In particular, the use of quantitative trading
Reply content:
The country is not very clear, only said the United States.
R vs. SAS
America's big financial institutions use SAS almost all of the previous articles. The most important reason is to use SAS to save due diligence. As a freeware, R has a large number of third-party libraries or packages. It's convenient for you to say you use it, but every time you use it you have to give it a due diligence. Otherwise the model validation in the Bank team,internal audit or the Fed will come back to pick up the trouble. Or you don't have to, but you have to write all the tools you need to yourself is tired, not to mention that they also need to write validate. Whether due diligence on third-party code, or the validation of your code, is expensive, involves many departments, and needs to follow the detailed process. (The Fed's demands are getting tougher)
The inverse of SAS, as a default tool, whether it is using base SAS or SAS EG, the intersection of money is good. For these organizations, SAS is really not expensive. We have recently helped a large line of Bulid Model,team to dozens of people, the customer did not care whether everyone needs, to each with a base SAS and SAS EG.
In addition, the bank is prone to hundreds of G-size data, R run up may be really too hard. The opposite of server-based SAS is very fluid.
Python vs. C + +
I don't know much about this. But Python's application in machine learning is getting better, so there should be a prospect in quantitative trading. And Python's docking with other languages is great, and it's his advantage. On the other hand, Python is really concise, even I see C + + on the headache of people can write some simulation. If the requirements are not very strict, many people will be very happy to accept python. Laxatives!! I'm so excited, wow ...
And no one answered ~ ~ ~ So I'll pretend to be a little bit. Hey ...
====================================================================
My point is: These two languages are useful in quantifying investments, but not enough ... In other places in the financial world, it is grossly inadequate.
First of all, the quantitative investment, need a lot of code practice place is mainly: model development, Model practice, Model revision
1. Quantitative model development, mainly including parameter determination, model backtesting, and pre-test
This part of the work needs to do a lot of statistical work, you want to do a neural network, or SVM, or Bayesian learning, and so on, even if simple as linear regression, in the final analysis is statistical work, a good statistical software can make your productivity greatly increased, right? ~~
As a statistical software, R's main opponents are: SAS, MATLAB, STATA, EVIEWS, and Oxmatrix and so on (my level is poor, probably know these kinds). Foolish thought Stata, eviews two software in the statistical ability has not been able to catch up R, the former two also rarely a spell. But r free Ah ~ ~ This is how valuable quality!
Back to the test also often write some script class things. Python will be very useful at this time. ~~r is also available, but Python is relatively handy. Matlab is also very good (I just love this software), but with genuine expensive AH (with piracy not worth advocating ha), Python or free. But it also has to do with your application scenario. For example, if you are modeling in Excel, then VBA will be very useful and better than python. That's why Python is not enough.
Another aspect of the lack of use is that you also need to manage the data, so you need to know the knowledge of the database and the corresponding language.
When the model is in front of the test, it is probably the "trial operation" stage. In addition to statistical analysis, you have to do a transaction record. If you are doing UHF trading, Python is definitely not enough. If not, there's nothing wrong with C + +, right? ...... I'm here / http quantnet.com
Asked, in addition to high-frequency trading, why do you stress speed? The man who answered was arrogant, didn't say any special reason, only said: I just want to hurry (why he is so hung) ... Still, it seems that "faster" is a very important reason.
There is also a limitation, that is, you use R/python corresponding data volume is not too large, the amount of data is too large r easy to hold, the main reason is R memory management is not the best. But most people don't really have that amount of data ... Note ha, is not enough, not to use HA ~ ~
2. Model practice, that is real thing trading ~ ~ This section see the previous test.
3. Model Correction ~ ~ Personal feel and development of the work content is similar (poor level, everyone forgive), so, see the first part.
=========================================================================
What about in the financial world?
In the financial world, it's a pretty big category. You say in the four big banks do backstage management, do account information management calculate? Development of the bank's app calculation, do not calculate the page management? These posts, in fact, I do not think how to use R python ...
Of course, many people ask, these are not all outsourced? Of course, the bank will buy a lot of professional software, but the bank also has to raise a group of their own technical staff. I can't say what I'm going to do, but I know a computer graduate student at our School of information, and when he talks with him, he mentions that some of his classmates were recruited to the backstage by the Bank of China, usually to write quite a lot of code ...
Well, I'm sorry for my ignorance. But R Python is not serious enough. This conclusion should be right.
