Races, Units, Buildings

[Arnuz]

That actually make sense, it's much more important for combat casting.
So making it Overland only might also be a fix after all. However, this'd indirectly buff life and to a lesser extent sorcery. Or, more appropriately, nerf chaos and death.

[Seravy]

If the towers only impact overland skill, then players who want to spam combat spells may need some extra skill investment via magic power.

While I see why you say this, I have no idea how to implement it.
To have separate combat and overland skill, the appropriate amount has to be subtracted from your overland skill when entering combat and setting combat skill. For Spellweavers, that means combat skill=2/3*overland skill, for everyone else, combat skill=overland skill.
If amplifying towers do not add to the total, then it becomes
combat skill = overland skill-7*amplifying towers for normal wizards and
combat skill = 2/3*overland skill -7*amplifying towers for spellweavers.
I don't think there is room to calculate all of this but let's assume I manage to do it through far calling elsewhere.
That still leaves one unsolvable problem : "amplifying towers" means city data. City data is not in the memory during combat and cannot be accessed.

tl;dr, it can't be done so there is no point thinking about it.

...unless, if I stored the Amp Tower bonus somewhere in wizard data...that could work? Like scoring.

Which means if you can trade one of your cities for one or two of their cities, you end up in a much stronger position. 

This actually is a fairly strong argument for raising the minimal rate of destroyed things when conquering. In a way, this even makes sense : if you are about to lose important factories or weapons in a real war, you blow them up so they won't be used against you if you have the chance to do so. Doesn't explain the dead population very well, but hey, these are uncivilized middle ages and we are using monsters to conquer on top of that so...yeah.
I believe the current rate is 20% (but 50% for neutrals). Maybe it would make sense if this rate increased in a similar way as difficulty levels, to counter the advantage AI cities have. Unfortunately, that would need ~50% destruction rates for the highest difficulty level which is ridiculous. (or maybe not...cities can't grow forever, eventually they max out and the human player catches up. The AI also produces relatively more units than humans, and population bonuses are lower than production. So 40% might work for the highest amount.

I would like to add however that as the human player my experience is the AI has much more stacks then me, even if not necessarily better. So if I'm leaving things undefended, I lose cities several times faster than I gain them. The 4 berserkers trick might be what makes it work for you, as you get a "full" garrison for only 400 production, but it's as effective as if it was worth 1000.

As building cost is fixed then the skill by towers might decrease with the number of towers in
-the world
-the plane
-the empire
depending on what you want to achieve.

This would be effectively the same as "3." from my list of options, except instead of getting SP which naturally scales down when converted to skill, we scale it down artificially. In other words it means either it's not worth building them in the late game (I get +2 skill when I could invest my 2000 power to gain 3 or 4 each turn, oh joy) or too powerful early. If neither then the scaling down is so weak it's not worth doing.

If it matters, around half of my amplifying towers in my example of 41 amplifying towers came from cities I conquered from the AI. I really really like waiting until the AI builds amplifying towers before taking the city, although it gets destroyed decently often, so it's not perfect.
...
I find I can't sustain buying amplifying towers

These also seem pretty strong proofs that conquest is the problem, not the building itself.

[Arnuz]

Nah you can also buy the building with raze, pillage, or node and lair money. I see myself buying several when the games is advanced enough, especially in undead games where I get free garrisons and free summons, and so mana and money are less of an important factor than in other kinds of strategies.
In my last ghoul game, advanced enough meant during the second AI fight, so I guess I started around y3.

I initially understood the SP solution to be linked with mana but of course it isn't. After some thought I think I like it, I guess it's not too difficult as there's already a similar mechanism for aether binding. I don't understand the 2000 power comment, once numbers are that high you've won anyway, this mechanism avoids the linearity of wizard towers. Then balancing them means assessing how many SPS they produce. Yes I definitely like it!

[Seravy]

About the amount granted by Amplifying Towers.

At 100 casting skill, a Wizard's Guild will take 140 turns to break even with an Amplifying Tower, if all of it is spent on skill. That's half the duration of the game.
At 200, it takes 280 turns so it'll never be the better option.
In both cases doing so means the next WG takes even longer to be worth it.

However, WG can and will produce further power after that. AT doesn't produce anything from day one, but costs maintenance. So while WG is an investment, AT is a loan. That in itself would make WG more desirable....if the numbers were a bit more favorable. At 140 turns though, I rather take the "loan" as the extra skill will earn me more through the spells I cast. It is the difference of something like being able to cast 2-3 more Streams of Life overall.

It's not surprising : AT is an attempt to provide a more viable alternative to satisfy the linearly increasing need for casting skill with empire size, which the default way of raising skill fails to provide. I don't think there is anything we can do about it, it's like Sawmill.

There was an old suggestion to change how casting skill works, instead of square root we could use a different root. That can either allow not having Amplifying towers, or they could look much less superior to using power.

At the current amount, +100% SP spent results in +41% skill in other words, each time your total investment doubles, your skill is 41% higher.
With a 1.9 root, this would be +44% every time.
With a 1.8 root, it would be +47%.

A wizard with 90000 SP invested would have 300 skill in the first system, 405 in the second and 565 in the last one.
In the first one it costs 201 SP to get from 100 to 101 skill. In the second, it costs 120 SP. In the third it costs 72 only.
As you can see, even a 0.1 difference means a drastic change, so this is a powerful tool, assuming I can even code fractional exponents in asm.
However, this change would not take into account territory size and would just flat increase all skill for everyone,  so it's probably a bad idea. Especially as it affects combat as well.

I don't understand the 2000 power comment, once numbers are that high you've won anyway, this mechanism avoids the linearity of wizard towers.

Not everyone wins the game early. Having 2000 power in 1420 with a Myrran AI who has 2-3 times as much and is quite hard to beat is perfectly normal.

Meanwhile I thought about making AT overland only, and I'm against it, even if I find a way to do it. The current balance of power between troops and casting power is good. Reducing combat casting power would mean troops (summoned or produced) become more relevant which is both a direct nerf to combat spells in general, and reduces the options of the AI greatly (as it will always be less effective at moving and using troops than the human). There is only one area when the AI is (almost) as good as the human player, and that is using combat spells. So this part shouldn't be any less relevant.

Changing the amount of skill, as long as it stays between 5-7 is acceptable, but unsure if needed, especially if conquered buildings break more often.
Reducing Uranus' Blessing (less skill, more power production, less cost) is definitely a good idea.
In my current game I'm finding there is no point casting it, as a 50 turns return of investment on the casting skill (no divine orders to lower the cost) is not worth it as that's generally too long to wait for at the time you have rares, unless you already won in which case the game won't last that long anyway.

[Nelphine]

I'm not sure about 40-50 turns being so long as to irrelevant. Either you're playing super fast, in which you don't do it at all regardless of the amount, or you're not, and 4-5 years isn't actually super long even past rares. Look at this lunatic game - since I'm playing conservatively, its been 5+ years since I got Uranus' blessing and my enemies still have 45+ cities between them.

[Arnuz]

I see, you're taking this from the point of view of the AI, that needs to garrison all its cities. That makes sense.

Oh well, guess we'll keep abusing it. It's a necessity when the AI gets so much cheating resource at high difficulty, I guess.

Too bad, I'd much rather reduce the tricks like this and then reduce the amount of cheating whenever possible. But if it isn't then it isn't, of course.

[Seravy]

There is one thing I didn't consider when comparing the AT and the WG.
Casting skill is not a resource by itself. To use it, you need mana. Mana, which ATs don't produce at all. This is true overland, and many times as much true in battles. It's safe to say you need at least 2-3 mana crystals every turn for each point of casting skill you have.
So the Wizard's Guild is a viable choice instead of an AT if your casting skill is high enough that your income can't keep up with it. Fortunately, there is a way AT can affect income directly, and that is maintenance.
Currently, WG costs 6 each turn, and towers cost 7. Not much of an incentive there, the WG does produce 11 resources total more (+10 power, +1 gold compared to the AT, for now I'm ignoring research, with that the difference is 14.) but 11, or even 50 is not a world shaking amount that is likely to affect decisions significantly, especially for a player who has 1800 gold to spare for buying the AT.

So I propose a maintenance of 5 for WG and 20 for Amplifying Towers. That way, WG has a clear +25 (28) resource per turn advantage making this a choice of 7 skill, or all the mana needed to fuel it (or indirectly, the loss of that mana through gold), and if we consider the numbers, 1400/28=50, get a return of investment in 50 turns on the Wizard's Guild compared to the Amplifier Tower at 100 skill, instead of the current 140 turns. As games typically last longer then 50 turns in the early and midgame, the Wizard's guild (or other economy buildings) are better choices, UNLESS the player can earn a massive advantage through using the extra skill - so it will resemble a true loan, where earning enough to pay the "interest" on it isn't trivial and requires a good comprehension of the game state.

Also, this change would have no detrimental effect on the AI, as it pays less maintenance - so it could remain the same threat as before, which wouldn't be the case if the produced skill was lower.

[Nelphine]

Hmm. I'd be willing to try that. I'd be afraid any city that can't build banks wouldn't be able to afford it without prosperity though. So 20 may be too high. (One major thing is, without alchemy, the player gas alternative choices on where to get mana - specifically nodes - but has no method outside cities and treasures to get gold. So cities still need to be able to support the gold maintenance, even if they don't need to provide the mana.)

[Seravy]

I'm not sure about 40-50 turns being so long as to irrelevant. Either you're playing super fast, in which you don't do it at all regardless of the amount, or you're not, and 4-5 years isn't actually super long even past rares. Look at this lunatic game - since I'm playing conservatively, its been 5+ years since I got Uranus' blessing and my enemies still have 45+ cities between them.

4-5 years isn't that long, yes. But not casting any other spell for several years first, when you are supposed to prepare for a major endgame war, probably means the difference between winning and losing, unless you are playing a combat spell strategy where those extra casting skill points are your "army".
So the cost (300*number of cities) is the real problem, as it would need to return the investment before the endgame war starts to turn profitable. If you skipped casting for 30 turns, then a war breaks out, you lose - return of investment won't happen because you lose the cities. If you used those turns to conjure some Djinn or Storm Giants or Sky Drakes, or made your cities fly or something...then you would not lose.

Typically, "rares" is a critical timing - if you strike before the AI has all their very rares, you are way better off than if you do after, even if "after" you have 50% more casting skill from UB. Depends on what very rares they get obviously, but you can't know in advance...but assuming the AI was already ahead, the stakes are even higher : the difference between an AI who just got the first "Sky Drake" and one who already owns 60 of them, at which point being able to summon your own 50% faster is too late.

[Nelphine]

True. I'm biased by the fact I typically finish bothering with rares around the time the AI first starts using rares. So, to me you do have time to cast a fair number before that end game war breaks out.

I do stand by my suggestion of 5 skill per amp tower (maybe maintenance of 12?) and 4 skill per Uranus' blessing.