Current Impossible Strategy

[Nelphine]

While that would be true if space were unlimited, its not. So for me, if there's a sweet spot you really want, it will develop into a hamlet with a sawmill faster if the AI settles it.

Then it becomes a settler vs a unit strong enough to conquer that hamlet. If the AI is big enough that usually means a unit strong enough to defeat a single swordsmen.

So then, marketplace (or any other building that is non military non amplifying tower) becomes compared against the units that could be produced instead, and what they can do.

And specifically what you are comparing is 'will this gold/power/food allow me to conquer more cities than the unit I could make instead'. Super early game (usually capital only, which is why I build my barbarian cavalry after only building two buildings; even bezerkers I usually don't have anything other than military buildings, and a granary and farmers market as food becomes a problem), units are better. Late game, your maintenance is high enough, and combat distance multiplier great enough that you need huge income (against sharee I was spending ~9000 mana per turn, plus whatever attacks she threw against me, provably ~13000 mana per turn total) - which means you to have been building that up all along.

So while the marketplace/settler comparison will be very interesting I think you must also compare it to a unit capable of conquering weak hamlets. (For instance, a fortress strike, and 6 other small attacks of 1 or 2 bezerkers, all on the same turn, is common for me in the early mid game.)

[Catwalk]

Would you ever build a marketplace instead of a combat unit?

[Nelphine]

Only in preparation for late game (which starts in like.. 1402 - you need to have enough income to defeat people who take more than one fortress strike to defeat) or because I need a farmers market for food.

Realistically, this means most cities build marketplaces not units. Because you can only defeat the first or second wizard without banishing them.

But the AI settles either fast (on hard) or incredibly fast (on impossible). Since space IS limited, that means building a settler isn't necessary, until you need to raze cities to defeat an opponent efficiently, AND you need more cities than you currently have to sustain a war against the next target.

[Nelphine]

I guess one of my key thoughts in reasoning is this. I only need a few* stacks of troops to defeat an enemy. Cities are there to build troops and/or to provide income to sustain attacks.

Since you only need a few stacks of troops, most cities are purely income generation.

Since the first few AI don't need much, you can kill them with one city.

The third AI you need to make sure you end with enough income to challenge the fourth AI.

The fourth AI is the last - as long as you have one city left and he doesn't, you win.

So you can afford to start losing cities against the last AI, and it doesn't matter. Defense doesn't matter. Just raze everything faster than he takes out your cities.

[Nelphine]

Note this is all in the context of some mythical 'best'. I don't really advocate playing this way.

I myself don't even fully play this way, though I come very close on impossible. Imposible is still very hard.

But for comparisons of what you 'should' build I believe this is relevant.

[Nelphine]

Facing an impossible AI, who has cult leader and sagemaster. Luckily no life or sorcery. 1410, he has about 230 casting skill and has around 700 power production. He's already finished researching at least 2 very rares, including clairvoyance. And he has doom mastery and survival instinct. If he gets any of the nasty chaos globals, I am going to have troubles, since I'm still trying to clean up the other 2 arcanus wizards (one of whom is a gnoll warlord tactician. His army strength is immense.) I haven't met the myrran wizard yet, but he's cast just cause.

Edit: for reference I'm still researching summon hero. I'm probably going to be facing very rare summons with my bezerkers soon.

[Arnuz]

I've noticed that each strategy seems to include omniscient. As it is now, I think it's a must-have and therefore OP. The problem is that it's front-loaded: with it and 5 books you get half the effect of several other retorts, including a cost 2 one, plus some additional production on top. The few books/many retorts approach, already the most powerful one, benefits the most from it.
I would transform it to have the bonus be more gradual - ideally, without any front-loading and only have a per book bonus.

[Nelphine]

No, I take omniscient for the production bonus. Its not OP, its that nothing else offers production bonuses.

For all the other things you're forced to spend more picks. For instance, if you want +16% power, you need a chaos book. If you want +2 pop (which is also unique, but I don't even take it in my strategy) you need to spend a pick on a nature book. Spending 1 pick on 16% power is squarely in line with other 1 pick choices such as specialist, archmsge, sagemaster, sorllweaver, channeler. In fact, if you consider +16% research compared to sagemaster, the omniscient bonuses seem weak.

Instead its rewarding you for taking multiple realms, which otherwise isn't all that great.

It does somewhat become good because cherry picking common spells that win you the game is a real thing, and omniscient then let's you get an extra bonus for doing so - but that's not a problem with omniscient, that's a problem with common spells being extremely good.

However I think alchemy is still far more important for impossible than omniscient.

(Also on another thread I asked if warlord is necessary for impossible. I've proved my strategy is workable, albeit difficult without warlord, plus your story of ghouls wouldn't need warlord. So, hurrah!)

[Seravy]

The point of this retort is to improve 4+ realm wizards, which are otherwise near unplayable.
Even with the retort, you are still trading long term power (rares/very rares) for a better early game (more commons and better economy) and additional diversity, which is a worse strategy now that rare creatures (and other spells) cost less to use, and AI wizards can still cast in combat while banished (so you can't get away with having no rare combat spells by preventing the enemy from using theirs anymore).

It (and other early retorts) appears this much on Impossible strategies for two reasons :
-You are very unlikely to survive the early game if you don't put all your picks into improving that, and due to the large AI bonus values, snowballing effect is amplified, so the early game is more important
-Winning is unlikely overall, so a high risk-high reward strategy that might sometimes provide an outstanding advantage is a better choice than a strategy that has less risk but always ends up just a bit short of winning. In other words, it's worth taking the risk of not getting the uncommon/rare spells needed to reliably win, in exchange of having a chance to win at all. On Hard, I wouldn't use this tactic, more books to ensure I have a robust list of rare+ spells that I can use to win works better.

[Catwalk]

Sagemaster and +16% from Sorcery is a bad comparison, as the Sorcery book does indeed provide a worse bonus than other books. +16% production, +16% power or +16% gold is far better than +16% research. Omniscient shouldn't use the same numbers for all books IMO.

Omniscient is powerful because it lets you ignore magic and focus on settling and spamming units. Only a few other retorts do that, and they're all powerful. I like Omniscient, but it's not balanced.