Welcome back to our third post on how to crate an efficient Trading Algorithm Model, today we're going to talk 'bout something slighly different than the previous, in fact as in the lasts post we proceeded with a Technical Analisys today we are going to explain how we can use Fundamental Analisys Tools to import datas into our Model and exploit those with some easy, yet effective, computer-based methods.
As you should know Fundamental Analisys, differently from Technical, isn't enitrely based on the title value itself, therefore today we are not just going to examine the title's trends' history, but we'll how non-numerical factors of different kinds can affect a title value and how they can be incorporated to our Model.
To go on We'd suggest you to read, if you haven't already, our second post on this project.
If you're not new to trading you should know what is a Economic Calendar (if you don't know don't worry check it out here), basically what we are going to explain now is not simply how to make your Algorithm proceed operations based on what your EC tells (there are many python programs on gitHub that can do that) what we want our Model to do to is tfind out how some different kinds of external facts influence the value of the title, and we will do this by working in a similar way like we did in the last post.
So, right now you might be asking yourself how to do this monumental and difficult job, don't worry we'll come in your help.
Basically there are many ways to get ther result we want, it is hard to tell which one is the best, the ideal 'd to use as many methods as you can but clearly this will resullt also in an increase of task, and their difficulty, to complete and therefore in a major energy consumption.
Here's a way to do this, let's get started!
As we saw in the second post, we can find, with our strategy, a bunch of similar Trend Sequences, our task now is to understand if this TS (Trend Sequences) are linked in any way to some external factors of the same nature, for example if we talk about Tesla an increase in the Title's value might, as it might not, be due to a new model's presentation or even be due to less straight-forward events like eg. climate change drammatic report.
As we saw in the second post, we can find, with our strategy, a bunch of similar Trend Sequences, our task now is to understand if this TS (Trend Sequences) are linked in any way to some external factors of the same nature, for example if we talk about Tesla an increase in the Title's value might, as it might not, be due to a new model's presentation or even be due to less straight-forward events like eg. climate change drammatic report.
Once we have a list of similar sequences clearly the first thing we have to analyse if the show any kind of periodocity, in that case we could try to approximate the next time that trend will appear. We can verify if they are periodical by calculating the mean value of the interval between them and then see if it is actually similar to any of the interval by summing that amount of time after the timme of a Sequence we have and check if it is even close to another one, if you have a decent result in terms of percentual success you can state that it is periodical, finally one tool you can use to analyse this datas is the Sd.
Take care, you might as well have in your list events called Outliers that'd make your periodicity's calculations useless, so you might want not to consider them in your Model, you'll be able to find out this values by analysing better the sequence for example making your program skip one or some of those at the time and escluding the values who are furthest from the difference of time by first sequence and the third (where for example the seccond one is the deleted value), a value can be considered periodic even if it is not completely accurate, but the set of the different differences from the prevision and the actual apparencies of the sequences (always in terms of time) still need to have (the set) a decent Sd.
We know it might not seem Fundamental Analisys but probably it could be linked to something external that we still don't know, but we can easly find out, and here's how.
The second thing to do is to now is finding out if the similar sequences or periodicities are in any way related to similar facts, this can be simply done (even if not in the most effective way) by automating a research about the title or company name, filtering this result with the date of each sequence you can obtain the facts you or your program are looking for, if similar facts accour for other similar sequencies than with the time you can establish a link between the two, in case you'd want this done by a computer you can see if particoular strings (that need to be part in a vocabolary you store for that title otherwise they might be random strings) appear for each sequence doing a research the automated research we suggested before.
Note very well that you will have to keep what we just said because you'll need it as an introduction to one of our next posts.
Clearly aafter a while you will able to see in real time if a tipe of fact reguarding the title generates some kind of sequence, after a while then, even if is a new fact or a new sequence you will have a big enough library of facts related to sequencies F-S that will help you, or your program understand better theese relationships.
It was a lot for today, if you have any question contact us, hopefully you learnt something and if you have any correction to make please let us know.
See you next post, StatsApplied.