Tactics for Rampart.

I see. I'm just curious how averaged this variant is?

Try taking the crypt with the troops and formation mentioned above using that scheme, and you'll take it 10 times out of 10 if you do everything right. The difference is +/- a few centaurs.

Most likely so, but how on the FIRST day can you determine 100% where to go so as not to lose tempo. Because if you run far, you might hit a strong block or a GO on the next turn, and if you clear the weak ones near the castle with your army,

There is no clear answer here. Each situation has its own priorities. In the first three days, try to send as many heroes as possible along the paths in different directions so they find the end (of the paths, that is), which will be either a GO, a passage to a secondary town, or some kind of block that's not worth going to at all yet. Moreover, scouts sent in different directions along paths should, if possible, not deviate even for chests, because later it will turn out they are a couple of tiles short and won't reveal some tasty treat. Or vice versa: later it will turn out there's nowhere to go, but you can no longer restart. Then, based on what's been scouted, we decide where to go with the troops to break through what we can handle. I try to break through to neighboring respawns as quickly as possible, capture a secondary town, buy some low-tier units there so they can start collecting free resources and scouting the territory. Besides, the more towns, the more money :). Not to mention if the secondary town turns out to be of your own faction)). Along the way, I clear money-generating buildings and grab resources.
This is just my imho, I'm not claiming I do everything right, and besides, this is just in a few words. If I were to describe it in detail for each individual case, it would be a whole book :)

What if the crypt is far from the castle?

If the crypt is far from the road and there's nothing else worth paying attention to, then there's no point in going there.

I agree about minimal stacks, but not entirely about minimal losses. Sometimes you attack fast monsters (like wolves, gargoyles, sprites) several times, you take hits on your main centaur stack and then you have nothing left to cover them with. And what if it's only the first few days?! You can't do much with just elves, and spending tempo on bringing dwarves, plus their speed (they're slow) might end up costing more!?

If there are no centaur external dwellings, they usually don't survive to the second week in enough numbers to be a main hitting stack. Their job is to cover the elves, taking hits themselves (if there's no other way), and to finish off what the elves didn't shoot. If you're planning to break through something very strong, then dwarves and even dendroids are indispensable helpers. It's clear they're slow, but they are only needed for a serious fight, and after it (if they survive :)), they are given to a squire, while the main force moves on with centaurs and elves.
And if I manage to take a maximum crypt with the losses mentioned above, I'll be as happy as a child, because in reality it doesn't always work out so well. You take the crypt, earn 5k gold, buy two heroes with it, and they'll collect another 5k each for you. Business :). Plus, they might even find a centaur external dwelling, from which you can replenish your losses. But if you save every single centaur, afraid to attack crypts, there'll be no benefit from it. "A rolling stone gathers no moss" :)
Upgraded sprites, upgraded wolves, and other unpleasant creatures are best left untouched if possible, but if there's no other way, there's a tactic for that (attacking them without it is like flipping a coin: morale in the first round means a wipe or heavy losses). You corner the elves with two evenly split centaur stacks (or centaurs+dwarves), and surround them with single-unit stacks that will take the first hit, then shoot at point-blank range and chop with centaurs, while not forgetting that you still need to cover the elves.
You can go on asking and telling, telling and asking like this for a long time. Theory is good, but it doesn't compare to practice. You'd better learn to play in real games against real opponents. You'll figure things out faster. For example, we have Loi playing for rating, who hadn't seen Heroes in the flesh (let alone online, but at all) until last autumn. And now I recall how he played six months ago and how he plays now. Believe me, the difference is simply obvious. In these six months (maybe a bit more), he's played over a hundred games online and significantly raised his skill level, winning against strong heroes players. This is because he didn't just ask, but tried it all in practice with people ;) And it's okay if there are losses at first — the wins will start coming soon ;)