Zelda II: The Adventure of Link is a technical swordfighting game implemented in a 2D side-scrolling viewpoint with some RPG elements. It was released in January 1987 on the NES, as a follow-up to The Legend of Zelda which was released only about a year earlier, February 1986. See the wikipedia page.
Swordfighting in Zelda II is far more technical than in Zelda I. The player has a total of 8 possible directions to attack or defend from: when facing left or right there is a standing attack (high) and duck attack (low), and in addition to these four there is also an up-thrust and down-thrust attack which can be made while jumping.
In addition to the sword, the player carries a shield which can only defend the four primary positions ((high, low) x (left, right)). The shield is capable of defending against many types of attacks. Attacking temporarily lowers the shield, a tradeoff shared with Zelda I.
Finally, the player has a fairly high level of inertia, and a moderate jumping ability. Jumping can be powered-up temporarily with a magic spell earned early. Player inertia plays an important role, and while the player can only move when in the (high) position inertia can be used to elicit movement in the (low) position.
Putting aside RPG elements for the time being, the game consists of many self-encapsulated levels, the chief of these being palaces. Progress through any of these levels revolves almost entirely around combat, with only a few environmental effects in place.
Each palace is organized more or less as a simple branching maze, with each branch consisting of fairly long linear segments. Each of these linear segments ordinarily involves a series of challenges, which are overwhelmingly based around swordfighting.
Navigating the branches themselves is not actually core to the game; later palaces have more complex layouts, however, which can be challenging.
Evironmental puzzles, when they are present, ordinarily involve destroying several blocks, taking an alternate pathway to avoid combat, or overcoming dangerous jumps. Even these environmental segments are only normally challenging in the context of enemies, i.e., most would be trivial without enemies present and the interplay with the core swordfighting mechanic.
Enemies themselves are very well-thougth out. Central to almost every enemy is how they interact with the (high, low) attack/defence stance. For example:
By far the greatest success in terms of gameplay is player control. All eight attack and defence positions are easily attained, and there is complex interplay with player inertia. This is very empowering and satisfying to the player.
The player jump is probably somewhat underpowered, but otherwise has an excellent sense of inertia, aided by extra frames of animation upon landing. It is possible to adjust attack position spontaneously during a jump, as well as to control the jump's direction mid-air.
Players may remember performing extraneous sword strikes while moving about in non-hostile areas, e.g., in town. My hunch is this is a pretty common behaviour, and really just shows that the core swordfighting mechanic is satisfying even when NOT in the presence of enemies.
Many of the enemies are very well-thought-out as well, with respect to the aforementioned core (high, low) tradeoff. Lots of different interplay between these different attack and defence positions is explored.
Finally, environments are interesting enough and do interact well with the swordfighting mechanic. For instance, small ledges can make a big difference in how enemies can be approached, and these are generally placed with this in mind.
In general, the RPG elements in the game consist of severl parts.
First, the player will level up after collecting a certain number of experience points. Each of attack, defence (life), and magic can be improved, however the player is encouraged to level abilities up evenly. Experience points are lost after the player runs out of lives, providing incentive to reach the next experience points goal. This works well, because in case the main challenge for the player is proving out-of-reach (e.g., completing a palace), reaching the next level can often provide an excellent "consolation prize."
Second, there are minor quests and puzzles to be completed in towns and other secondary areas. These are interesting enough and work to provide the player the ability to become more powerful (e.g., through obtaining additional heart containers) in order to complete the main challenge which is finishing the palaces.
Overall the RPG elements are a minor part of the game, but are enjoyable and well-thought-out.
Where the game fails is chiefly where it proceeds to break up its own central conceit. For instance, the "floating eye" enemies described in the above segment are difficult to approach because they do not "play by the rules". Instead of being in a fixed (high) or (low) position, they vary smoothly between them.
Another example of this would be the flying deer heads, which follow a sinewave pattern (these are much more easily managed as they don't change direction.) As well, certain stages have enemies which attack from directly above or below, positions which cannot be defended against. This means the player is forced to resort to a change to their left/right movement to overcome these obstacles, which often hurts the rhythm of the game (e.g., the bubbles that rise up.)
Another major mis-step in design is that the player is often put in a position, intentionally, where it is not possible for them to make a correct decision as to their attack/defence position. A typical example of this would be when being attacked from the left and the right by flying cougar heads. Taken from one direction or the other, the player has to choose (high) or (low) at any given instant for either a successful attack or defence. However, when approached by these enemies from both side, their attack projectiles often arrive so close to one another than it's not feasable for the player to turn quickly enough to defend against both.
This is unfortunately only one example among many.
Other enemies are not generally defeatable using the core swordfighting mechanic. An example of this would be the dog-headed brutes who throw axes. The axes are thrown in the (high) position but in fact the players shield has no effect. The jump would be sufficient to avoid these, except that they are thrown too quickly to avoid with jumps. As a result the player is forced to take damage in order to approach these enemies, or perhaps resort to the "JUMP" spell (which is still not quite sufficient.)
A much better design would be to make the axes either less frequent, or better yet make them defendable, but carry with them a knockback; this could perhaps be balanced to make it neccessary to jump some of the axe throws in order to approach the enemy.
Other enemies require a powerup to attack, e.g., casting the fire spell on your weapon.
Casting spells would full a much better role if it merely buffed the player's position; i.e., don't put the player in a position where they can't overcome an enemy without a spell, but make the spell almost always beneficial in some way.
Finally, enemies can sometimes be unpredictable. This destroys the flow of the game because the player cannot use their inertia to properly plan attacks.
One example would be the skull-birds, who can unpredictably change direction mid-air. Without this extra feature, these enemies could be compensated for; however, because of the impossible-to-predict direction change, any attack the player plans can possibly be foiled by the birds.
Another example would be the lowly blue slimes. Seemingly, these telegraph their jump-attacks by vibrating. However, they only do this about half of the time. The other half of the time, their jumps are spontaneous and can't be predicted. This means sometimes the only way to safely approach a silme is to wait until it jumps, as there is a reliable cooldown between which it won't take another jump. This cooldown ought not to be relied on, if the enemy reliably telegraphed its attack.
As well, Darknuts only provide a very short interval to predict their attack as coming from (high) or (low) position. An earlier or perhaps persistant signal would improve the players' experience tremendously.
As a final example, there are bat-enemies fairly early on which dive down from the ceiling when you come into range. Unfortunately, it is not possible to determine whether they will dive toward or away from the player, and this complicates attack planning significantly. Instead of developing a rhythm for approaching these enemies, the only successful strategy is to move close enough to trigger them, and move away-- regardless of whether they move towards you or away from you. If they move towards you, it's then possible to attack. If they move away, it's necessary to repeat the cycle.
Zelda II's truly excellent core swordfighting mechanic and player control are by far its strongest point. Unlike e.g., the Castlevania series, the player has more fluid control over their inertia and attacks, which most would surely agree are very satisfying.
However, also unlike the Castlevania series, non-adhering or unpredictable enemies, as well as those requiring special abilities to defeat serve only to work against this core idea and as a result frustrate the player.
While similar games such as Castlevania or Ninja Gaiden may fairly be described as challenging, Zelda II is perhaps more correctly regarded as merely "difficult".
The biggest lesson learned here seems to be this: whatever your core mechanic is, don't put challenges in place that clearly and obnoxiously violate it. Certainly, variety is a good thing, and lots of types of games will rely on special abilities or explicit counters; but this is really more about fairness.
Zelda II is a game that clearly focuses on one mechanic: swordfighting. But then, perhaps to create extra "difficulty" it has gone and simply violated that mechanic in places. It's more difficult, to be sure, but at its core it's unfair, because the player doesn't really have the tools at hand to approach the challenges in a satisfying way.