I think cels has nailed just what was wrong with the side frames—they're too side-on.  That's what I get for using other people's sprite work as reference without checking that they got it right. 

Fresh attempt at the front frame, redone mostly from scratch:



Not quite sure about the legs, shorts, ears, or perspective, but we've gone from robotic-expressionless to, "Hey, I'm having a bad day, and if you get in my way, you're going to have a worse one."  That's probably an improvement.  He's also acquired a shadow, and some muscle definition.  Looking back at the original sprite, I'm not quite sure what I was thinking in some regards.

About the tiles—I was mostly just kvetching, truth be told.  The only terrain tiles that would be affected are what would have been shadowed wall corners under the old lighting scheme.  The item tiles that will need to be reworked are mostly 1. mushrooms or 2. armour, and some of the armour tiles are pretty terrible anyway.

The perspective is intended to sort of match the cabinet perspective typical of SNES-era Squaresoft games like Final Fantasy VI, but the sprites in those are so small it's difficult to figure out whether they're perfectly side-on or matching the terrain.  I'm not doing the weird Zelda perspective where you can see all four walls at once or anything like that.

I do have this broken down into chunks/small milestones, although I sometimes get a bit ahead of myself.  Next set of art assets required:  the goblin and his move, attack, and die animations, and one monster (giant rat) and its move, attack, and die animations.  Next coding milestone:  hook up certain bits of the UI so that they're displaying real information and not placeholders, and debug the minimap that's supposed to be appearing on the bottom right.  At that point, I can either work on basic combat or sidestep into dungeon feature placement for a bit.

Existing terrain tiles (I'm not thrilled with the floor tiles, but they don't look quite as bad in context as they do here): 


Existing item tiles (the mushrooms will not be grey in-game): 


Terrain tiles in context (procedurally generated level with old goblin sprite and unfinished UI bits): 

Here beginneth my posting of the laughable assets for my game project, in the hope of improving them a bit.

This is my goblin.  You can call him A, because his placeholder name in the unfinished game UI is "A. Goblin".  A is the protagonist of my roguelike game project, currently known as "Goblin Mage".  He's presented here against a background colour that matches the main colour of the most-common in-game floor tile (we'll get to the tiles later). I suspect his arms in the side views are too long, but I'm not sure by how much or whether there's a better way to position them (or is the perspective just playing tricks on me?). I'm also not sure about the shading around his ear in the bottommost, right-facing view.  Hopefully it's at least clear that the light source is above-left. I suck at drawing humanoids.   And I really need to do my best to fix this guy, because the player is going to be staring at him all the time.

A has a walk cycle.  The walk cycle is terrible.  This is partly on purpose (the idea is to use the smallest possible number of frames, because I have to do four-directional movement animations for ~30 different creatures, which will go much quicker if I use step-pause-step-pause 1-2-1-3 walk cycles out of the 8-bit era), but in this case, I know there should be some torso rotation on the two "step" frames but can't convey it without making things look worse, in addition to the problems from the static frames.


I've included the full game palette in the walk-frames pic.  The idea is to keep it under 64 colours.  Currently I believe it's at 51.  The game is grid-based, with a nominal grid-square size of 32px, but sprites can slop over a couple of additional pixels to the right and an indefinite distance upwards without causing significant breakage.  (The intention is to have some 64x64 boss monsters, but I haven't gotten to that part of the engine code yet.)

Seriously, muscle-powered propellers do have some history in the real world, but it's with early submarines rather than dirigibles.

The clouds in the photo could possibly be rolling surf, but not treetops.  I think fskn is right about them being more readable because of value contrast.

Anyway, the figure in the foreground looks better now than in the original version of the picture, including the mask reading more clearly as a mask.

(And if you think your skill level is low, wait until I actually get around to posting the state of my current (game) project.  I expect plenty of laughter will ensue.)

Here are the things that jump out at me:

1. The lighting's inconsistent.  The leaves appear to be lit from almost directly above (maybe just a hair to the right), but the trunks are lit from our right (and not from above at all).  Keep in mind that the trunks will be in the shadow of the branches and leaves.

2. The bottom edge of the highlight on the foliage of the narrower tree needs to be more curved.  Currently it looks messy.

3. The cutoff between the different shades of brown on the trunks is so abrupt it looks like you'd get hexagonal logs if you cut these trees down.  Break up those straight lines with bark texture or dithering.

4.  This is partly due to the stylized look you've chosen, but the transition from highlight to mid to shadow on the foliage are a little tidier than I'd like.  There should be occasional patches of shadow in the brightly-lit areas where a gap in the leaves lets you see further back into the tree, and occasional brighter patches among the midtone leaves near the highlight where some branch sticks out a bit further.  Experiment.

Just my $0.02 worth.

Hmm.  The yellow sprites do stand out pretty well against the (rather dark and unsaturated) terrain tiles.  I do wonder how easy it is to quickly tell one sprite from another, though—they all have the same limited, low-contrast palette, and it's often distributed rather similarly.  The life/energy/whatever-they-are bars hovering over the individual sprites' heads are difficult to pick out where they happen to overlap with another sprite, so I hope it isn't necessary to the game to be able to figure out the readings on them quickly.

The dark sprites (enemies, I presume) aren't going to stand out nearly as well as the yellow ones.  That's a potential problem from a gameplay perspective.

The animation looks smooth, and I think it's telegraphing the movements clearly.  Realism isn't that much of a concern when the sprites are already this stylized.  (Take that with a grain of salt, though—like cels, I'm not exactly a master animator.)

From a purely artistic perspective, cels is right:  you could do with more contrast.  On just about everything.

From a *game art* perspective, making the enemies stand out more (and making it clear which sprites are enemies) is probably the most important thing.  My suggestion would be to never use yellow on the enemies, only the blues and grays (and make sure each enemy has at least some blue on it, so that it doesn't vanish into the shadows).  Pick a third accent colour for neutral units or objects, if there are any.

(I also wouldn't cover such a large chunk of the top-right corner with the UI, or at least experiment with it being in a clearly delineated area at the top or bottom, but I do realize that may not have been your idea.)