Test games played

[Nelphine]

Personally, I don't care about the time it takes to raze the human cities. The human can design a doomstack that will raze an AI city every turn, and never take casualties (easier with some realms, cough magic immunity, than others). The human also can actively decide to raze every single city. So the human can raze all of Myrror in about.. 40 turns, even if it has 35 cities.

On the other hand, the AI won't always go to the nearest city with their doomstack. They may waste time attacking other stacks (although that's unusual). They will generally never build a doomstack immune to damage, unless they happen to get lucky with regenerating creatures but even that is extremely difficult. And they won't raze EVERY city.

Evem assuming the AI manage to get a stack that is immune to attritional damage, I can't imagine the AI wiping out 25 cities in 50 turns. So even with 10 less cities, I think it would take the AI at least 10 turns longer than the human. And it's probably an even bigger gap than that. I would expect on average the AI to take at least 75 turns for 25 cities. Once you account for actual attritional damage, I would expect it to take closer to 100 turns for the AI to take one unstoppable doomstack and destroy 25 cities.

That by itself is a fundamental problem, making AI simply less dangerous than the human. That's why they need the quantitative advantage; they really need to have at least 3 doomstacks just to keep up with one human doomstack.

This is also probably why a 2 way early war feels so deadly in the beginning of the game; there are multiple AI, all building doomstacks, that are all being sent against you.

Unless you can solve that, no I don't particularly think the size difference actually makes much difference. Even if both planes are the same size (call it 30 cities), we expect the human to take 1.14 turns per city, so 34-35 turns to destroy those 30 cities; and the AI will take 4 turns per city, so 120 turns.

[Seravy]

This game the AI conquered or razed like 2-3 of my cities each turn and kept doing that for some turns. Once all cities near the towers were down, things slowed down, to like 1 city each 2-3 turns. However, if I hadn't researched Holy Word to stop it completely, they would have likely been faster.
On the other hand, the AI is indeed not razing every city so while they lose Myrror, they actually gain more cities you have to destroy which is the advantage on their side. So maybe it's good as is? idk.
If I had two doomstacks, I could have been the faster one and two doomstacks are still reasonable (my current doomstack is 3 heroes and 3 stone giants. If I had a bit more time to prepare, two of these would have been trivial)
...also, this lack of units worries me. Is the AI targeting the wrong plane again? They returned to Arcanus so it's not impossible. Oh wait there are no units on the other side to come over. Guess that explains it, this continent is already abandoned, without cities on it they can't have units to send into the tower... oh max farmers everywhere I think the units are already starving to death. Well THAT is a problem more Myrran cities can be relevant in solving. It's a problem if the AI starts to lose stacks 10-15 turns into the war... the ruined cities they conquer and fail to raze are enough to keep them from losing the game, but they don't contribute much to feeding armies...

[Nelphine]

No, I don't actually think that will help the food problem. The time the AI have to build up is such that they will have huge piles of medium tier units lying around. Those units aren't actually dangerous, but they contribute to starvation. The number of units a given city produces, is more likely to be more than that city can help feed later on. The strongest units are the ones that fight, and therefore the ones that die. The ones that starve, are the ones that can't fight, and they can't fight because they aren't relevant. Starvation really doesn't matter to the AI until you're starving units that were created in the same year.

[Seravy]

Starvation matters a LOT. A stack of 6 units is a lot weaker than a stack of 9 units, even if the missing units weren't that great. A stack that drops below 5 units receives a penalty due to being too small and won't be able to attack nearly as much.
Also starvation isn't one unit at a time - a city with 15 farmers are feeding 60 units. So for each such city the AI goes below the threshold, they are losing a large number of units.
You are wrong about how unit production works too - a large, fully built city might contribute enough units to use up its own food or more, but smaller cities still making buildings are not, and these smaller cities are what's missing if the land size is reduced (as all cities have already grown big enough to focus on unit production).
But the real key here is every turn matters. If starvation is pushed back 5 turns due to the AI having 5 extra cities, those 5 turns allow those troops to conquer more stuff on the other side, which will let them feed themselves. If they disband, those cities fail to get conquered.

I think the best solution might be to keep the "+20% size" for the plane which has 3 wizards, but remove the -20% modifier on the other plane. That extra 20% on my side is a major contributing factor why I can't properly finish conquering Arcanus and all the lairs, and break towers first. (same way how larger landmass games are always harder)
This much size difference also favors Arcanus way too much for placing lairs on the types that are random. 50% more overall land means 50% more lairs.

Oh, btw there is a max unit cap. If it's reached cities can't produce more units but they still contribute to avoiding starvation. So at least on higher difficulties that ensures there is a limit to how much useless units the AI can have.

[Nelphine]

I rarely play games that hit the unit max.

However, I disagree. A stack of 9 unbuffed experienced halberdiers literally poses no more threat than a stack with only 3 of them. NO difference. Neither of them are strong enough to attack any human garrison, neither of them are threatening enough for the human to bother attacking them. They literally don't matter.

I would argue that even stacks of 9 naga don't matter by the time the towers break. So whether it's 9 naga or 3 naga, literally makes no difference. 

(Yes if I'm playing my spearmen garrisons that don't even have an alchemists guild or barracks, then yes, 9 naga or halberdiers can still attack. But even with the penalty, the stacks with only 3 such units can still attack because 8 spearmen are THAT POOR. And my combat spells mean I win whether the 9 attack or the 3 attack, AND I use exactly the SAME combat spells whether it's 9 or 3. So it's still the same result.)

So no, starvation doesn't matter until 'modern' units starve; modern meaning units that can actually threaten the human. When towers break, that means units built with a war college, an armored guild, and at least mithril.

[Seravy]

A stack of 9 unbuffed experienced halberdiers literally poses no more threat than a stack with only 3 of them.

9 of them will attack your city and the wizard will kill your units with combat spells, conquering it. 3 of them won't. Even 5 of them likely won't.

Not everyone has the luxury of having berserkers in garrison.

[Nelphine]

No. Either you're playing with strong garrisons (like magicians), and those 9 units don't have enough strategic strength to attack OR you're playing invincible garrisons (like invisibility or supreme light) and you don't lose a single unit when those units attack. Either way, those units do nothing.

Bezerkers aren't used to prevent riff Raff stacks from attacking. They're used to prevent the strongest doomstacjs from attacking. You can stop 9 unbuffered experienced halberduers with literally like 2 elite bezerkers that have holy armor and holy weapon.

[Seravy]

No. Either you're playing with strong garrisons (like magicians), and those 9 units don't have enough strategic strength to attack OR you're playing invincible garrisons (like invisibility or supreme light) and you don't lose a single unit when those units attack. Either way, those units do nothing.

Somehow in all my previous 3-5 games I had neither of those. Players don't have infinite time and resources to build garrisons especially when the Arcanus plane is that much larger. So yes, that's what a player should aim for, but doesn't mean they actually have it.

In particular, Magicians are great but if you have been doing a war with a Chaos wizard, you can forget using them because Fire Storm is a thing.

Maybe I'm somewhat biased as I have been doing nothing but overextending games due to playing Gnolls, but I do think imperfect garrisons are a thing. (In fact the Myrran wizard in question might use the Fire Storm or other nukes themselves. This one in the last game had Call the Void in the last 20 turns of the game. If the AI has a stack next to it when those spells hit, it's theirs. If not, I get to keep it and Stream of Life heals the damaged units.)

[Nelphine]

I suppose that's possible. I found I couldn't even beat expert reliably until I had learned to have one of those two garrison types, and stuck to it. Overextending (for me) has always meant not being able to garrison properly. Its one reason I like life so much, because sometimes heroism by itself was enough, and therefore let me get away with very weak (in terms of cost) garrisons. So I have my own bias too. I'm willing to concede your point, as it appears you understand mine.

[Seravy]

Any recommendations on race/retorts/strategy for a Ghoul game?
10 Death books and Arcanus plane preferred, but I'm ok with 10 Death 2 Sorcery if Focus Magic is important to include. (In this case Difficulty will need to be dropped to Advanced though, otherwise the usual Expert.)