Notes about a Killer AI
The Zulu were the runaway AI in this game. They had the most expansion room of any Civ. They built the most Ancient and Middle Ages wonders, increasing their edge. They had a war with America which lasted from around 1000BC to 450AD - I think that helped them by reducing their military upkeep. They had an early GA presumably due to the war with America.
I've now examined them more closely during the 590AD to 840AD interval and learned that:
o They were running very close to four turn research. They learned Communism, Fascism, Steam Power, Industrialization, The Corporation, and Refining during this time, taking five turns for Fascism and learning the others in four turns each.
o They traded for resources and luxuries. They used their tech lead to maintain five or six luxuries and I think to also drain what money they could from the other AIs.
o Their gross city income was around 2900gpt during this time.
o I don't think they were running a deficit. At one point they were earning about 2900gpt gross and the tech they were researching needed 4800 beakers. (For them, for the human it would be 12000 for the same tech.) So they'd need to produce 1200 beakers/turn to get it in 4 turns. They probably had enough libraries and universities to get 1200 beakers from 800gpt or less. That leaves 2100gpt plus whatever gold they drained from other AIs to be used for corruption, maintenance, and unit support. I'd guess that with their huge military they were probably around break-even and did not have to run at a deficit.
o Most of the other Civs had no surplus funds at all. They'd overbuilt their militaries and couldn't afford fast research. The result was that the Zulu were the only AI with positive cash flow and the ability to research and thus their lead just kept increasing.
Do the AIs use Deficit Spending?
Earlier in this thread Arathorn guessed that they do. I've found strong evidence to support that guess.
I examined many saves and didn't find any case where an AI's research rate was under 20%. One case which seems quite clear: at 1270AD Egypt had just learned a tech and was starting a new one. She had 0g on hand and had a gross income of 462gpt. She was in Monarchy, had just 12 cities, and I'm fairly sure she had at least a few hundred units at that date. Her net income after corruption and maintenance could hardly have supported such a military. And nonetheless she ran research at 20% for the next few turns, as far forward as I checked. She was at 0g on hand in each turn as one would expect if she were running a deficit.
Another case I examined also suggests that the AI doesn't often re-evaluate its research capacity. At 810AD the Inca had just learned Communism and had 449g on hand. They started their next research at 20% science. In 820AD they had just 33g on hand. In 830AD they had 0g and stayed that way for many turns, but their science rate remained at 20% (deficit spending.) In 890AD they suddenly had a positive balance after going to war with the Zulu - some gpt deal(s) had been terminated. But they remained at 20% research for many turns, finally learning Fascism in 970AD. At that point they had 1613g on hand and set their science rate to 60% for their next research. It seems clear that they didn't increase their science rate when their income permitted it.
I haven't found a case of this but speculate that the same thing might happen in the other direction - if an AI starts research of a tech at a rate it cannot sustain later on, perhaps it leaves the rate unchanged and does some deficit spending at higher than the 20% rate until finished that tech.
At this point I hypothesize that:
1) The AIs research at a minimum of 20% regardless of whether they can afford it. (High probability hypothesis.)
2) The AIs continue research at the same rate until they finish a tech, even if they started at a high rate which they could initially afford and subsequently can't afford that rate. (More of a guess
)
At any rate, it seems clear that the AIs do sometimes spend at a higher rate than they can afford. It is more likely to happen at higher levels when they can quickly build a large military. In some cases I expect this will get them useful tech at a discount. The map parameters, especially size, will have an effect on the value of this deficit spending.
This may also result in some unintentional abuse of AIs in other game situations. I've often played in a fashion which keeps some AIs at zero gold most of the time. In the past I assumed they'd compensate by adjusting their sliders. But that doesn't seem true. Keeping an AI at zero gold may force the disbanding of something every turn. And that might keep a small AI weak, perhaps having a noticeable effect in early stages of a game.
All this suggests that a possibly easier route to winning a Sid game (easier does not mean easy at this level!
) is to start with a map where most or all Civs are isolated on separate islands. In that situation I expect that all AIs would reach a point (at a guess, in the late Middle Ages) where they can't afford to research at higher than 20%. (And can't even afford that actually, but it seems they'll do it nonetheless.) When the AIs reach that "wall", they should slow down and the human will have a chance to overtake them in research and control the game, enabling a diplomatic or space win without a great amount of conflict.
Signing Off
That's the end of things I intend to report on in this game!
Time for me to dedicate a few days to real life and then see if I can play the current GOTM in time to submit...
.
