Sullla AGA6: Emperor Testing
v.60
Knowing that a new version was due out soon, I decided to play out a couple of openings on Emperor to see how things went. I first set up a Medium Fractal map and picked random civ, getting Elizabeth and the English. I chased after Hinduism and to my surprise actually got it in that game (3250BC), despite not being a Spiritual civ. I did have a high-commerce tile to work though, some kind of luxury resource that was giving me 3 commerce/turn; no doubt that helped out a lot. Anyway, I played until I had gotten a second city, which I placed to block off all Chinese expansion in my direction (once borders expanded, it would seal off a penninsula entirely). See below:
York grabs a couple resources and after it finishes its obelisk, the Chinese won't be able to pass. The problem, however, is that the Germans are already expanding into my territory (they beelined straight for me, unfortunately) and a barb city has popped up right where I was going to send my next settler. All that jungle certainly isn't helping things either. I rated this start as "difficult" but certainly playable. I also noted that the AI seems to be MUCH faster on Emperor compared to Monarch; frankly, the other civs were destroying me in score (despite getting early cultural expansion from founding a religion) and they had tons of archers and scouts everywhere. The Oracle was also built in 1360BC in this game. Wow. That was... fast. At the time of that picture above, the Germans had 3 cities to my 2 and looked about to found more. This certainly would have been a challenge.
I tried a couple more starts, but the map generator struck out several times. One game gave me a good capital city site on a river - but almost completely surrounded by desert. Not going to try and play out of that on Emperor, no sirree. 
Another start was buried in plains forests with only Calendar-enabled luxury resources in range. That would have been hard too. Then a tundra "unter-start" with furs. Map generation clearly still needs some work.
Finally, I got something playable. Game specs were Emperor, Russia - Catherine, Small/Fractal/everything else Normal. Here was the start:
This isn't fantastic, but it's not bad. Clams on the water, plus horses and pigs after a border expansion. Fresh water on a lake too. I decided to forego pursuing a religion and try to focus on expansion (in Civ3 style for high-difficulty starts) and see how that went. Thus the first research went into Fishing; I would put 5 turns of production into a warrior, then immediately swap to Work Boat to hook up the clams. This would let me get a 4-food tile into play WITHOUT having to stop growth to produce a worker. I don't know if this is good strategy yet, but I decided to give it a shot. Moscow spent the first 5 turns working the clams resource for 3 food/3 commerce (Cathy is Financial). Once Fishing was discovered in 3700BC, Moscow swapped from the partially complete warrior (5 shields invested) to a Work Boat, and swapped from working the clams resource to working a grasslands forest to get 2 shields/turn.
Now my scout popped two huts very early on, getting maps and gold. I found a Roman scout in 3750BC, and met an Indian (Asoka) scout shortly thereafter in 3550BC. More surprising were the religious pop-ups: Buddhism was FIDL [Founded In a Distant Land] in 3650BC, Hinduism FIDL in 3550BC! Both gone before turn 10 of the game! And, just to round things out, Judaism was FIDL in 3100BC. Wow. Emperor is... fast. It was a good thing I didn't chase after an early religion! 
(I probably ended up with one of those "AIs rabidly pursue religion starts here, like in SPGT20. This is fine, so long as it doesn't happen every game. I WAS able to get a religion in my first game, so this probably is ok.)
The Work Boat finished in 3050BC (turn 19), and that got me a 4-food tile. I went back to the warrior to finish it (I still had NO defense up to this point!) and also so that Moscow could grow to size 3 before starting on a settler. It might just be me, but I think that starting a settler at size 2 is probably a strategic blunder, since it will take so very long to reach 100 shields, even counting in surplus food as shields. Anyway, in 2880BC Moscow hit size 3, and the next turn finished the warrior and started on a settler. That was the point of this picture:
Now notice the situation I'm in here. Rome is right there, just 10 tiles away from Moscow. I know this is a Small map - but that still seems a little too close. We appear to be in a sardine can here. But, IF I can get a settler produced early enough, I can put a city exactly 5 tiles west of Moscow and completely block off Caesar from any expansion whatsoever! Of course he might attack me to get through - but it would be interesting regardless to see how the AI approached that situation. So the race was on, my settler was due in 14 turns and it was my fervent hope that I could beat the AI to that site. Now I had gone Work Boat/Warrior/Settler for my start here and timed the build almost perfectly for reaching size 3, so I don't think I possibly could have gotten the settler out any sooner. And yet...
I still lost the race. And it wasn't even close. 
My scout spotted the Roman settler (defended by two archers) in 2680BC. Since it was already well out of the capital, I calculated the turns backwards and estimated that it must have been produced in 2840BC. That's turn 24 of the game - yikes. Well, I based my strategy on the assumption that I could equal the AI's production of the first settler (in Civ3, by sacrificing the early granary you could, with a decent start, get the first settler out at the same time as the AI even on Emperor). Obviously this is no longer the case with Civ4. Still, the fact that the AI was able to produce a settler FOURTEEN turns before me, even when I sacrificed everything in favor of getting the first settler out as soon as possible - that's pretty brutal. Emperor is playing a LOT harder than Monarch right now - but I'll get to that in a minute.
So Antium was founded in 2640BC and I snapped this picture a couple of turns later. I picked out a new site for my second city at about this time; I would put St. Petes on the tile 3 spaces NW of Antium, where it would grab gems, rice, wheat, and should be able to grab the horse from Antium as well. After all, Catherine is Creative, and I expected I could get the cultural lead from that free culture. The settler was built in 2280BC, and I sent him out along with my ONLY defensive unit, my warrior. Moscow stopped to build an archer so I would at least have ONE unit in my capital (again, I sacrificed everything for the first settler).
Then in 2040BC, Antium's borders expanded. My jaw dropped. The city had been there for 15 turns, and it not only built an Obelisk, but also accumulated 10 culture for the border expansion? OK... this was starting to border on the ridiculous. In any case, I managed to found St. Petersburg.
Only to have a barb archer walk up out of the blackness right next to the city. Now if I understand it correctly, barbs can only appear in areas covered by the fog of war on the map (is that right?), so that meant there was exactly ONE tile where that archer could have appeared, the wheat tile before St. Petes was founded. Sure enough, that one tile produced a barb archer at exactly the wrong moment for me. It attacked in 1960BC and my city was captured. That's... some pretty unbelievably bad luck there. Needless to say, I didn't have the heart to play any further than this. (Barb archers this early in the game? Pre-2000BC? This may need some tweaking, because they are extremely rough on the player in the Emperor starts I played.)
This was an absolutely cursed start, heh. The barbs could spawn archers because all of the AI civs started the game with Archery on Emperor difficulty, just as they do in modern Civ4.
Let me try to figure out what took place here in this game. I thought that I had pursued a decent strategy, but it blew up horribly in my face. Part of that was due to my pursuit of an erroneous gameplan (it now appears to me that the player will never, ever be able to produce a settler before the Emperor AI) and some of it was also due to bad luck. But the Emperor AI is really tough to beat right now (which is not necessarily a bad thing), and I wanted to try and figure out why. I need a bigger sample size to come to more definite conclusions, but I think I have some tentative suggestions to make at this point.
Right now, the Monarch AIs can give me a pretty good game but don't stand much of a chance to beat me. I creamed them pretty badly in SPGT19, and was running even with the leaders in a Five-City Challenge on a Large map in SPGT20; in short, they don't really scare me. Yet the Emperor AIs are completely leaving my civ in the dust in all of the short starts I've played. Like, we're not even playing the same game in the early going. I might be able to eek out a come-from-behind win (I haven't had the time to try this yet), but in short, right now Civ4 Emperor is playing a lot like Civ3 Deity. That is, the player falls hugely behind early on and has to scramble to catch up from the back of the pack. Now there's nothing wrong with that necessarily; I have no problem with the difficulties being hard (if anything, we should err on the side of being too hard). And maybe I'm just overlooking some strategies because I haven't played enough games of Civ4 as yet. However, I've been hearing the same comments from a number of others on these boards. Monarch is readily doable, yet Emperor kicks their butts. There does not appear to be a smooth scale on the difficulties; Noble -> Prince barely plays harder at all, but Monarch -> Emperor is a huge jump. To figure out what's going on, I took a look at the XML file on the difficulty levels (Civ4HandicapInfo.xml in the Assets/XML/GameInfo folder). Here's what I found:
I made up this table summarizing many of the most important changes in the different difficulties. I left out some of the information that was redundant, dealt with the frequency of barbs appearing (obviously there are more on the higher difficulties) and the goody hut probabilities. I'll also apologize in advance if I mis-interpreted what some of these categories mean. Most of the categories have a pretty smooth sliding scale of increasing difficulty. The research costs are a good example, with all the techs getting a little more expensive with each level. Maintenance and civics costs get higher, the barb bonus goes down, etc. Two things though really jumped out at me when I looked at this.
1) AI Barb Bonus: Notice that I circled the Chieftan bonus against barbarians. What the XML reveals is that the AI civs get the Chieftan bonus on ALL difficulties. An AI scout will get a 70% bonus against barb animals, to be added on to the 100% bonus it already gets - and that's STILL not counting the terrain bonus! Their scouts will NEVER die to barb animals, EVER, and will stand a good chance against non-barb animals too. Not to mention all the free promotions these units will be getting. In short, the AI civs getting the Chieftain bonus against barbs basically ensures that they will never lose any units to the barbs. This strikes me as profoundly unfair for the player. I LIKE the fact that the barbs are more dangerous, but if the AI never faces any kind of threat from them, if it ONLY applies to the player... well, that just seems patently unfair to me. Why not give the AI the Noble bonuses, which are still significant but present some chance for the unit to lose? I predict that there will be much player frustration over the AI "cheating" against barbs if the game is released in this state.
(And if you guys already tested this and determined that the AI needs this bonus to avoid screwing itself up, my apologies. )
This has never changed: the AI still receives massive bonuses against the barbarians. And it still manages to lose workers and settlers to them on a regular basis!
2) More importantly, look at the red box I put on the table. This highlights the biggest difference between Monarch and Emperor, and is probably why that difficulty is playing so much harder right now. If I understand the titles correctly, "Growth" is the rate at which the AI cities grow, "Train" is the rate at which units are produced, "Construct" is for buildings/wonders, and "Create" is for culture (?) Anyway, taken together these roughly correspond to the "Cost Factor" that determined so much in the Civ3 Editor. Now there is not much of a jump from Noble to Prince (5%), and not much more from Prince to Monarch (5.26%). Yet Monarch IS significantly harder than Prince, but that's because the AI gains a free starting worker on Monarch, NOT because things are significantly cheaper. (By the way, getting a free starting worker is a HUGE advantage for the AI, not quite as strong as a free settler but not far off either.) So the jump on these difficulty levels comes not from the minor AI discounts, but from the AI's free worker on Monarch.
Yet on Emperor, the AI not only retains that very-strong free starting worker, there is a drastic jump in AI discounts, from 90 to 70 in all of the categories highlighted. That's a discount of 22% over the Monarch values - a huge difference compared to those 5% ones! All you have to do is look at the table to see that going 100 -> 95 -> 90 -> 70 represents a gigantic leap over the previous values. The AI starting with 2 archers and 1 explorer, plus the techs that require those units, certainly doesn't make things any easier either, but that should be left in place. The big jump in these AI cost factors is what causes the difficulties to scale up at such an uneven rate.
Now it may well be too early to start fine-tuning the difficulties; in fact, it almost certainly is. For that matter, it may well be a design decision to make Emperor harder than it was in Civ3 - and again, I would be fine with this. I'm just reporting what I'm finding as a tester, and that is that Emperor is playing very tough right now, and that there is NOT a smooth scale on the difficulties at all (right now, I wouldn't want to touch Deity with a 10-foot pole!). Whether we need to start making changes is a design decision for people higher up than me in the ranks to consider. 
This was a real issue for the testing group as Emperor difficulty was much, much harder than Monarch. Our individual testing games had suggested this empirically and then my poking around in the XML revealed that the big culprit was that massive jump from 90 to 70 on the various cost discounts. Note that this was significantly harder than the same XML settings in modern Civ4:
The current version of Emperor has a value of 85 for those AI cost discounts, a value of 80 for Immortal, and a value of 60 for Deity. In other words, Emperor back then was harder than Immortal today! (Immortal and Settler difficulties weren't added until the expansions.) Of course this was also back before Blake's AI improvements added in the expansions but it's still striking how much these numbers were watered down over time to make the game easier. This was also back when the health and happiness caps scaled by difficulty, with the player only getting 3 happiness and 1 health by default on Emperor and Deity, which made the game *MUCH* harder before this feature was removed in the expansions. I always felt that was a mistake and I wish that Civ4 had kept the sliding health/happy scale since it took real mastery of the gameplay to work with such a low happiness cap. Anyway, to wrap this up, Emperor was very difficult at the time and almost no one was playing and winning games there. Deity was completely out of the question and no one was even touching it at the time. Given that all of us were also learning the gameplay and didn't know what we were doing yet, it's not a surprise that most of these games were played on Noble or Monarch.
