Frequent readers will remember my run at a high score game using Rome a while back. That game topped 1.2 million in score, but it turns out that that's peanuts. There's a way to go much higher, pioneered by WastinTime of CivFanatics, who topped 3,100,000.
This report actually comes after several attempts. Some fizzled out after I rushed badly or developed too slowly or missed a key slingshot. One try I played all the way through, which led to my analysis of the score normalization formula. (No, that article didn't come from idle curiosity.) That game came up short of my score goal, but I learned a lot.
The approach is to play on a Big & Small map, which is perfect for many reasons. The civs usually start on a pangaea-type landmass, allowing lots of early conquering and resource gathering. Then the myriad islands come with gobs of fish, clam, and crab resources, which can be used with the Sid's Sushi corporation for massive amounts of food. And the islands themselves allow building up tons of population with tons of cities. The rule about minimum distance between cities doesn't apply across separate islands, and each city occupies a very small number of land tiles towards the domination limit. Stated another way, water tiles are a perfect way to build a high score, since they provide food but don't count towards domination.
There's the setup. I went with a standard map size; it's an ongoing discussion whether the bigger maps can return more score for the longer time taken and reduced corporate productivity per resource. Tropical climate generates more jungle and less ice, and Massive Continents produces the most land and the fastest fertile starts. Tiny Islands are essential to cram in more cities and seafood, and Islands Mixed In causes them to cover the whole map rather than leave a big ocean of dead space.
WastinTime seemed to prefer the Inca for quechua rushing, but I think Darius is a better choice. Organized is much better than Industrious in the long run, and the Apothecary UB is excellent for this. Immortals are quite good enough for rushing even if not quite as fast as quechuas. (You do have to reroll a map if you can't find horses.) The difficulty is Immortal too, no pun intended. Deity does have higher scoring potential, but needs a lot more to go right for early rushing and taking vassals for researching. Immortal is breezily easy on Marathon and the lower tech costs let you research fast enough on your own.
I chose my opponents based on a combination of many factors: won't declare war at Pleased, favorite civics commiserate with those I plan to use, low probability to build military units, a spread across the spectrum of BasePeaceWeight so that they won't all BFF each other, and avoiding problematic traits (Creative and Protective) and UUs (Holkan). The one unusual choice is Qin Shi Huang, who is Protective and can declare war at Pleased; but I took him anyway since the rest of his profile is great: likes Bureaucracy, low BasePeaceWeight (most of the other picked leaders skew high), and low military production.
As with the Rome game, I am playing by CFC Hall of Fame rules, although not with the mod. All victory conditions are enabled, some game options are disallowed (notably Unrestricted Leaders), and the maximum opponent count for a standard size map is 10. I do think that maximum opponents is the way to go: more workers to steal and more start-normalized capitals to conquer. The big downside to more opponents is a tighter domination limit, but that is tolerable. Finally, I noticed that with more opponents I would get a lot more starts with gems, which does make sense. Few civilizations get scattered around the edges of the Big landmass; using more civilizations forces some to occupy the interior where jungle-gems exist and get start-normalized into grassland.
I'm not going to tell you how long I spent rerolling and partially playing some starts, but here's the one that developed into the full game.