Test games played

[Kaiku]

I noticed when I saw them smashing Earth Elementals with ease. I've made peace with the Sorcery wizard, which is nice as now I don't need to defend my new cities (his old) as badly. So now's a good time to go around the map and take those neutrals.

The rewards from lairs are a bit nuts, I feel. I had massive resource issues and was barely scraping by. Took one lair and I'm set for the next dozen turns or so.

[Kaiku]

1406. Went to war with another wizard, Sorcery and Life. Razed most of his cities, but he (also) has several doomstacks of Focus Magic'ed Naga, which are very hard to deal with as they pick off my Berzerkers like they're Spearmen. Luckily he doesn't attack my Berzerker stacked cities, as he could conquer them easily. But on the positive side, I got The Black Knight. Didn't knew it was possible to get champions so early. I have no items for him, but with a couple of buffs he's nearly immune to everything they're throwing at me already. Don't expect the rest of this game will be very hard with him. My casting skill is absolutely pathetic, though.

[Seravy]

Started a game as well.
5 Death 5 Sorcery Sage Master Omniscient High Men.

Early game is a lot more interesting in the current system. I was able to juggle goals of having settlers, units and research without losing anything critical. I did not summon max creatures, and got away with it despite playing high Men. In fact I only ever summoned 1 Naga and some Skeletons until I got to Shadow Demons in research.

Land for settlers runs out quickly at least on Fair land. Not surprising, if we assume there is room for 25 cities on the map, divided by 4 players, it's only 6.25 cities on average but the real number is around 5 because Neutrals also take up space. This however is merely an average, so if starting closer to the AI, it can easily go down to 3-4 only. With 2 starting settlers, that means you barely have room for 1-2 more cities. This puts settlers into a "raze and replace" role, assuming you need more cities of your own race.

Not ideal, I'm not sure I like all free land being already claimed in 1403. It's definitely an improvement over the previous system though. Unfortunately the AI can't play with higher settler costs (or few cities in general) so it probably can't be helped. Fortunately, razing cities and rebuilding is perfectly viable this early.

I do have one interesting (but probably stupid) idea how this sort of a "settler spam" gameplay can be avoided.
As is, settlers have an exponential return - the earlier you plant them the more you get out of the city they grow into. Thus, building settlers later is not an option - a pop 8 city with barely any good buildings is of no value compared to a pop 16 metropolis with a wizard's guild and amplifying tower and armorer's guild.
So...there should be a mechanic that penalizes early cities. Preferably in a transparent, intuitive, easy to understand way. If building cities 50 turns earlier is not more beneficial, the AI doesn't have to force spamming settlers above all else, and can be set to do, or don't do it based on personality. Obviously they'd still need to start doing it eventually, but it'd be a choice of doing it on turn 1 or turn 40 or 60.
The only such mechanic I can think of is growth rate. Specifically, outpost and population growth rate needs to be lower based on turn count, in a way that an outpost planted on turn 0 and turn 50 grow to roughly similar size by turn 150.
I see two major problem with this approach. One is, as stated, the AI plays poorly from fewer cities. However by turn 50 the player is likely in a position to intercept and kill all AI settlers or outposts and prevent them from building any. So in the end the AI still ends up losing the game if not building them earlier. Two, there is no way "population growth is slower in the earlier turns" make any sense whatsoever for this game. It might for Civ where one can get away saying "technology is more advanced so you grow faster" but not here.
So in the end this is just an impossible idea.

Back on topic of the game.
I'm satisfied by the 7 figure skeletons. They actually saved my city from being taken over by AI horsebowmen. Also, the AI is now diversifying their forced building by race makes for more interesting games. At the very least this is the first time I get a chance to play against larger stacks of horsebowmen.
I'm also very happy with Crusaders. They are much better than it looks like. Not only are they far better than priests (I tried, the priest I build didn't survive the first combat) because they actually live long enough to cast their healing spells and even fight enemies, but also because they can heal themselves or each other, so not only do they tank ranged damage but can even keep refilling the health of whichever unit is targeted. And the best part, if you fight weaker enemy stacks, you can even use the heals to make sure the stack is at full health at the end of combat. This is pretty much the first time we have a cheap enough healing unit that also has the durability to serve in combat.
I haven't produced any Knights yet - so I'm not sure if both new units are needed, or one of them is enough to fill the gap we had.

Adding the "No positive AI diplomacy until turn 40" seems to have worked fairly well. Two wizards have a pact with each other while both were at a war with the third one, albeit one of the wars already stopped and it allowed that wizard to switch over to being at war with me instead. Meanwhile the army power of the one with barbarians already skyrocketed, but I happen to have a wizard's pact with them and will likely have something from Death to deal with the Berserkers before I have to fight them. They'll probably hit every lair and node before I get to doing it though. Fortunately, my Shadow Demons are excellent at lairs/nodes and can hunt both planes so I can at least count on them getting some. I had no early node this game but the demons did get a Sorcery node in 1405 for me.

[Seravy]

Still thinking about AI settlers and wondering if it would be better if the (Arcanus) AI produced them in two waves, one to cover the "better" half of the map, then a short gap to produce buildings and some armies, and then more settlers to fill the rest.

While cities are best when built ASAP, those pop 6-8 spots aren't all that great even for the AI to really prioritize them over economy buildings in the first wave of cities, and this both allows the human to get slightly more use for settlers if you want to, and have more cities of their own race, while at the same time not providing them with a major advantage, effectively letting them pick between the overextending and the normal role instead of the AI hogging the overextending one every single time.

The problem with this would be that if landmass is uneven, whoever got the "big" continent will fail to take advantage of it (albeit they might be in a better position to defend it once the second wave of settlers come, by then the leftover settlers from others might steal the territory.), although that might actually result in more even distribution of resources between the 4 AI players, which cuts down excessively powerful AI starts.

The other problem is, it's ultra hard to come up with a formula for the timing of the "gap".

So yeah, just some brainstorming, it's probably also a bad idea.

Or...there was this old suggestion "cities smaller than X are auto razed". If people can't take over pop 2-3 settlements, early wars would suddenly result in a lot more space to fill with settlers. The downside to this one? Obviously the player can just decide to wait until the city reaches the size X and get it then. So it's not really helping and encouraging such waiting is bad. (It's already bad enough that I tend to wait for the AI to finish the sawmill when playing high difficulty)

[Seravy]

So, this is the current game I was playing on Expert.

   

Idk, either having 10 cities is 1410 is really that poor performancy unworth of this level of difficulty (I don't think so, while I didn't have a good start, I eliminated a wizard entirely and own all their cities)...or the difficulty levels are still not properly balanced (Or, not properly balanced again...we did do minor AI improvements even since 5.0 was first release)

I'd expect this kind of AI progress on Master, not Expert, or am I wrong in thinking that should be the normal?

Note that purple is Barbarians, so keeping up with Green, the Myrran wizard is expected. Both being over 3 times my power despite me holding two wizard's worth of territory...is not.

It might be possible the 2 free settlers actually helped the AI more than the player despite common sense saying it helps the player? I wonder. Either way, I added one more red mark for Expert in my test results. So far that's 4 red, 3 yellow and 5 green so...leaning towards that it's somewhat too hard.

I believe we'll need to start over AI testing after we do the income and research adjustments though. It'll have a pretty large influence, especially as it makes the strategy the player uses to win, pushing research, less effective.

...and now for something entirely different. I'm going to play a few games on Easy with 1 opponent to gather data on expected power/research incomes for various turns, that way I don't need to worry about the AI interference. I'll pretend there are enemies and keep reasonable garrisons instead of going all out on economy, and will not overextend. I'll use random wizards but will try to play well. I think this is as close as we can get to gain this information without spending weeks on fighting real AI opponents.

[Nelphine]

In not sure. Barbarians can easily sextuple your strategic strength, even without buffs. Since overall is a combination of 3 factors, if one of those factors is 6 times higher than normal, then being twice as powerful is not unexpected.

10 cities to me sounds low - I would expect 35-40 cities, so I would expect you to have 15-20 cities if you have the territory of two wizards.

So to me you actually have the territory of one wizard.

Now the myrran AI could actually have the territory of 4 wizards, plus inherent better power and strategic strength frok being myrran. Call it 5 times the expected power of 1 wizard. 

However, 1410, he won't have it all yet, so 1/2-2/3 sounds right (3 times as powerful us 3/5).

So myrran guy looks right.

What about barbaruans? Well, I've already decided twice as strong is reasonable just on having bezerkers. But those bezerjers don't just stand and look pretty - they go and do things, and since they are strong, they can do 6 times as much in terms of conquering.

On average, I would say barbarian AI have 3-6 times many lairs and nodes, as well as a higher number of neutrals and conquered enemy AI cities. That should be about 3 times as much power production as a similar AI.

That's 1.5 times as powerful, and that multiplies alongside the strategic strength.

So 3 times an individual AI is exactly what I'd expect.

But wait, 3 times as much power production is actually too high - barbarian cities suck for power. However, this is negated because the higher power production also results on higher spell power.

So 3 times as strong as an average player is about what I expect for a barbarian AI. (A mono life barbarian AI would be about 50% stronger on top of that.)

[Seravy]

Well the game is already a loss so I looked at the map.
Purple has 9 cities, less than me but barbarians are barbarians. Them being triple my strength is actually ok.
Red however has 8 cities, two less than me. And we both were at war with the yellow player, but I got all of their territory while red got none - this is how cities ended up 10-8 in my favor despite a corner start. They have no business being twice as strong as I am on this difficulty level, or am I wrong? For Master, and Lunatic, definitely, but on expert an AI with the same empire size shouldn't be twice as powerful?
Speaking of which, Expert AI with Gorgons in 1410 and five nodes? They are playing Orcs, probably the bottom tier race for AI as their strategic strength is pretty weak too. I admit Life+Nature are decent at raising the graph with summons buffs and just generally having pop 25 cities but still...

Green has 15 cities - only colonized half the plane. That, and being Dark Elf, so she is only at like, a third of her final power.
But the question is not "Is the Myrran wizard as powerful as the territory implies", but "Is the Myrran wizard as powerful as what the player can reasonably beat on this difficulty level?". If not, we need to revise Myrran AI specific mechanics to fix it.

[Nelphine]

That sounds like the AI found Gorgon's in treasure, and went crazy with it due to strategic strength of Gorgon's. Unusual indeed.

What's the power bonus on expert? 160%? If theyve got 8 cities better developed than your 10 (for instance all 8 have amp towers), then yeah, they'd be twice as strong at expert.

That's on of the problems I have, that the overall graph really had no relevance on how strong an AI really is.

[Seravy]

No, I checked, they finished all uncommons already, and it's their first pick in rares.
The Myrran wizard has Angels, probably in the same way but maybe that was found, didn't check there. They already summoned 14 of them which reminds me of Lunatic, not Expert.

Power bonus is 180%. It's high to support the combat MP spending, mainly. The AI still spends several times as much as the human in battles, even if unnecessary. I don't think we can have that significantly lower but we might be able to shave off a few % over time.
Except the Myrran wizard has no such combat spending. So they have yet another major advantage over all others on top of no competition for territory and the Myrran extras - twice the normal budget for SP/RP spending.

[Nelphine]

Not that high. They still have lairs to deal with, which is notociably relevant until late uncommon.

Still higher than arcanus wizards at war, but lower than arcanus wizard at peace.