The Rise of Wesnoth campaign replays (hard)

[baruk]

I finally finished this campaign on hard difficulty without using save-loading, but with many scenario restarts.

I include replays for each of the non-plot scenarios. I couldn't find a replay of the Fall, or A new land, so I made substitute attempts where I achieved the main objectives I had carried to the next scenario (about the same amount of gold, and all 3 heavy infantry in the Fall).
I've also included a saved position of the final position in the final scenario.

I figure this should be useful resource if anyone is stuck on this campaign, you can check my replays to see how much in the way of gold and recalled troops I needed to finish each scenario.

The order of the elvish tasks I chose was: cursed isle, troll hole, lizard beach, the dragon. I chose Swamp of Esten rather than the Midlands. I numbered my replay files so the order should be obvious.

I finished with my inflicted damage about 1% higher than expected, and my sustained damage about 2% lower than EV. I sustained 105 losses (I found the campaign challenge enough without trying to also get through without losing a unit).

[Shotor_Rider]

I had finished tRoW on medium long time ago. Seeing your post made me do it again. I started it on hard, and with not a single save and load I got to the clear port. This scenario is not fun at all:(

Anyway the replays of the first 5 scenarios are attached..I think I have it almost to the perfection. Nevertheless I have lost two soldiers.

[Elvish_Pillager]

I've reached Clearwater Port with no losses, but the scenario is awful to play.

I think that was on Medium. I don't want to bother digging up the replays though.

[Shotor_Rider]

HAHA! the power of save-loading: Finished clearwater port on hard with no losses and killing all the three orc leaders...obviously my EV is a lot different from the actual values.

The more sensable thing is to stay back home and just defend, as baruk has done but this is a boring strategy.

To make it more challenging Elvish Pillager if you are the master of wars do what I did in a more honset way, with casualities of course.

Here is turn 26 where the last king is cornered...enjoy killing him.
I also include the start of scenario if you want to try..you have a good amount of gold and plenty of high level units.

[Elvish_Pillager]

When I want to do a scenario with no losses and no work, I hack my Wesnoth to always generate hits for side 1 and misses for other sides.

[Shotor_Rider]

I agree cheating is cheating, S/R or hacking the codes. I tried to go back and at least finish up to clearwater port with no losses. But no matter how many times I try in Harrowing escape I lose one unit. This is really frustrating becuase its a relatively long scenario.

I have never seen any of your no loss replays EP. But I think its not all strategy, you also need to get really lucky. When you play for no losses, at some point you should have a unit with at least 50 percent chance to be killed. Correct? or you play it out so perfectly that each unit has less than say 10 percent chance to be killed in all the turns?

At any rate I'd love to see one of your no loss replays. I was trying to open some of the replays for HttT that you have posted but they were corrupt or maybe for previous versions...I am using the most recent stable version of wesnoth. For example finishing bay of pearls with no losses requires a super amount of luck. I also found some replays by (S)elfish wierdo that he claims has finished several campaigns with no losses. But either he has restarted some scenarios 100 times or he is really lucky. He has units with at least 90 percent chance to be killed and yet he gets away with it.

[Elvish_Pillager]

I have never seen any of your no loss replays EP. But I think its not all strategy, you also need to get really lucky. When you play for no losses, at some point you should have a unit with at least 50 percent chance to be killed. Correct? or you play it out so perfectly that each unit has less than say 10 percent chance to be killed in all the turns?

If I remember correctly (because this was a while ago) I very rarely had a unit with more than 1/3 chance to be killed. I only occasionally (~once per scenario) had over 10% chance to be killed. Overall, I played each scenario 2-3 times on average.

Granted, this was on Easy difficulty, but then again, it was the old Easy, before the devs went and made Easy actually easy.

At any rate I'd love to see one of your no loss replays. I was trying to open some of the replays for HttT that you have posted but they were corrupt or maybe for previous versions...I am using the most recent stable version of wesnoth.

Yeah, they're all ancient. Maybe you could view them if you got some Wesnoth version like 0.8.4. I don't remember exactly, although I think it's recorded in the savefile.

I'll remember this and maybe play some more lossless games in the future. But not HttT. I'm sick of HttT; I'll never play The Lost General or beyond unless they change it a lot.

TRoW is tricky. Maybe I'll try that one.

For example finishing bay of pearls with no losses requires a super amount of luck.

Indeed. It took me about 10 tries and I had a unit with almost 50% chance to be killed, unlike any other scenario.

I also found some replays by (S)elfish wierdo that he claims has finished several campaigns with no losses. But either he has restarted some scenarios 100 times or he is really lucky. He has units with at least 90 percent chance to be killed and yet he gets away with it.

He restarts over and over. He beat Test of the Clans(sic) where I didn't, for instance, by reloading from the beginning of the scenario until he got about 200/330 damage taken, and it's probably even better than that, situationally.

IIRC, I had about 10% over EV on damage dealt and 20% below EV on damage taken, at the end of the campaign.

[baruk]

For example finishing bay of pearls with no losses requires a super amount of luck. I also found some replays by (S)elfish wierdo that he claims has finished several campaigns with no losses. But either he has restarted some scenarios 100 times or he is really lucky. He has units with at least 90 percent chance to be killed and yet he gets away with it.

I've recently started playing HttT again, on hard difficulty, version 1.0.2. I was never entirely satisifed with my first completion of it, having save loaded at a couple of points in Valley of Death (it has nagged away at me).
I'm going for a no losses approach, my first such attempt at any campaign (as well as no save loads, with the exception of scenario start). As challenges go, it seems a positively-reinforcing approach: the difficulty of avoiding casualties is offset to a minor degree by the benefit of having long lived units, whose experience will begin to tell as the campaign goes on. I wonder if anyone has attempted a more self-impairing challenge, such as a "no recalls allowed" attempt?
So far I've reached the Seige of Elensefar with no losses, and after a few cursory attempts, I'm confident of getting through this scenario also. The orc is a pushover, and the time limit is not a worry. I think I will have to briefly retool to deal with the undead troops (by recalling some thugs), they usually take out a unit or two before I've finished the orc leader.

Bay of Pearls has been the toughest challenge so far. It took somewhere between 10 and 20 attempts, maybe a little more. I would say that my progress in the scenario has not been solely down to luck. I believe I've hit upon a reliable and reusable no losses strategy for this scenario. It won't work every attempt, but you won't have to go through 100 restarts to succeed.

I would concur with Elvish Pillager, in the respect that most turns, the worst risk of unit loss I encounter is less than 10%. I encounter greater risks if I make mistakes, if I have a few bad fights in a row, or if I'm in "time-trouble", and have to throw units into danger in order to achieve the objectives. The recruit order of the AI leaders has turned out to be not as important as I had first thought. The number of level 2 orc warriors recruited does vary, but shouldn't matter with a careful enough approach.

For Bay of Pearls, the answer is to fight on your terms rather than the AI's. I opened the first 3 cages, sending 3 or 4 mermen to the northwest (I've tried keeping one merman back for defence, with moderate success). I had about 200 gold, recalling my most experienced troops (including 4 level 2s).

As long as you keep the mermen at least a full movement away from the enemy units arriving from the sea castle, I've found that they can avoid fighting until they reach the 2 nagas guarding the NW cage. I try to keep the mermen in the NW of the map for a couple of turns or so, until the first wave of bats and nagas have hit my main force on the landmass.

My land forces are usually enough to wipe out the bats and nagas, you can usually tempt the nagas onto the land for easy kills, or just use mages or marksmen to nuke them out of the water. With enough forces available, you can destroy these guys before the orc forces arrive.

After making several unsuccessful attempts at holding the line against the orcs, I decided to take a different approach. As soon as the bats and nagas were wiped out, I instead withdrew my forces to the small stretch of water between the castle and the mainland (using the island village as the central point). It usually gives you an extra couple of turns before the orcs arrive (crucially getting you out of the night into daytime). The other advantage is that you now have a more defendable frontline of 5 hexes. With tough enough units in the frontline, you will be able survive the initial enemy attack. In your turn, simply rotate the injured units out of the line, and strike back at (hopefully injured) enemy units in 20% terrain.

Meanwhile, the mermen in the northwest should have made their move as soon as the main force of nagas and bats are engaged with your land army. I usually take out any stray nagas and bats first, before opening the final cage. I try to attack the sea castle leader in daytime, when his maximum potential damage is 24, using storm trident followed up by melee attacks if he's down to low hps. Merman activity near the land is usually enough to draw some orcs into the water. This will have the useful effect of taking some pressure off your land forces. From here the scenario is straightforward to wrap up, so it is usually where I start making mistakes, it is quite easy to over extend your forces when they go on the offensive. The second orc leader is usually a little more risky for your mermen to attack, as he tends to have more units guarding him.

As long as the AI acts predictably, taking no losses in this scenario is not too difficult. If your merman force gets bogged down fighting with an equal or greater force of nagas, then casualties are inevitable, barring an improbable run of fight results.

For the rest of the campaign, the scenarios that worry me in terms of being remotely completeable without losses include: Valley of Death, and the final 2 scenarios. Some of the cave levels could be a challenge, as you have to rely on a bunch of new level 1 units.

So far, Cursed Isle has also provided a tough challenge. I haven't yet managed to kill both leaders, even with losses. My no loss attempt at it was simply a survival victory, I wasn't even able to kill one of the liches. I've tried using merman assassination squads, but with little success (and many losses).

[Shotor_Rider]

Baruk, post your replay for bay of pearls.

Here is a tactic that I sometimes use, do you guys consider it cheating? I play one scenario first with saving and reloading, trying different approaches, to see what AI does. Then I start over and play the whole scenario with no S/R when I have that figured out.

I still have not figured out how is AI thinking? Here are what I have noticed:

1. AI never attacks a unit that is about to be leveled up and is relatively healthy. These units are great! even better than leveled up ones sometimes..for example in clash of the armies, I have Kapoue at the bottleneck and elves and mermans never attack! I have not managed to do this, but imagine such a unit sitting at the enemy's keep..LOL.

2. AI if it can always attacks a unit with no ranged attack using a unit with a ranged attack. This can be a good way to lure a unit to a place that can be killed safely. And the reverse situation also holds, you can put a mage in the atatck zone of two archers that can easily kill him with their ranged attacks, but the stupid AI attacks with mellee.

3. An AI leader with less than half hp usually does not attack.

4. Damage is more important than a village for AI. So if there is a village that could be captured by an AI unit, leave a weak unit in his attack range. Of course a unit that can survive the attack and retreats the next turn.

[baruk]

Baruk, post your replay for bay of pearls.

I have attached my replay. Watching it again, there are several risky moments, usually when I place a unit in an exposed position to finish an enemy unit. Some instances are calculated risks: I knew the unit could get very badly wounded, but not killed, in one round, but on others I rode my luck.
I just finished Siege of Elensefar also, but it proved more difficult than I had thought: when fighting multiple orc warriors at night, it is quite easy to lose a unit. In my completed scenario, I was lucky in that orc crossbowmen were recruited instead. In this case, the chosen recruits did reduce the level of challenge of the scenario.

Here is a tactic that I sometimes use, do you guys consider it cheating? I play one scenario first with saving and reloading, trying different approaches, to see what AI does. Then I start over and play the whole scenario with no S/R when I have that figured out.

That isn't cheating in my opinion. As a wise man once said, "Know your enemy".
Playing a game with save/loading is not a problem if you are treating that particular game more as an intelligence gathering operation than an actual challenge.
I often have to restart a scenario many times before I can finish it to my satisfaction. The result of many restarts is that you pick up a knowledge of how the AI behaves. Working out how the AI behaves is an interesting part of the game, a kind of side-challenge. By playing the game, you cannot avoid picking up an idea of how it behaves.

Save/loading may or may not be cheating, but it certainly reduces the challenge aspect of the game. I find I need a certain level of challenge for a game to interest me. Without the challenge aspect to it, I find I will get bored with a game very quickly.

[Shotor_Rider]

wow! I saw your replay, very very nice! just one comment: since you won't carry any gold to the next scenario, there is no need to rush capture the villages unless they were already enemy's villages (I didn't really pay attention) I would not even recruit the first few turns so that I won't have to pay upkeep. when the orcs get in a closer range I start recruiting and line them up in front of the castle (they have %30 def. as opposed to 20) put a druid behind them and once they get injured recruit new units to replace. the only drawback is that one village in front of castle which you should let go.

[Elvish_Pillager]

I've recently started playing HttT again, on hard difficulty, version 1.0.2. I was never entirely satisifed with my first completion of it, having save loaded at a couple of points in Valley of Death (it has nagged away at me).
I'm going for a no losses approach, my first such attempt at any campaign (as well as no save loads, with the exception of scenario start). As challenges go, it seems a positively-reinforcing approach: the difficulty of avoiding casualties is offset to a minor degree by the benefit of having long lived units, whose experience will begin to tell as the campaign goes on.

Just you wait until you get to the Valley of Death. It'll be difficult, but at least when I got to it, I had about a dozen level 3 units. The thing about No Losses is that any experience gained is permanent.

Unfortunately, that leads to even more instability in later levels: Some, you will steamroll over with hordes of advanced units; others, like the Test of the Clans, your units will not be tough enough to handle. One thing I haven't considered is selectively AMLAing a dedicated crew of point-guard units. Of course, AMLA has not been around since I got that far...

Wait - you're going for No Losses on hard difficulty? I don't think anyone has gotten past the Siege there - or did I? hmm... at any rate, it took me about 5 tries and then I won or gave up (I don't remember which.)

I wonder if anyone has attempted a more self-impairing challenge, such as a "no recalls allowed" attempt?

I've been through most of HttT without recalling units. It's not hard. However, if I used no recalls and allowed no losses... Valley of Death was hard enough with hordes of level 3s.

So far I've reached the Seige of Elensefar with no losses, and after a few cursory attempts, I'm confident of getting through this scenario also. The orc is a pushover

and the time limit is not a worry. I think I will have to briefly retool to deal with the undead troops (by recalling some thugs), they usually take out a unit or two before I've finished the orc leader.

I use Mages. They'll serve you much better later on.

The number of level 2 orc warriors recruited does vary, but shouldn't matter with a careful enough approach.

Hmm? When I played, I always made sure that the enemy didn't recruit any warriors. Of course, they've probably also fixed the AI since then.

...or if I'm in "time-trouble", and have to throw units into danger in order to achieve the objectives.

Funny; I never get into time-trouble. The important thing is that killing enemies quickly keeps you from dying much more than pure defensive strategies to, and if you're on top, you can use the time to move forwards, as well as killing enemies and forming a defensive formation.

As long as you keep the mermen at least a full movement away from the enemy units arriving from the sea castle, I've found that they can avoid fighting until they reach the 2 nagas guarding the NW cage. I try to keep the mermen in the NW of the map for a couple of turns or so, until the first wave of bats and nagas have hit my main force on the landmass.

My land forces are usually enough to wipe out the bats and nagas, you can usually tempt the nagas onto the land for easy kills, or just use mages or marksmen to nuke them out of the water. With enough forces available, you can destroy these guys before the orc forces arrive.

You use your land force to kill their sea force? Weird, I never thought of that!

My strategy: March a Druid out into the water and surround it with mermen. Let the nagas break on the mermen and cycle any wounded mermen back to a safe place. This should be even more effective now that the limit on total healing by a healer has been removed (at leased, IIRC it's been removed.)

After making several unsuccessful attempts at holding the line against the orcs, I decided to take a different approach. As soon as the bats and nagas were wiped out, I instead withdrew my forces to the small stretch of water between the castle and the mainland (using the island village as the central point). It usually gives you an extra couple of turns before the orcs arrive (crucially getting you out of the night into daytime). The other advantage is that you now have a more defendable frontline of 5 hexes. With tough enough units in the frontline, you will be able survive the initial enemy attack. In your turn, simply rotate the injured units out of the line, and strike back at (hopefully injured) enemy units in 20% terrain.

Gee, you did this scenario in a totally different way!

My strategy for facing the Orcs at daytime: Charge out ASAP and meet them in the middle. Don't fight yet! Keep Horsemen back just out of Orc range; it makes them cower. Then, when they chase you at Night, you have plenty of room to back off, collecting reinforcements. Then, I win by killing them as fast as possible during the day. (For extra trickyness, I kill all the Archers, but leave the trolls, so that my 5-move units can escape the battle the next night.)

As long as the AI acts predictably, taking no losses in this scenario is not too difficult. If your merman force gets bogged down fighting with an equal or greater force of nagas, then casualties are inevitable, barring an improbable run of fight results.

Straight mermen vs. straight nagas in open sea works like that. Healing is very important in the sea battle.

For the rest of the campaign, the scenarios that worry me in terms of being remotely completeable without losses include: Valley of Death, and the final 2 scenarios. Some of the cave levels could be a challenge, as you have to rely on a bunch of new level 1 units.

Valley of Death, Dwarven Doors, The Lost General, and Test of the Clan(s) were the worst. When I played, Dwarven Doors was the hardest scenario overall. However, it was also hard without No Losses and it is probably easier now on Hard than it was then on Easy.

The Battle for Wesnoth is easy to beat if you have dozens of level 3 units. The "no losses" bit makes it a bit of a challenge instead of a total joyride.

So far, Cursed Isle has also provided a tough challenge. I haven't yet managed to kill both leaders, even with losses. My no loss attempt at it was simply a survival victory, I wasn't even able to kill one of the liches. I've tried using merman assassination squads, but with little success (and many losses).

I've beat the liches on easy, with losses.

IMO, the best way to win is NOT to beat the liches. If you don't beat them, then Mormirmu (if you get him) shows up in the Valley of Death with a bunch of White Mages, and you don't even have to worry about them dying, because they're allies.

[baruk]

wow! I saw your replay, very very nice! just one comment: since you won't carry any gold to the next scenario, there is no need to rush capture the villages unless they were already enemy's villages (I didn't really pay attention) I would not even recruit the first few turns so that I won't have to pay upkeep. when the orcs get in a closer range I start recruiting and line them up in front of the castle (they have %30 def. as opposed to 20) put a druid behind them and once they get injured recruit new units to replace. the only drawback is that one village in front of castle which you should let go.

The money question was quite interesting: I actually needed early village gains in order to fund my 10th unit (I think it was 10 recalls in total). As there are only 2 castle slots, I felt it necessary to recruit early rather than wait: I wanted early troops so that I could face the initial wave of nagas and bats, and annihilate them. 2 of my first 4 recruits were loyal, so my income was not too badly affected.

I think there was a gold bonus for finishing early, though not sure on this as I finished the second leader on turn 21 of 21. Something I didn't notice before is that the leader you kill determines the choice of next scenario: sea leader for cursed isle, land leader for peninsula, kill both for a choice. You seem to get your 100 basic starting gold for the isle, whereas you don't for the peninsula: you have whatever you finished bay of pearls with.

Holding a line in the middle of the water may seem a little insane with ground based troops. However, I had taken out the amphibious threat of the nagas, so only faced enemies on one front. If I had held the line further back, the orcs would not have had the movement to engage my forces. As I had many available hit points to spare, I wanted the orcs to attack, whilst taking heavy wounding in return, making them vulnerable for my attacks the following turn (using fresh, uninjured troops from the second line).

Wait - you're going for No Losses on hard difficulty? I don't think anyone has gotten past the Siege there - or did I? hmm... at any rate, it took me about 5 tries and then I won or gave up (I don't remember which.)

I just reached the princess' revenge. I had to use a fairly conservative, defensive approach for crossroads, and I have 250 gold available.

Just you wait until you get to the Valley of Death. It'll be difficult, but at least when I got to it, I had about a dozen level 3 units. The thing about No Losses is that any experience gained is permanent.

....IMO, the best way to win is NOT to beat the liches. If you don't beat them, then Mormirmu (if you get him) shows up in the Valley of Death with a bunch of White Mages, and you don't even have to worry about them dying, because they're allies.

Sounds like I might be in trouble..
I have a grand total of 2 level 3 units, but also one level 2 within 2 exp of levelling. I have about 10 level 2 units.
Having chosen cursed isle, I am looking forward to Moriremu's arrival, though surviving until then will be tough enough.

baruk wrote:
So far I've reached the Seige of Elensefar with no losses, and after a few cursory attempts, I'm confident of getting through this scenario also. The orc is a pushover



baruk wrote:
and the time limit is not a worry. I think I will have to briefly retool to deal with the undead troops (by recalling some thugs), they usually take out a unit or two before I've finished the orc leader.

I use Mages. They'll serve you much better later on.

In my first couple of attempts at the Seige, I seemed to do well against the orcs, only losing a unit when the skeleton backup arrived. But in the following attempts, the orc started doing better and better against me, often dropping one of my units in the first exchanges (two orc warriors, getting a good run of hits, seem to do enough damage to kill any of my units).
In my successful attempt, the orc only recruited crossbowmen and assassins, and suddenly it became very straightforward. I recalled 2 thugs and a mage to help with the skeletons. I am beginning to regret not exping more mages earlier on, I would really like to put together a silver mage squad for the later levels.

baruk wrote:
...or if I'm in "time-trouble", and have to throw units into danger in order to achieve the objectives.

Funny; I never get into time-trouble. The important thing is that killing enemies quickly keeps you from dying much more than pure defensive strategies to, and if you're on top, you can use the time to move forwards, as well as killing enemies and forming a defensive formation.

I think I lost the isle of anduin once or twice due to the time limit: I was being too defensive, and had to find ways to be more aggressive (whilst keeping units safe) before I could finish it.
My first successful no losses attempt at crossroads is a 23 (of 33) turn victory. However, it was far too slow as I had also recalled heavily, and could only take the default 100 gold to the next level. I had decided to kill both leaders, in case there was a bonus (and for the extra exp), but it took too much time.

I agree about the aggression thing. The trick is to get just the right amount. My method is to get a no losses victory using my most defensive approach first of all. If the result is not good enough, then I try again with more risk taking (or if I'm losing too many units, I try to be more defensive, and so on). Assassination is a further alternative.

My strategy for facing the Orcs at daytime: Charge out ASAP and meet them in the middle. Don't fight yet! Keep Horsemen back just out of Orc range; it makes them cower. Then, when they chase you at Night, you have plenty of room to back off, collecting reinforcements. Then, I win by killing them as fast as possible during the day. (For extra trickyness, I kill all the Archers, but leave the trolls, so that my 5-move units can escape the battle the next night.)

Check out my Bay of Pearls replay in my previous post on this thread, our strategies may have more similarity than you think.
My first two recruits are and elvish rider and a horseman (Haldiel). Their job is to take the villages along the south of the map. In doing so, they draw some attention from orcs and bats (it buys some time as the enemy are forced to waste movement points). After the naga/bat force strikes, I move them back in order to provide support. And they retreat with the rest of the army at dusk.
Another factor is that the orcs tend to split up if your mermen start threatening their leader. If you can kill the sea leader early enough, you can really upset the orcs by taking all the sea villages and moving near the coast. This makes the land army's job a lot easier, as any orcs facing them arrive piecemeal, rather than in force.

baruk wrote:
As long as the AI acts predictably, taking no losses in this scenario is not too difficult. If your merman force gets bogged down fighting with an equal or greater force of nagas, then casualties are inevitable, barring an improbable run of fight results.

Straight mermen vs. straight nagas in open sea works like that. Healing is very important in the sea battle.

I focussed on getting the NW cage, and the trident, as early as possible. After the fight, I left the injured rest for 2 hp a turn.

Druid healing sounds like another viable way to go. As the 3 accessible cages will give you 4 mermen, I would probably need to use more support from land forces to beat the nagas.

baruk wrote:
For the rest of the campaign, the scenarios that worry me in terms of being remotely completeable without losses include: Valley of Death, and the final 2 scenarios. Some of the cave levels could be a challenge, as you have to rely on a bunch of new level 1 units.

Valley of Death, Dwarven Doors, The Lost General, and Test of the Clan(s) were the worst. When I played, Dwarven Doors was the hardest scenario overall. However, it was also hard without No Losses and it is probably easier now on Hard than it was then on Easy.

The Battle for Wesnoth is easy to beat if you have dozens of level 3 units. The "no losses" bit makes it a bit of a challenge instead of a total joyride.

My thinking on the final scenario is that facing so many level 3 units, with so many individual battles, will mean that there are perhaps too many chances to lose units.

Dwarven doors could be tough. When I completed it on hard before, I had the advantage of using expendable units to slow down the orcs.

My memory of lost general is fading, but I seem to remember level 2 trolls being involved. Lots of damage combined with the low defence of cave terrain will be a tough combination. Sceptre of fire will have the same difficulty.

[Elvish_Pillager]

3 level 3s, even after the Princess of Wesnoth? Did you spread the XP out too much, or is there less XP now on hard than then on Easy?

I usually start out with a more aggressive strategy - some scenarios can be one by charging out with a few high-level units, and if you start out defensive, you won't get that.

As to the Battle for Wesnoth scenario: The enemy also has a lot of level 2s and level 1s, and by that time, you should have Dwarvish Lords, who are even better at absorbing damage.

The Lost General is hard, but the Sceptre of Fire isn't. In the Lost General, you have to fight level 2 orcs and skeletons with level 1 dwarves, but in the Sceptre of Fire, you have to fight level 2 orcs and trolls with level 3 dwarves. You have to level up those Dwarvish Fighters ASAP, that's all.

[baruk]

3 level 3s, even after the Princess of Wesnoth? Did you spread the XP out too much, or is there less XP now on hard than then on Easy?

I got that slightly wrong. I had 3 level 3's at the start of the Princess of Wesnoth scenario. Tricky at first, but once I had found the right tactics to use, it was straightforward, and enabled me to gather some needed exp. I found you have to treat groups of multiple swordsmen with respect: often I would hold off from attacking until the AI had made a positional mistake, allowing me to drop his units using overwhelming forces without risk of counterattack. I was up to 4 level three units at the end of that encounter.

Just today, I managed to finish the Valley of Death: the Princess's Revenge scenario (with no losses and no save loads). I have attached the replay, it is fairly entertaining (turn 4 and turn 8 were particularly fun ). I was surprised, as it probably took fewer attempts and less time than the Bay of Pearls. I had been expecting weeks of hair-pulling frustration, until I would finally give up, but not so.
My strategy was to rush the SW leader (sending 2 horse units to collect holy water in the meanwhile), and then defend in the SW for the remaining turns (with help from Moremirmu and friends).
My recalls: grand knight, outlaw, white mage, mage, 2 thugs, bandit, elven shyde, elvish marshal, 2 knights, elvish rider. 5 of these units along with 2 leaders all levelled in the scenario, providing much needed boosts to hps at crucial points.
Having Moremirmu appear up on turn 7 was fairly important for my attempt, as on turn 8, among other instances, his team was vital in helping to eliminate 6 wraiths on the right flank of my army. My previous (with losses) attempt at this scenario on hard was a bit of a nightmare, as having completed the peninsula level previously, I had no help from Moremirmu. Incidentally, Moremirmu and crew are controlled by you in the version I played (version 1.02), and I was able to keep them alive (I figured they would count as losses).
At the start of Gryphon mountain, I now have six level 3 units and 13 level 2's available in my recall list. If I'm lacking in exp, it may partly be because Delfador has got so much of it (about 250 so far) and that my somewhat defensive tactics have allowed a higher than usual number of enemy units to survive.

I usually start out with a more aggressive strategy - some scenarios can be one by charging out with a few high-level units, and if you start out defensive, you won't get that.

I would call that an assassination-type tactic, where you are strong and mobile enough to ignore the enemy army, and go directly for the leader.
Usually in my first few attempts at a scenario, I get a feel for what to do, what kind of approach works. I'll usually give the assassination tactic a try in that time. My preferred tactic at the moment is more defensive, as I feel I can reduce the risk of losses a lot easier that way.

As to the Battle for Wesnoth scenario: The enemy also has a lot of level 2s and level 1s, and by that time, you should have Dwarvish Lords, who are even better at absorbing damage.

The Lost General is hard, but the Sceptre of Fire isn't. In the Lost General, you have to fight level 2 orcs and skeletons with level 1 dwarves, but in the Sceptre of Fire, you have to fight level 2 orcs and trolls with level 3 dwarves. You have to level up those Dwarvish Fighters ASAP, that's all.

I agree, the crucial hurdle is levelling your dwarves: if you can do that, the cave levels (and levels beyond) can become a formality.