Do we really need to raze Hangzhou? If we can get Roosevelt to keep Chengdu (or capture it and gift it to him ourselves), then we can leave Hangzhou in place. We should tell Roosevelt to attack Chengdu every turn.
You're absolutely right that we could do it that way. That was my original plan. There are two reasons it's better to do it the other way, imo.
Reason #1.
Capturing and holding New York.
Take a look at the map and think about what Roosy might do after we capture New York, if he has units in Washington, Philadelphia, Xian, and Chengdu! Now bbp could send one of our units to Roosy Land after we capture Guanzhou, but AIs tend to like to stack their armies in newly captured cities (Xian) or their city closest to the enemy (Chengdu) that is often under cultural pressure, or both.
We'll have a relatively small army--our current maces plus leftover trebs plus ~10 new cannons. No time to build lots of defenders, so we don't want to have Roosy behind and in front of us. Furthermore, we'd have to defend New York and Guangzhou and even Beijing.
Note that we'll want to capture New York first because we'll be suffering major war weariness from a long campaign against Chruchill and New York has the Statue of Zeus (+100% war weariness).
Reason #2.
Stealing multiple techs from Roosy fast.
We'll want to steal 4, if not 5 techs from Roosy very quickly, assuming Gandhi doesn't research one or more of them in the meantime: Guilds+Banking+Constitution+Economics+(Corporation)
The odds of our tech-stealing spy mission
failing go up significantly as soon as we stack more than one spy in the target city, which means we only want to station one there at a time. But that means we can only steal 1 tech per city every 6 turns, because it takes 6 turns to put a new spy in there and have him be stationary for 5 turns. So if we use only Trojan City, it will take us 18t to steal 4 techs, 24t to steal 5 techs. Too slow. To compound matters, chances are at least one of those missions will fail anyway...
Now, if we have three such cities, we'll pay somewhat more in distance costs, of course, so we attempt to steal the cheapest techs farthest away, to minimize that damage, but we can steal as many as 3 techs on T0 and hopefully the other two on T+6.
So what we really need is:
- T-6: Three spies stationed
- T-1: Cities settled at least 1t before T0 (for the culture -50%)
- T+0: Confucianism spread
- T+0: One more spy stationed right outside each city with 1 movement available
Does anyone know if the EPP Differential is based on current unspent epp or is it total epp generated the whole game?
Total epp generated the whole game against all AIs. This is well known and
well reported in strategy articles. No need to test it.
Edit: Mitchum, you might ask, why not just gift him Hangzhou and use that as one of the cities to steal tech from? That would work too, true. But a a higher cost, because it already has Buddhism in it, unfortunately, Roosy's religion. For T0, that's an immediate -15% we lose. That's not killer, but then if we want to steal another tech 6t later, then he has now put 6t of culture in there and our culture -50% has dropped significantly, depending on how long we owned it before gifting it in the first place.
-50% * YourCC/(TheirCC+YourCC)
So, -50% * 3/(3+6) = -16% if we own it for 1t.
2t: -50%*6/(6+6) = -25%
3t: -50%*9/(9+6) = -30%
etc.
EDIT2:
Our odds of failing the steal tech mission against Roosy right now:
19% 
With 2 or more spies on the same city tile, our odds of failing the tech steal for each spy increase to
38%
That won't improve much by the time we steal the techs because Roosy has spent so much already and is spending tons still with all his multipliers. In fact, we should do a counter-espionage mission on him asap, and then every 10t thereafter forever.
.
) We paid 2*110g to upgrade 2 swords to maces. A mace costs us 43g to build in Moscow. So we lost 67g per upgrade. 
jk (I know how painful it is)
