attack a unit in the open which is based on attacker strength, defender strength (including terrain bonuses), attacker HPs, and defender HPs. And this formula seems to result in defender HPs having a greater effect as they get higher.
When a defender's HPs are relatively high (e.g. a 10HP army vs. a 4HP attacker), the attack strength must be high relative to the defense strength before the AI will attack. One consequence of this is what we've seen in the AI's reluctance to attack armies. I suspect there are more consequences including some which may affect modded games (higher HPs) in unexpected ways.
More to come on this...