Step 3: Balancing and aligning strokes
So we have considered the average height and width of our glyphs, we have decided if we want them to have ascenders and descenders, and we have figured out how to align their heights. What's next?
This step is quite lengthy, and deals with two distinct elements: structural characteristics and aesthetic characteristics. In the first part, we will look at how we can analyse the basic structure of various scripts. In the second part, we will look at how scripts can bend those rules for aesthetic reasons.
Structural outline
The next step is to look at what goes on inside the "home row" so to speak: the main space occupied by all (or nearly all) glyphs. This is very easy to explain with the Roman alphabet, due to its relative simplicity and uniformity. Take a look at the lower case Roman alphabet, stripped of (nearly) all of its stylistic elements and placed within guidelines as before:
You may have noticed that this image has one more guideline than the previous image, and that the new guideline runs directly through the middle of the home row (which is what I will call it from now on, for lack of a better term). With the addition of this single line, though, you will notice that every horizontal (or nearly horizontal) line, every juncture, intersection, and angle change occurs on a guideline. This is a rule that limits the vertical stroke density (a term I assume I have just made up).
"But wait! The circle of the hook g isn't quite on the line! the horizontal of ‹e› in Times New Roman is above the middle line!" you may say. I will of course get to this later, but for now, you should recognize that this minor non-conformity of ‹g› and ‹e› is an aesthetic, rather than structural concern. As I said, I will get to aesthetic concerns after I finish explaining structural concerns, which is more fundamental and needs to come first. The fact that there is a single guideline in the home row means there can be a maximum of one horizontal within the home row, even if it's not quite on the line.
Back to the image for a moment. We can make a small observation about the middle guideline. In all letters with a stroke that uses the middle guideline, there is something both above and below the stroke in question. this can either mean horizontals above and below it, as in ‹a g e s›, or diagonals stretching away both up and down, as in ‹k x›. This is not the case in Cyrillic, though, which allows letters like ‹ь› and ‹ч› which employ the middle guideline with open space above or below. As the designer of your own script, you get to define the rules (this could be a rule of the Roman alphabet or just coincidence, but it doesn't really matter).
So that is pretty simple. But maybe you want something more complicated. Let's see some ways that we can add a bit of complexity. If you want, you can just add more guidelines. Just make sure that you use them consistently. Look at Devanagari, for example, which uses 7 guidelines:
Thank you, D'source
We can actually combine the simplicity of the Roman system with the complexity of the Devanagari system. For this, we'll take a look at Malayalam, which, at first, appears to follow the same five-line pattern as the Roman alphabet:
At first glance, it appears to follow the five-line system fairly well, but with a little more scrutiny, this description does not hold up. For example, why does the loop of ‹ഇ› not connect to the line below it, since they both rely on the base line to govern their position? Why do the loops of ‹ േ› not touch each other, since their position is determined by the middle guideline? How can this system explain the coexistence of ‹ദ› and ‹ഭ›?
The reason is because Malayalam actually allows for the addition of secondary guidelines between the main guidelines, but, with the exception of only a few outlying glyphs, this can only happen when the guidelines above and below are already occupied. Sometimes, especially with older ligatures, these can even be further subdivided, but rarely, and only when necessary:
So you can consider the guidelines not to be absolute, and add more when they become necessary.
Finally, some scripts seem to defy being defined by this sort of classification entirely. There are a few reasons why this may be:
1. The script is too complex:
For scripts like Egyptian and Chinese, there are just too many possible stroke positions to bother trying to align them, not to mention the fact that they can be written both horizontally and vertically. Instead, these scripts rely on a different technique, namely:
2. Glyphs are centred
Instead of aligning to a base line, like Roman or Arabic, or aligning to a head line, like Devanagari or Tibetan, a script may instead align to a central line. If it does, then a script may derive its unity through a sort of symmetry with its other half, without reference to other glyphs. This is common in vertical scripts. Nevertheless, unless the script is a logography, most glyphs are probably going to have a similar density of strokes.

Horizontal Egyptian text is generally aligned along a central writing line.

As you can see here, Chinese is not always based on the exact centre of the character, but the perceived visual centre based on the weight and shape of the strokes.
3. Writing direction has changed
Sometimes, script direction may change. This happened with Mongolian, but since all the glyphs were rotated along with the writing line, there is no real change. However, for scripts like Chinese (as mentioned) and Kulitan, when the writing direction changed, the glyphs were not rotated to compensate.

Like Chinese, Kulitan glyphs are more or less aligned on a central writing line.
4. The progression of text is non-linear
For the Nastaliq style of Arabic script, which is the default style for writing Urdu, each word is written along a gradually descending diagonal line that often overlaps with the previous one somewhat.
5. Some other reason
If you have some design idea that destroys this, then go ahead and use it. Maybe the ascenders of a single letter with multiple ascenders get progressively taller throughout the letter and then reset for the next letter. Maybe glyphs of varying heights stack and are squashed to fit in a consistently-sized home row. Maybe something else.
One of my own scripts, Naduta has a fairly light level of complexity for basic glyphs, but there are a couple different heights foor ascenders, and when glyphs are stacked, they will always reach somewhat higer above the head line than they will below the base line. The glyph on top will generally be lowered below the had line when possible to fill gaps in the base glyph, leading to a lack of alignment with the various upper and lower extents of glyphs. The Naduta Common script, which evolved from cursive Naduta, inherited a lot of the disorder of its parent script.
Aesthetic considerations
Everything I said above is only meant to explain how the letters are governed at their most basic form. Many scripts have aesthetic touches that can alter these significantly. In curved Roman letters such as C, the end points actually appear slightly below/above the head and base lines. In letters like ‹e›, where the lower curve angles up and approaches the middle horizontal stroke, or ‹g›, where the upper circle needs a little extra space to avoid looking cramped, lines may be pushed out of the way a bit to avoid clutter and create a greater visual balance.
You can also have different elements take different heights, such as different types of ascenders and descenders, or what have you.
As an example, if you try to apply straight, horizontal guidelines to the
Georgian alphabet, you will give yourself a headache, because all the joints, endpoints, and so on fail to align neatly on the same horizontal lines as each other. However, you will notice that no glyph has more than two end points/junctions/angle changes within the home row. Additionally, any letter with a particular style end point etc. will place it at the same height as all other letter with the same type of element, but it seems that each different type of element has its own specific height at which it appears. For example, the letters ‹ვ კ ფ ც ჳ ჴ› all place the middle endpoint in exactly the same place, i.e. slightly above the base line; the middle endpoints of ‹ლ ო რ უ ღ› all stop at the halfway point of the home row, and the bottom hooks of ‹ე ვ კ ჟ უ ფ ქ ყ ჭ ჴ› all end at the same place, i.e. at the base line. You'll notice also that the presence of a descender shrinks the size of a circle, so, ‹ფ› has a smaller and higher circle than ‹თ›.
So what this means is that you can define the basic permissible shapes with the guidelines, but you don't need to follow them rigidly when designing the final appearance of the script. You can break the rules as much as you want, so long you do it in a consistent or understandable manner (even if the rule break is seemingly as arbitrary as "‹t› has a shorter descender because that's what I feel like doing").
Here are the new scripts:
Script A:

I've given this one a single guidline inside the home row, and the lines are rigidly fixed on this line. This has caused me to eliminate or rearrange some strokes.

This script has two guidlines inside the home row, but the lines are not rigidly fixed on them. They adjust their exact positioning based on their surroundings. In addition, there is a single letter that has an extra guideline.
I think we can say that these scripts have taken on a little character and are distinct form each other, but they are still lacking in strong identity. The next step is where you can add real character to your script and finish it off.