About down+away DI in Marth v Marth:
Most of Marth's launchers are Sakurai angles (so 45° above a certain low knockback threshold), and you usually can't combo against down+away DI because they are either out of hitstun in time, in a knockdown situation or offstage. However, using a move with a higher angle (around 60°-70°) will allow you to combo against down+away DI like you combo against no DI with Sakurai angle launchers. For Marth, this is mainly tipper fair (67°).
I managed to combo into pivot f-smash at 20%, there should be a percentage range around it where it works but I think it is rather narrow because on the lower end Marth can jump out, while on the high end the launch speed is too high, and while you barely manage to catch up to Marth, by the time you perform the f-smash (during which you stand still) he is out of range again.
That doesn't take away from the fact that tipper fair is still good in this situation, you just have to choose an aerial followup.
You will be able to register your tipper hit by the time you have to commit to the pivot tipper, so you won't run the risk of throwing out a potentially whiffing f-smash.
The biggest risk I see in this setup is them crouchcancelling the tipper fair, but that is always the risk with attacking low-percent opponents, and if you do the fair right before landing, they won't be able to punish on reaction, so you have counterplay, making it a viable tool.
U
UnderTheKnife
Match analysis helps in many ways. I'd argue the most important one is that you learn to recognize more patterns at the same time, and this one will translate very directly into your play because your brain just becomes faster at figuring out what the situation is in real time. Knowing the solution in that particular situation would be pointless if the time window to apply the solution is already over by the time you recognized the situation.
Watching your own videos can be useful because while playing, noticing your mistakes is more difficult. But, chances are you either already know what you are doing wrong, and a lot of the time you could figure out how to be better at individual microsituations, but the knowledge gained from this is very "uncondensed" and one tends to be overwhelmed by it, to not be able to apply it in real-time, to forget it or to apply it in situations that look similar to you but need to be treated very differently.
The most useful thing about watching your own videos is recognizing what is costing you most at the moment and what areas of the game you need to focus on.
Note that it is different if a player with higher game understanding looks at your video. They might be able to point out your core misunderstandings, which can help a lot and is one of the reasons coaches are so effective in all fields of learning. This happens most commonly if they had to deal with the same misunderstanding at some point and therefore easily recognize it, also it requires them to be good at explaining things. If these conditions are not met, they will most likely point out individual mistakes you made that are also recognizable to you, although they might mention a few general principles that apply in these situations in the process, or you can take something from the way they talk about the game, so it is not completely unhelpful either.
When watching matches between strong players, you need to understand why the options they choose are so strong. My method for this is to take one reaction time span (I always assume 15 frames which is an average reaction time, it is usually a little more in reality because what you register is more complex than a simple yes/no reaction, but working this way around is safer than assuming more), and replace it with different actions to look at possible outcomes.
For example, not being able to hit the other player with a certain move with the altered inputs in the reaction time span means that the player is outside of that move's threat zone.
Other things to look out for are how top players use threat zones to retain pressure if they have no true punish, or to shift the risk/reward-ratio in their favour when using unsafe moves.
You can do this when looking at your own games, too, and see if you are always dashdancing inside threat zones without noticing or something...
For understanding new concepts and condensing your knowledge of individual game situations into a solid gameplan, discussion, especially with strong players, is very helpful in my opinion. It often requires an idea coming out of nowhere and usually more people have more of them than one person has. Also make sure to test these ideas and possibly find exceptions to the rule, because in a game as complex as Melee those exceptions almost always exist.