Like I mentioned earlier - if I can make it "clickable" (have a GUI) that would be the best. But if the code or script is somewhat easy to modify, that works too. While I agree with the saying in general, I don't quite understand what you are referring to. "Garbage in" as in paying little for it and I should not expect it to work / or the output not to be trusted? As long as the data is good, I would think it would be relatively easy to cross-check the output from the program to see that it's correct?
I was referring to the post where it was suggested something can be programmed, tested, and documented in 15 minutes to accomplish what you want. That is nonsense and you would get what you pay for, trash. Coding and testing alone just for the examples you posed would take almost a day. And I am talking of quality code and rigorous tests that validate the results as intended. In 15 minutes I write code to read 2 time series from yahoo, store in dataframe and run the correlation between the two time series, including testing. The problem with a lot of CS self proclaimed hotshots is that they rush something, polish a little on the outside to make it shine but the inside is broken and garbage. Especially in trading accuracy and attention is extremely important. The problem with R and Python is that it makes people blindly rely on packages. Many those packages, such as yfinance, have tons of issues and quirks that one must perfectly understand before using them. Just pulling a time series from yahoo's api does not make the data correct or clean or adjusted for whatever the user needs. In many cases the package does not even behave as intended. For example there are tons of non documented and outright problematic issues with the yfinance package and how it deals with aligning timestamps when requesting multiple tickers at once.
It is actually that simple. But we are not talking gui or any fancy solution here. Just a script or notebook you run and it spits out some shares. Gui, fields for settings and so forth is a whole different ballgame. Not a lot more complicated, but will take more time. Then you probably have to pay the fivverr guy $50 or so
Why don't you take that job, it takes 15 minutes as you claimed, right? That extrapolates to an hourly salary of 200. (it does not and you are completely simplifying all of this, while not complex, quality takes more than 15 minutes, I am willing to bet that if I gave you the name of a few stocks you would completely fuck up without even knowing it, just because your awesome yahoo finance python package did not do what you expected it to do)
Just ignore him. He is mostly here to argue. My experience is that the Yahoo data is mostly good. However, obviously if you are blindly going to follow the script and put a fortune on what it tells you, you need to cross-check, confirm and have a completely different level of code quality than some random guy at Fivverr. But I assume your goal is to just filter for further research before you trade?
I am mostly here to set facts straight and be truthful and logical. Your claim of 15 minutes is simply speaking a joke. As you said in any endeavor where money is put on the line 15 minutes never cut it. Thanks for correcting yourself.
Of course. But yfinance, the python package for yahoo finance, often used, does that under the hood, unfortunately sometimes incorrectly. The "issues" section on github does contain numerous posts to detail the potential pitfalls.
Don't forget survivorship bias. You've asked for a backtest involving stocks in the S&P 500. Today's S&P 500 is not the same as it was 1,2,3,4,5 years ago etc! e.g. Consider Monsanto that delisted June 2018 (takeover by Bayer) - it was a member of the S&P 500. Do you even have this data in your system? All of a sudden this system gets way more complex....